Role of grinding method on granular activated carbon characteristics

Abstract

A coconut shell (AC1230CX) and a bituminous coal based (F400) granular activated carbon (GAC) were ground with mortar and pestle (MP), a blender, and a bench-scale ball milling unit (BMU). Blender was the most time-efficient for particle size reduction. Four size fractions ranging from 20 × 40 to 200 × 325 were characterized along with the bulk GACs. Compared to bulk GACs, F400 blender and BMU 20 × 40 fractions decreased in specific surface area (SSA, −23% and −31%, respectively) while smaller variations (−14% to 5%) occurred randomly for AC1230CX ground fractions. For F400, the blender and BMU size fraction dependencies were attributed to the combination of (i) radial trends in the F400 particle properties and (ii) importance of shear (outer layer removal) versus shock (particle fracturing) size reduction mechanisms. Compared to bulk GACs, surface oxygen content (At%-O1s) increased up to 34% for the F400 blender and BMU 20 × 40 fractions, whereas all AC1230CX ground fractions, except for the blender 100 × 200 and BMU 60 × 100 and 100 × 200 fractions, showed 25–29% consistent increases. The At%-O1s gain was attributed to (i) radial trends in F400 properties and (ii) oxidization during grinding, both of which supported the shear mechanism of mechanical grinding. Relatively small to insignificant changes in point of zero charge (pH_PZC) and crystalline structure showed similar trends with the changes in SSA and At%-O1s. The study findings provide guidance for informed selection of grinding methods based on GAC type and target particle sizes to improve the representativeness of adsorption studies conducted with ground GAC, such as rapid small-scale column tests. When GACs have radial trends in their properties and when the target size fraction only includes larger particle sizes, manual grinding is recommended.

1. Introduction

Granular activated carbon (GAC) is commonly used in water treatment to adsorb dissolved organic contaminants and metallic ions that cannot be effectively removed by conventional physical, chemical, or biological water treatment processes [16,45]. The adsorption process is affected by the physical and chemical properties of both the adsorbent (e.g., GAC) and adsorbate (i.e., target compound for removal) as well as the water matrix composition [16]. Furthermore, GAC adsorption in water treatment usually occurs in fixed-bed adsorbers (FBA) where the adsorbate removal efficiency depends on various operating conditions [66]. Therefore, to design and optimize a GAC treatment system, it is necessary to conduct adsorption tests with the selected GAC and target compound(s) under representative conditions.

Batch kinetic and isotherm experiments evaluate adsorption kinetics and equilibrium adsorption capacity, respectively [66]. Because of kinetic limitations due to intraparticle diffusion (i.e., pore and/or surface diffusion), GAC adsorption of some water contaminants can take years to reach equilibrium [57]; therefore, GAC is often ground to small particle sizes to shorten equilibrium time by reducing diffusion pathlengths (i.e., particle radius). Adsorption kinetics and capacity are the basis for selecting the best performing GAC and for determining the dimensions and operating conditions of the FBA for a specific treatment purpose [66].

To determine optimal operating conditions for the FBA, dynamic evaluation of FBA treatment efficiency is conducted in a flow through system. A commonly used bench-scale evaluation tool is a rapid small-scale column test (RSSCT) [66]. By using ground GAC, RSSCTs simulate full- or pilot-scale FBA performance using only a fraction of the originally needed time and resources. To scale down pilot- or full-scale FBAs, RSSCTs use ground GAC whose mean particle diameters (d_p, usually < 110 μm) are determined using a known scaling procedure [18].

A fundamental assumption of using ground GAC in batch experiments and RSSCTs is that grinding does not affect GAC properties. Despite some concerns about possible changes in accessible surface area, little attention has been devoted to investigating the effects of grinding on GAC characteristics that are potentially essential for its adsorption capacity. Several studies found no substantial change in GAC adsorption capacity for phenol [40], o-chlorophenol [55], and humic substances [57] after grinding to smaller sizes (0.45 μm < d_p < 3,524 μm). However, Yu et al. [77] reported an increase in adsorption capacity of a ground (d_p < 100 μm) coal-based GAC towards perfluorooctane sulfonic acid (PFOS) and perfluorooctanoic acid (PFOA) compared to its bulk size (100 μm < d_p < 900 μm).

Depending on GAC type, adsorbate, and solution characteristics, various surface properties can be important for the adsorption process and may or may not affect the adsorption capacity. For a given adsorbate, adsorption onto GAC surfaces may be controlled by (i) physical interactions (e.g., microporosity effects and molecular size exclusion) that are mainly functions of the GAC accessible surface area and pore size distribution (PSD) [25,35] and (ii) chemical interactions (e.g., hydrophobic or electrostatic effects) that are functions of surface functional groups, elemental composition, and the point of zero charge (pH_PZC) [33,60,78]. Additionally, Alfarra et al. [2] suggested the concept of hard and soft acids and bases, where various metallic ions are trapped either in the soft zones in the flat aromatic sheets or in the hard zones on the edge of the sheets. Furthermore, grinding may affect different types of GACs differently as GAC surface and structural properties are influenced by its precursor material and activation process [45]. Utilities often use RSSCTs to compare the predicted full-scale treatment efficacies of different GACs and therefore, it is crucial to systematically evaluate and minimize the potential grinding effects on the surface properties of different GACs, to ensure the representativeness of RSSCT results.

