Phenol Molecular Sheets Woven by Water Cavities in HydrophobicSlit NanospacesPiotr Kowalczyk,*,† Marek Wisniewski,‡ Artur Deditius,† Jerzy Włoch,‡ Artur P. Terzyk,‡
Wendell P. Ela,† Katsumi Kaneko,§ Paul A. Webley,∥ and Alexander V. Neimark⊥
†School of Engineering and Information Technology, Murdoch University, 90 South Street, Murdoch 6150, Western Australia,Australia‡Physicochemistry of Carbon Materials Research Group, Faculty of Chemistry, N. Copernicus University in Torun, 7 Gagarin Street,87-100 Torun, Poland§Center for Energy and Environmental Science, Shinshu University, 4-17-1, Wakasato, Nagano-City 380-8553, Japan∥School of Chemical and Biomedical Engineering, University of Melbourne, Parkville, Victoria 3010, Australia⊥Department of Chemical and Biochemical Engineering, Rutgers, The State University of New Jersey, 98 Brett Road, Piscataway,New Jersey 08854-8058, United States
*S Supporting Information
ABSTRACT: Despite extensive research over the last several decades, the microscopic characterization of topological phases ofadsorbed phenol from aqueous solutions in carbon micropores (pore size < 2.0 nm), which are believed to exhibit a solid andquasi-solid character, has not been reported. Here, we present a combined experimental and molecular level study of phenoladsorption from neutral water solutions in graphitic carbon micropores. Theoretical and experimental results show highadsorption of phenol and negligible coadsorption of water in hydrophobic graphitic micropores (super-sieving effect). Graphicprocessing unit-accelerated molecular dynamics simulation of phenol adsorption from water solutions in a realistic model ofcarbon micropores reveal the formation of two-dimensional phenol crystals with a peculiar pattern of hydrophilic−hydrophobicstripes in 0.8 nm supermicropores. In wider micropores, disordered phenol assemblies with water clusters, linear chains, andcavities of various sizes are found. The highest surface density of phenol is computed in 1.8 nm supermicropores. Thepercolating water cluster spanning the entire pore space is found in 2.0 nm supermicropores. Our findings open the door for thedesign of better materials for purification of aqueous solutions from nonelectrolyte micropollution.
1. INTRODUCTION
Adsorption processes are playing a central role in waterpurification and clean-up from nonelectrolyte micropollutionincluding toxic industrial chemicals, dyes, chlorine andchloramines, pesticides, pharmaceuticals, personal care prod-ucts, and endocrine-disrupting chemicals, amongst others.1 Ithas been estimated that ∼80% of industrial water contaminatesare phenolic derivatives.2 Of these, most compounds arerecognized as toxic carcinogens.2 Thus, is it not surprising thatphenol, a planar molecule with a hydroxyl group attached tothe benzene ring, is a recommended probe for testing potentialadsorbents for water purification and clean-up by adsorptionprocesses.3,4
Nanoporous carbon adsorbents, such as granular activatedcarbons (GACs) produced from natural precursors andactivated carbon fibers (ACFs), have been used in portablewater purification systems and water treatment plants for theproduction of clean drinking water.5,6 It is generally acceptedthat the wettability and the nanopore structure of carbonadsorbents are the most important properties for theoptimization of their performance toward adsorptive removalof nonelectrolyte contaminates.7−9 Yet, the specifics ofcompetitive solute−water adsorption in nanopores of the
Received: August 20, 2018Revised: October 23, 2018Published: November 17, 2018
molecular size are still poorly understood. As a result, a “try itand see” approach rather than theory-informed design hasbeen commonly used in search for advanced adsorbents.The cooperative effects of the nanopore size and wettability
in adsorption of phenols are subjects of subsequentinvestigation.10,11 A variety of experimental methods havebeen used to study the mechanism of phenol adsorption fromaqueous solutions at different pH, including amongst others,adsorption, calorimetry, Boehm titration, and X-ray photo-electron spectroscopy analysis.12−14 Although these methodsprovide a viable insight into the mechanism of phenoladsorption from aqueous solutions, they lack a detailedanalysis of the molecular configurations of the phenol−watermixtures in nanopores. Moreover, the methods of classicalphenomenological thermodynamics have been commonly usedfor modeling and interpretation of adsorption isotherms.15−17
However, the homogeneous representation of adsorbed phasesby classical thermodynamics is questionable, when the sizes ofpores and adsorbed molecules are similar. This is because ofthe molecular packing effects and strong lateral interactions inconfining geometries.18,19
It is generally accepted20−22 that adsorption of phenol fromaqueous solutions on carbon adsorbents is maximized atneutral pH, when the nanopore blocking and coadsorption ofwater molecules on the oxygen-containing functional groups(so-called “solvent effect”) is the lowest. In hydrophobicultramicropores and narrow supermicropores with pore sizes<1.4 nm, the pore filling with pure phenol phase has beenpostulated based on empirical correlations.20,22 In widersupermicropores with hydrophobic pore walls (pore size >1.4 nm), the coadsorption of water is expected, however, boththe composition and the structure of the adsorbed phenol−water mixtures is unknown. In hydrophobic mesopores (poresize > 2.0 nm), the adsorption of phenol is limited to spacesthat are close to pore walls (e.g., contact and second adsorbedlayers), whereas the remaining part of the mesopore is filled bywater molecules.23 The term “hydrophobic” is not preciselydefined because all nanoporous carbon materials are oxidizedto some extent upon their exposure to air.24,25 Therefore, onecan expect that the adsorption of phenol from aqueoussolutions is either comparable or lower than the adsorption ofpure phenol from the saturated vapor pressure at the sameoperating temperature.We have recently reported the super-sieving effect in phenol
adsorption from aqueous solutions on nanoporous carbonbeads synthesized from pure polymeric precursors.18 In thecurrent work, we investigate the microscopic mechanism of thesuper-sieving effect by combining the experimental results withgraphic processing unit-accelerated molecular dynamics (GPU-MD) simulation of phenol adsorption from aqueous solutionsat neutral pH. Studied activated carbon samples and theexperimental methods are described in Section 2. To the bestof our knowledge, the GPU-MD simulations of adsorptionfrom aqueous solutions are reported for the first time. Thedetails and description of the computed properties are given inSection 2. Experimental and theoretical results are criticallycompared and discussed in Section 3. Final conclusions aregiven in Section 4. The study demonstrates the importance ofthe synergy between the experiment and atomistic levelsimulations in design and smart fabrication of novel nano-porous carbon adsorbents for purification and production ofultrapure water.
2. MATERIALS AND METHODS2