Another often undiscussed topic is how GAC is ground. With no standardized procedure, the most mentioned method is careful crushing with mortar and pestle (MP) [17,34,54,77]. MP is time-consuming, labor-intensive and poses the risk of carbon dust inhalation. Wet sieving is an option to avoid carbon dust when collecting between two sieves but cannot be used when collecting all the particles passing through one sieve is desired [5,62]. Therefore, some studies opted for use of mechanical tools such as ball milling units (BMUs) [28,36,48] and blenders or coffee grinders [4,30,72]. Such arbitrary selection of grinding methods could result in variability in the ground GAC, adding further uncertainty to the sustainment of the bulk GAC properties and further challenging the validity of the assumption that grinding does not affect GAC adsorption capacity. To date, only a few studies have experimentally investigated the impact of grinding, reporting some changes in surface characteristics such as a noticeable decrease or increase in specific surface area (SSA) when using MP or a blender [39,42,53] and external surface oxidization and a substantial shift in PSD when using a BMU [52].

Therefore, the current research aims to study three commonly used grinding methods (MP, BMU, and commercial blender) to (i) evaluate and compare the particle-size-reduction effectiveness of the three methods, (ii) provide a systematic experimental investigation of the impact of the three grinding methods on various GAC properties, and (iii) investigate the dependency of the grinding impact on GAC type and particle size.

2. Materials and methods

2.1. Materials

A coconut shell-based GAC (CS-GAC) – AquaCarb 1230CX (AC1230CX) (Evoqua Water Technologies; Pittsburgh, PA) and a bituminous coal-based GAC (BC-GAC) – Filtrasorb 400 (F400) (Calgon Carbon; Pittsburgh, PA) were evaluated. F400 is widely used in water treatment and AC1230CX was chosen a representative of CS-GACs that are marketed as sustainable alternatives to BC-GACs. Manufacturer provided properties are included in the Supplementary Information (SI) Table S1.

2.2. Grinding methods

Three griding methods were evaluated: (i) MP, (ii) a Waring Xtreme blender, and (iii) a bench-scale BMU. The BMU was a 1-L stainless-steel jar, containing stainless-steel balls of different sizes, that fits into rotary agitators. The top ball size (B_top = 0.98 in = 25 mm) was estimated (Eq. (1)) based on Fred Bond Theory [10].

$$B_{top}=\sqrt{\frac{F{W}_i}{K{C}_s}\sqrt{s/\sqrt{D}}}$$ (1)

In Eq. (1),

B_top – top grinding ball diameter, in;

K – 200 for balls and when using inches for B_top and feet for D;

F – the grain size where 80% of the fresh feed passes (1,680 μm);

S – specific gravity of the feed (0.445 for AC1230CX, 0.540 for F400);

W_i – Bond work index of coal at the feed size F, 13 per Weiss [75];

C_S – Percent of mill critical rotational speed, 75 per Don and Robert [24];

D – mill diameter, 0.564 ft.

Per Cho et al. [15], the second ball size (B_second = 0.59 in = 15 mm) was set at approximately half of B_top, and the make-up of grinding balls was set to 29% B_top and 71% B_second. Per Perry’s Chemical Engineer’s Handbook [24], the total number of balls (20 of B_top and 50 of B_second) was determined to occupy about 25% of the jar volume, and the GAC (i.e., the feed) was loaded into the BMU until the ball voids were filled. As a result, the BMU accommodated either 100 g of AC1230CX or 140 g of F400, and the amounts ground with the other two methods were set to the same masses used in the BMU.

Bulk AC1230CX and F400 samples were each ground using three grinding methods (Fig. 1). Particle size distributions of the ground products were investigated with sieve analysis containing six size fractions: 20 × 40, 40 × 60, 60 × 100, 100 × 200, 200 × 325, and below 325 mesh size sieves. Three of the size fractions (20 × 40, 60 × 100, and 100 × 200) were selected for characterization along with the bulk GAC (Fig. 1) to investigate (i) how grinding impacted GAC surface properties and (ii) if grinding impacts depended on particle size. An additional 200 × 325 ground fraction and a 20 × 40 fraction obtained by directly sieving the bulk GAC (i.e., without grinding) were characterized for SSA and PSD due to appreciable variation (see Section 3.2.1) among the originally chosen three size fractions.

Fig. 1.

Experimental layout overview.

Note: 20×40 fraction for surface area and pore size distribution analysis also include sieved only 20×40 fraction, as listed in Table 1.

2.3. Characterization methods

Fig. 1 summarizes investigated surface properties and corresponding characterization methods. Nitrogen (N₂) physisorption tests were performed at 77 K using a Quantachrome NOVA- 2000e surface area and pore size analyzer. The Brunauer–Emmet–Teller (BET) equation [8] was used to calculate the SSA. The PSD was obtained as follows: (i) the Quenched Solid Density Functional Theory (QSDFT) [49,58] method in Quantachrome’s data reduction software was used to calculate the distribution of micropores ( <2 nm) and mesopores (2–50 nm) [65], (ii) total pore volume was calculated from the volume of adsorbed nitrogen at the near saturation relative pressure point (P/P₀ = 0.985), and (iii) macropore (>50 nm) volume was determined by subtracting the micro-and mesopore volumes from the total pore volume [19].

The surface morphology was characterized by environmental scanning electron microscopy (SEM, Philips XL 30 ESEM-FEG). Surface crystallite structure was examined by X-ray diffraction analyzer (XRD, PAN-analytical Xpert diffractometer) at λ = 1.5406 Å using a Cu Kα radiation source. The transmission electron microscopy (TEM) examination was performed with a JEOL (JEM-2010 F) high resolution-transmittance electron microscope (HR-TEM) operated at a 200 kV field. The surface elemental composition was characterized by a Kratos Axis Ultra X-ray photoelectron spectroscopy (XPS) system fitted with a Al Kα X-ray radiation source. The binding energies of the machine were standardized by C1s (electrons emitted from the 1 s orbital of a carbon atom) level at 284.6 eV as a reference and calibrated using Ag standard.

Point of zero charge (pH_PZC, the pH at which the total surface charge of the carbon particles is zero) was determined using the pH drift method [19,43]. Briefly, for each experiment, 100 mg of GAC was added into 20 ml of 0.1 M NaCl solutions (in Type 1 water, resistivity = 18.2 MΩ·cm) in 25 ml vials, and the initial pH was incrementally adjusted from 2 to 12. The vials were tightly sealed and agitated on an orbital shaker (200 rpm) for 48 h before the final pH measurement. The 0.1 M NaCl solution was made carbonate free by boiling and cooling down under argon flow. In addition, the entire process was carried out under argon flow (50 ml/min) to minimize ambient air carbon dioxide exchange that would impact solution pH. All samples, including blanks for each pH point, were prepared in triplicate for quality control. The pH at which no drift occurred after 48 h was defined as the pH_PZC.

3. Results and discussion

3.1. Particle size distribution after grinding

The common criterion for all three grinding methods was that grinding continued until more than 99% of the 100 g F400 or 140 g AC1230CX sample passed through the #20 sieve (i.e., largest sieve). Such a criterion ensured that the ground samples represented the bulk GAC sample and varying particle sizes between the #20 (840 μm) and #325 (40 μm) sieves were generated for characterization. The grinding time to reach the criterion was one hour for MP or BMU and 30 s for the blender. The resulting six ground products showed different particle size distributions (Fig. 2). Calculation of percent masses in each size fraction is based on the total mass after grinding to account for some minor losses (<4.4%) during grinding and sieving. Manual grinding with MP was the least efficient in size reduction as the D₅₀ (d_p below which 50% the particles were collected) was 420 μm (F400) and 480 μm (AC1230CX) for MP, 180 μm (F400) and 280 μm (AC1230CX) for the blender, and 140 μm (F400) and 190 μm (AC1230CX) for the BMU. The blender produced comparable results to the BMU with only 30 s of grinding time, proving to be the most time-efficient method for size reduction. RSSCT studies often use the expected log-mean diameters of ground particles collected between two sieves in scaling equations. Based on the different particle size distributions of ground GACs produced by the three methods, it is possible that the actual log-mean diameter between two sieves could differ from the expected log-mean diameter to some degree, depending on how the sorbent was ground.

Fig. 2.

Particle size distribution of AquaCarb 1230CX (AC1230CX) and Filtrasorb 400 (F400) produced from three grinding methods: mortar and pestle (MP), blender, and ball milling unit (BMU).

Grinding AC1230CX (Fig. 2, filled symbols) consistently yielded lesser size reduction compared to grinding F400 (Fig. 2, open symbols), especially with the blender and BMU (up to 19%). Properties of the two GACs and the physical principles of the three grinding processes may offer a plausible explanation. Fig. 3c, d, and e display a schematic representation of the three grinding methods. GAC in the BMU (Fig. 3e) is mainly influenced by gravity, centrifugal, and frictional forces. The grinding balls and feed material travel up along the jar wall then dropdown. As a result, the GAC particles that are caught in between the jar walls and the balls or between two balls are impacted by both shock (vertical, fracturing) and shear (tangential, abrasion) forces [44,51]. Similarly, in a blender (Fig. 3d), depending on the point and direction of impact exerted by the blender blades or jar walls, shock and shear forces impact the GAC particles. In contrast, when crushing with MP (Fig. 3c) and even though the twist of the pestle can exert abrasion to a degree on particles (depending on the operator), the two relatively large surfaces (the rounded end of the pestle and the bottom of the mortar) mainly exert vertical shock forces that mainly fracture the particles (Fig. 3f) as opposed to abrasive shear forces (Fig. 3g). Therefore, the greater abrasion number, a percent reduction in mean d_P after particles rubbing together or particles rubbing against another surface [6], of AC1230CX (85, Table S1) than F400 (75, Table S1) is more evidently reflected during the blender and the BMU grinding.

Fig. 3.

Schematic representation of (a) Filtrasorb 400 (F400) and (b) AquaCarb 1230CX (AC1230CX) being ground by three methods: (c) mortar and pestle, (d) blender, and (e) ball milling unit. The grinding methods reduce particle size by two mechanisms: (f) shock and (g) shear forces. The ground particles are sieved into various size fractions (h). Note: (1) gray regions represent outer layers, and black regions represent particle core; (2) Fig. 3c, represents the movement of the mortar during grinding in the current study. The movement can be different depending on the operator.

3.2. GAC physical properties

3.2.1. Surface area and pore size distribution

Surface area is a major factor in determining GAC adsorption capacity because the large internal surface area created by the internal pores provides almost all of its adsorption capacity [37,76]. Table 1 shows the SSA and PSD of both AC1230CX and F400 before and after grinding. The effect of grinding on SSA varied depending on the GAC type, size fraction, and the grinding method.

Table 1.

Specific surface area (SSA), pore volume, and pore size distribution (PSD) analysis of two granular activated carbons (GAC) before and after grinding with three grinding methods.

GAC	Grinding Method	Size Fraction	Mean Particle Diameter dp (μm)	SSA (m2 /g)	SSA% change from Bulk GAC	Total Pore Volume (cm3 /g)	% micro-pore (<2 nm)	% meso-pore (2–50 nm)	% macro-pore (> 50 nm)
AquaCarb 1230CX	Bulk GAC	12×30	1130	1050	NA	0.58	77.75	13.99	8.26
	Sieving only	20×40	630	1152	9.64	0.64	72.99	18.86	8.15
	Mortar and Pestle	20×40	630	1046	−0.45	0.59	74.34	17.17	8.49
		60×100	200	1062	1.13	0.60	75.49	16.14	8.38
		100×200	110	1090	3.79	0.60	77.59	14.35	8.07
		200×325	60	1057	0.65	0.61	73.19	18.37	8.44
	Blender	20×40	630	1082	2.99	0.61	74.74	16.77	8.48
		60×100	200	1103	5.01	0.61	75.11	16.63	8.26
		100×200	110	1054	0.39	0.59	74.84	17.01	8.14
		200×325	60	903	−14.01	0.51	74.83	16.83	8.34
	Ball milling unit	20×40	630	922	−12.25	0.50	81.28	11.31	7.41
		60×100	200	1107	5.40	0.61	76.29	15.31	8.40
		100×200	110	1081	2.94	0.60	75.27	16.48	8.25
		200×325	60	952	−9.39	0.53	76.14	15.68	8.18
Filtrasorb 400	Bulk GAC	12×40	1050	769	NA	0.50	62.58	28.64	8.78
	Sieving only	20×40	630	889	15.67	0.58	59.29	32.36	8.35
	Mortar and Pestle	20×40	630	764	−0.59	0.48	65.05	26.98	7.97
		60×100	200	867	12.82	0.56	64.91	26.75	8.35
		100×200	110	832	8.19	0.54	62.22	27.66	10.12
		200×325	60	833	8.42	0.53	63.37	28.86	7.76
	Blender	20×40	630	596	−22.52	0.41	62.88	28.04	9.09
		60×100	200	748	−2.72	0.48	63.36	28.27	8.37
		100×200	110	826	7.43	0.53	63.82	27.78	8.40
		200×325	60	776	0.99	0.50	62.96	28.89	8.15
	Ball milling unit	20×40	630	528	−31.35	0.34	66.67	24.46	8.88
		60×100	200	734	−4.48	0.48	63.55	28.19	8.26
		100×200	110	805	4.75	0.52	62.73	28.84	8.43
		200×325	60	816	6.12	0.52	62.45	29.16	8.40

Note: NA = not applicable.

Compared to the bulk F400, the SSA decrease in the F400 20 × 40 fraction was considerable for the blender (−23%) and the BMU (−31%) but negligible for the MP (−0.6%). For the F400 100 × 200 fractions, relatively smaller percent increases from the bulk F400 were similar for all grinding methods (5 to 8%). Due to this size-fraction-dependent variation, an additional F400 fraction (200 × 325) was characterized, and the SSA percent difference from the bulk F400 was 8% for MP, 1% for the blender, and 6% for the BMU, showing a similar variation with the 100 × 200 fraction. Patni et al. [53] observed a similar SSA decrease from bulk GAC (around 40%) at d_p = 550 μm (comparable to the 20 × 40 fraction; mean d_P = 600 μm) with two BC-GAC (Filtrasorb 400 and 600) ground with a blender.

The contrasting SSA changes in the F400 20 × 40 fraction produced with MP versus the two mechanical methods are, again, explained by a combination of GAC properties and grinding mechanisms. Buczek [13] used abrasion to gradually remove the external layers of a coal-based, steam-activated GAC which is similar to F400 in terms of the precursor material and activation process [41]. Buczek [13] observed increasing particle density and decreasing ash content towards the particle center (schematically represented in Fig. 3a), indicating a decrease in the percent burn-off (BO%) of carbonaceous materials. BO% influences the formation of pores and surface area during activation. Generally, for a given precursor material and activation procedure, a greater BO% (usually < 80% by weight [wt%]) results in a greater pore volume and surface area per unit weight of GAC [9,14,20,45]. Therefore, in the blender or BMU, smaller size fractions (e.g., 100 × 200 fraction) may have contained more particles that resulted from the abrasive removal of outer F400 particle layers (Fig. 3g and h, gray regions) with a relatively greater BO%, pore volume, and SSA compared to bulk F400, leaving the larger F400 size fractions (e.g., 20 × 40 fraction) to mostly contain the inner F400 particle layers (Fig. 3g and h, black regions) with a relatively less BO%, pore volume, and SSA compared to bulk F400. The absence of an SSA decrease in the F400 20 × 40 sieved only fraction relative to bulk F400 (Table 1, Sieving only) supports the proposed hypothesis because the particles in the sieved only F400 20 × 40 fraction were not subjected to any potential removal of outer layers imposed by grinding. Furthermore, being at the smaller end of the bulk F400 particle size range (12 × 40), their shorter radius facilitates greater overall BO% compared to larger particles and results in a greater SSA (16%, Table 1) relative to bulk F400. Radial pore volume decreases towards the particle core have also been observed with a peat-based GAC (Norit R3ex) and a hard coal and wood char mix based GAC where core samples (outer 67 wt% removed) adsorbed up to 50% less nitrogen or benzene compared to the outer 33 wt% [12,22,23]. Considering the reductions in SSA from the blender and BMU at the larger particle size range (e.g., 20 × 40), MP is recommended for obtaining ground F400 at such sizes.

For AC1230CX, the SSA percent change of each size fraction compared to the bulk AC1230CX was between −0.5% and 4% for MP size fractions, −14% and 5% for blender size fractions (greatest deviation for 200 × 325 fraction), and −12% and 5% for BMU size fractions (greatest deviation for 20 × 40 fraction). Unlike F400, there was no consistent trend in the SSA percent change with size fraction or grinding method. Two possible explanations are: (i) AC1230CX was resistant to external layer removal and possibly mainly fractured during all grinding methods due to its greater abrasion number, and (ii) the development of the surface area and pore structure is heavily influenced by the precursor material and AC1230CX may not possess a radial trend in BO%. The two possibilities will be revisited in the later sections.

For both GACs, total pore volume and SSA were positively, linear correlated (0.98 R² for F400; 0.97 R² for AC1230CX; Table 1 and Figure S1). This is expected as the internal pores provide GAC’s large internal surface area [37,76]. Regardless of grinding method, the PSD was consistent for both F400 (micropore 63.19 ± 1.57%, mesopore 28.26 ± 1.60%, macropore 8.55 ± 0.54%) and AC1230CX (micropore 75.96 ± 2.09%, mesopore 15.80 ± 1.91%, macropore 8.23 ± 0.25%), indicating negligible effects of different grinding techniques on PSD. The PSD consistency supports that the observed decrease in the F400 blender and BMU 20 × 40 fractions are not due to potential pore collapse or pore enlargement, which would have resulted in a significant shift in PSD, but due to the difference in total porosity in different layers of F400 towards particle core. The greater percentage of micropores in AC1230CX helps create more surface area, and therefore, for a given total pore volume, AC1230CX has greater SSA than F400 (Figure S1).

3.2.2. Surface morphology

To further investigate the observed changes in SSA, SEM micrographs of the bulk F400 and the F400 20 × 40 MP, blender, and BMU fractions were compared (Fig. 4). SEM images for the remaining F400 size fractions are included in the SI (Figures S2 and S3). Images were taken at lower magnification (Fig. 4a, c, e, and g) to display the shape, size, and outer surface morphology of the particles. There was no obvious difference in particle morphology among the F400 20 × 40 (420–841 μm) fractions (Fig. 4c, 4e, and 4g). All the observed ground particles as well as the bulk F400 (Fig. 4a) featured a rough outer surface with ridges, as reported in previous studies [70,74]. The images taken at greater magnification (Fig. 4b, d, f, and h) show a closer view of the surface ridges and crevasses with some visible submicron (50 nm–1 μm) macropores. Micropores (<2 nm) and mesopores (2–50 nm) require greater magnification to be visible. Regardless of the grinding method, all ground F400 samples (Fig. 4c–4h) showed a similar surface structure with the bulk F400 (Fig. 4a and 4b). Therefore, at the SEM magnified levels (1 μm), the three grinding methods did not introduce visible alterations to the F400 GAC surface and pore structure, such as collapsing or creation of visible large pores. The consistency among the SEM images of bulk and ground F400, together with the little variation in the PSD among the bulk, sieved only, and ground GACs (Table 1), supports the explanation that the changes in the SSA of F400 was due to the difference in the overall particle porosity in the blender and the BMU ground fractions that reflected inherent radial trends in the F400 particles as previously discussed (Sections 3.1 and 3.2.1).

Fig. 4.

Scanning electron microscopy micrographs of Filtrasorb 400: bulk sample (a) and (b), and 20×40 fraction obtained by grinding with mortar and pestle (c) and (d), blender (e) and (f), or ball milling unit (g) and (h).

In a similar manner to F400, the shape and surface structure of ground AC1230CX did not vary among different grinding methods (Figures S4–S6). In contrast to F400, the AC1230CX surface showed some well-developed large macropores (> 1 μm) and smaller submicron (50 nm–1 μm) macropores. Note that macropores (>50 nm) covers a wide range, and many smaller macropores cannot be seen even in the high magnification SEM micrographs. Therefore, even though F400 and AC1230CX had similar macropore percentage (Section 3.2.1), F400 micrographs show less macropores. CS-GAC typically have larger macropores due to their honeycomb structure (Figure S6c, S6e, S6g), originating from the cylinder-like tube structure that runs through raw coconut shells [31] and cross-linking of the cylinders by their reactive points during carbonization and activation [1,56,68]. The resulting channels that run through the granules (schematically represented in Fig. 3b) potentially facilitate a more homogenous BO%, supporting that AC1230CX may not possess a notable radial trend in BO% (Section 3.2.1). In addition, although adsorption mainly occurs in micropores and/or mesopores for most adsorbates, macropores are important conduits that provide access to the interior mesopores and micropores [45,59]. Depending on the molecular characteristics and adsorption mechanisms of the adsorbate, larger pores can benefit adsorption kinetics and capacity. For example, studies have attributed the lower poly- and perfluoroalkyl substances (PFAS) removal efficiencies of CS-GAC compared to BC-GAC to the more microporous (tighter) structure of CS-GAC that possibly imposed kinetic limitations [3,46,63]. Similarly, bamboo-derived GAC had greater perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA) adsorption capacity than BC-GAC due to its enlarged pore structure that aided faster diffusion [21]. The CS-GAC in the current study (AC1230CX) has a greater microporosity (75.96 ± 2.09%) than F400 (63.19 ± 1.57%) and is an enhanced product with modified pore structure and superior mesopores compared to historically produced CS-GACs, aiming to improve its adsorption performance in applications where BC-GAC has been preferred [26]. A recent comparative study reported nearly equal PFOS and PFOA removal efficiencies of AC1230CX and a BC-GAC, Fitrasorb 600 [32].

3.3. GAC chemical properties

3.3.1. pH_PZC

The effect of pH_PZC on adsorption can be significant when the adsorption driving force is the electrostatic interactions between the GAC surface and target contaminants that are charged at drinking water relevant pH (e.g., pH 6 to 9), such as natural organic matter (NOM) [33,50] and many PFAS, including PFOS and PFOA [60,78]. Table 2 summarizes the pH_PZC values of all size fractions. Both AC1230CX and F400 had pH_PZC >10, which is advantageous for the adsorption of various anionic contaminants (e.g., PFAS, NOM) as the GAC surface would be positively charged at drinking water relevant pHs. Despite the subtle changes in ground GAC pH_PZC compared to the bulk GACs (−6% to 4%), a Tukey’s multiple comparison test [11] revealed some statistically significant variations when all size fraction and grinding method combinations were compared for each GAC (Figure S7). For F400, significant pH_PZC differences only occurred in blender and BMU 20 × 40 fractions, which were the same F400 fractions that exhibited an SSA decrease (Section 3.2.1). For AC1230CX, significant pH_PZC differences randomly occurred, exhibiting no trend with grinding method or size fraction.

Table 2.

Point of zero charge (pH_PZC) analysis for two granular activated carbons (GAC) before and after grinding with three methods.

GAC	Grinding Method	Size Fraction	Mean Particle Diameter dP (μm)	pHPZC	pHPZC std. deviation	pHPZC %change from Bulk GAC
AquaCarb 1230CX	Bulk GAC	12×30	1130	11.57	0.03	NA
	Mortar and Pestle	20×40	630	11.81	0.22	2.04
		60×100	200	12.00	0.10	3.69
		100×200	110	11.79	0.30	1.86
	Blender	20×40	630	11.49	0.07	−0.73
		60×100	200	11.55	0.05	−0.22
		100×200	110	11.38	0.09	−1.67
	Ball Milling Unit	20×40	630	11.47	0.02	−0.93
		60×100	200	11.30	0.04	−2.40
		100×200	110	11.33	0.02	−2.06
Filtrasorb 400	Bulk GAC	12×40	1050	11.25	0.09	NA
	Mortar and Pestle	20×40	630	10.92	0.17	−2.85
		60×100	200	10.91	0.30	−2.94
		100×200	110	11.16	0.08	−0.80
	Blender	20×40	630	10.77	0.03	−4.15
		60×100	200	11.07	0.10	−1.55
		100×200	110	11.08	0.07	−1.43
	Ball Milling Unit	20×40	630	10.61	0.09	−5.56
		60×100	200	10.85	0.08	−3.42
		100×200	110	11.02	0.12	−1.95

Note: NA = not applicable.

3.3.2. Surface elemental composition

Given the changes in pH_PZC, which is related to the ratio of surface basicity to acidity [27,78], surface elemental composition of all bulk and ground GAC were investigated because oxygen containing functional groups are often related to surface acidity [19,33,78]. XPS analysis provided atomic percentage concentrations (At%) of C1s (At%-C1s) and O1s (At%-O1s) (Table 3). Compared to bulk F400, F400 20 × 40 fractions from the blender (13%) and BMU (34%) showed the greatest At%-O1s increase. Buczek [13] observed an increase in oxygen complexes contained in the carbonaceous material toward the GAC particle center due to a decreased BO% towards the particle core (i.e., inner layers). Therefore, the increase in At%-O1s for the blender- and the BMU-produced F400 20 × 40 fractions was potentially due to these fractions containing more core portions of the GAC particles compared to bulk F400 from the shear grinding mechanism previously discussed (Section 3.1 and Fig. 3).

Table 3.

Elemental composition of two granular activated carbons (GAC) before and after grinding with three methods.

GAC	Grinding Method	Size Fraction	Mean Particle Diameter dP (μm)	At%-C1s	At%-O1s	At%-O1s%change from Bulk GAC
AquaCarb 1230CX	Bulk GAC	12×30	1130	88.41	11.60	NA
	Mortar and pestle	20×40	630	85.32	14.68	26.56
		60×100	200	85.16	14.84	27.94
		100×200	110	85.05	14.95	28.93
	Blender	20×40	630	85.28	14.72	26.91
		60×100	200	85.51	14.49	24.97
		100×200	110	88.33	11.67	0.65
	BMU	20×40	630	84.99	15.01	29.45
		60×100	200	86.55	13.45	16.00
		100×200	110	89.53	10.47	−9.70
Filtrasorb 400	Bulk GAC	12×40	1050	85.49	14.51	NA
	Mortar and pestle	20×40	630	84.63	15.37	5.96
		60×100	200	84.85	15.15	4.41
		100×200	110	84.58	15.42	6.27
	Blender	20×40	630	83.68	16.32	12.51
		60×100	200	85.30	14.70	1.34
		100×200	110	84.19	15.81	9.00
	BMU	20×40	630	80.56	19.44	34.02
		60×100	200	86.25	13.75	−5.21
		100×200	110	86.88	13.12	−9.55

Note:.

NA = not applicable.

At%-C1s: Atomic percentage concentrations calculated by measuring the kinetic energy of electrons emitted from the 1 s orbital of carbon atoms. At%-O1s: Atomic percentage concentrations calculated by measuring the kinetic energy of electrons emitted from the 1 s orbital of oxygen atoms.

As MP had little preferential removal of outer layers, the relatively small At%-O1s increase in all MP F400 ground fractions (4% to 6%) compared to bulk F400 is potentially due to surface oxidization during grinding, providing a baseline At%-O1s increase, as opposed to the currently observed up to 34% increase in the blender and BMU 20 × 40 fractions, if no radial trend in BO% existed or was reflected in different size fractions. Surface oxidization during grinding occurs rapidly and primarily on the exposed external surfaces [47,52]. Therefore, for the blender and BMU, oxidization during grinding is expected to be less evident in the already exposed (i.e., already oxidized) external layers that are contained in 60 × 100 and 100 × 200 fractions and more evident in the core parts (e.g., 20 × 40 fraction) that are newly exposed. The change in At%-O1s compared to bulk F400 were negative for the BMU F400 60 × 100 (−5%) and 100 × 200 (−10%) fractions likely due to containing more already exposed outer layers with less oxidization potential in addition to already having greater BO% and less oxygen containing carbonaceous material compared to the core and the bulk F400. The blender-produced 60 × 100 (1%) and 100 × 200 (9%) fractions showed At%-O1s increases comparable to MP and greater than the same fractions obtained with the BMU. The greater oxidation for MP and the blender could result from MP being an open system with ample oxygen and the blender being a greater energy and speed grinding process.

Similar size-fraction- and grinding-method-dependent trend in oxidization was also observed in AC1230CX. The At%-O1s percent changes from bulk to ground fractions showed a (i) consistent increase (27–29%) in all MP fractions that was similar to the blender 20 × 40 (27%), blender 60 × 100 (25%), and BMU 20 × 40 (29%) fractions; (ii) lesser increase from the BMU 60 × 100 fraction (16%); and (iii) a slight decrease (−10%) and minimal increase (0.7%) in the BMU and blender 100 × 200 fractions, respectively. The 27–29% AC1230CX MP oxidization is substantially greater than F400 MP (4–6%). CS-GAC generally contains less ash content (1–3 wt%) compared to BC-GAC (6–20 wt%) [45] and therefore, possibly having more oxidization potential. Recalling the two possibilities proposed in Section 3.2.1 and revisited in Section 3.2.2, if AC1230CX was mainly fractured during all grinding processes due to its greater abrasion resistance, the At%-O1s change in the blender- and BMU-produced ground fractions should have been similar to MP. However, in both blender and BMU, the 100 × 200 fractions (potentially containing more outer layers) had the least amount of oxidization, and the 20 × 40 fractions (potentially containing more core parts) had similar At%-O1s increase to MP. Therefore, AC1230CX was also abrased during mechanical griding, and thus, the second possibility of AC1230CX not possessing a radial trend in BO% was further supported.

As mentioned earlier, an increase in oxygen content has been linked to an increase in surface acidity and thus, a possible decrease in pH_PZC [19,27,33,78]. The At%-O1s increases in F400 20 × 40 fractions compared to the bulk F400 corresponded to small but statistically significant decreases in pH_PZC in the same fractions. Only one small but statistically significant pH_PZC increase (4%) in the MP AC1230CX 60 × 100 fraction did not correspond to a At%-O1s decrease in the same fraction, which could be due to experimental errors or due to other unknown changes that possibly increased surface basicity.

3.3.3. Crystalline phase structure

The randomly oriented stacking of the flat aromatic sheets caused by heteroatoms and defects creates the microporous crystalline structure of GAC [7], which is the source of its remarkably high surface area [37]. The lack of high-intensity narrow peaks in the XRD diffraction patterns demonstrate the amorphous nature of GAC. As shown in Fig. 5, the presence of a broad peak at diffraction angle 2Θ = 22.87° and 23.99° (002) and a second weak peak at 2Θ = 43.83° and 43.49° (101) indicate the relation of GAC structure to graphite crystalline structure [29,73]. The small sharp diffraction peak at 26.55° (002) in F400 samples (Fig. 5b, 5d, and 5f) indicate traces of pristine graphite [38,67].

Fig. 5.

X-ray diffraction spectra for four different fractions from two granular activated carbons: (a) AquaCarb 1230CX (AC1230CX) and (b) Filtrasorb 400 (F400) obtained with mortar and pestle, (c) AC1230CX and (d) F400 obtained with the blender, and (e) AC1230CX and (f) F400 obtained with the ball milling unit.

The diffractograms varied, again, depending on the GAC type, size fraction, and the grinding method. In the F400 20 × 40 blender (Fig. 5d) and BMU (Fig. 5f) fractions, the broad 002 peaks (2Θ = 23.99°) were stronger compared to other ground size fractions as well as the bulk F400, possibly indicating an increase in the graphite-like crystalline structure. Such a pattern supports the radial trend in the F400 particle properties, that is, containing more carbonaceous (graphite like) materials towards the particle core due to less BO%. Lee et al. [38] observed a similar change in the 002 peaks where the left side of the peak became increasingly flattened (Fig. 5d and f) as the crystallite fraction (as opposed to the amorphous fraction) decreased with the increase of BO%. The peak shape of the F400 20 × 40 MP fraction (Fig. 5b) was similar across all other ground and bulk sizes. Therefore, the BO% decrease towards particle core that was reflected in the SEM, pH_PZC, and XPS analysis of the F400 20 × 40 blender and BMU fractions was again confirmed in the XRD diffractograms.

The ground AC1230CX diffractograms (Fig. 5a, c, and e) displayed a noticeable broadening of 002 peaks in 60 × 100 and 100 × 200 fractions compared to the bulk AC1230CX across all grinding methods. Similar broadening of a sharp 002 peak has also been observed in graphite after milling due to the creation of lattice defects and changes in the interplanar spacing [64]. During grinding, shearing energy mainly dissipates as thermal energy through friction while shock energies are absorbed by the particles, causing fissures, defects, and internal strains in the crystalline structure that can lead to homogenous or heterogeneous changes in interplanar spacing [61], as shown in Figure S8 (see also SI Section S1. Transmission electron microscopy (TEM) examination of crystalline phase structure). Crystallite size reduction and strain are the two main sources of peak broadening [71]. The crystallite sizes of the AC1230CX 60 × 100 and 100 × 200 fractions were relatively larger than that of the bulk size (Table S2). Therefore, the broadening of peaks in the ground AC1230CX indicates a greater crystalline strain as particle size decreases regardless of the grinding method used. Overall, the variations in the XRD diffractograms of the ground size fractions of F400 and AC1230CX obtained with the three methods, again, reflected the combined effect of grinding mechanisms and GAC inherent properties on the characteristics of ground GACs.

4. Conclusion

• Among the three grinding methods, blender and MP were the most and the least time-efficient for particle size reduction, respectively. The particle size distributions generated by the BMU (D_{50, F400} = 140 μm and D_{50, AC1230CX} = 190 μm) were comparable to the particle size distributions obtained by the blender (D_{50, F400} = 180 μm and D_{50, AC1230CX} = 280 μm), but MP produced bigger particles (D_{50, F400} = 420 μm and D_{50, AC1230CX} = 480 μm).

• The varying percent changes in the investigated properties of ground F400 and AC1230CX obtained with MP, blender, and BMU compared to their corresponding bulk GACs can be attributed to the combined effect of the different particle size reduction mechanisms and the different inherent properties of the GACs. Major findings:

  • In the blender and BMU, both shock and shear forces are at play. Shock forces fracture particles to produce smaller particles of random sizes. Shear forces mainly abrade the particle outer layers into small pieces that fall into smaller size fractions (e.g., 100 × 200) while the sanded-down core parts remain in the bigger fractions (e.g., 20 × 40).
  • For F400, BO% decreases towards the particle core and forms less pore volume and SSA. Less BO% means more carbonaceous material that can potentially contain oxygen complexes, which was reflected in the statistically significant pH_PZC decrease, substantial oxygen content increase, and the strengthening 2Θ = 23.99 peaks in the blender- and BMU-produced 20 × 40 fractions.
  • For AC1230CX, SSA and pH_PZC variations in the ground fraction from bulk GAC were relatively small and random across all grinding methods and size fractions, which could result from two possibilities: (i) AC1230CX was resistant to abrasion and was mainly fractured during all grinding methods, or (ii) AC1230CX does not possess a radial trend in BO%. SEM micrographs and At%-O1s changes in the ground fractions compared to bulk AC1230CX supported possibility (ii).
  • The findings suggest that GAC type and the target particle size range should be considered when choosing a grinding method. For BC-GAC or any GAC or other types of sorbents, such as biochar, bone char, and ion exchange resins [5,62,69], that potentially possess radial variations in chemical and/or structural properties, MP grinding is recommended if a larger particle size (e.g., 20 × 40) is desired because the resulting shock forces result in more representative particles compared to the shear forces realized by the blender and the BMU. For smaller particle sizes (e.g., 100 × 200 or 200 × 325), grinding methods can be selected based on cost, labor intensity, and operational safety because the various methods of characterization either demonstrated less physical/chemical changes compared to the original or similar changes across all the methods when the particles were ground to such small sizes. Investigation into the effect of wet vs dry grinding on GAC properties is recommended for future research.
  • The applied result of the above characterization and analysis supports the ability to use the blender or BMU in RSSCT experimentation when using small particle sizes as opposed to the more commonly referenced MP method. For larger experiments requiring more ground GAC, such a conclusion can reduce the grinding time and labor to less than 10% of what originally was needed.

Data availability

The data will be published at the following DOI: 10.23719/1527,926.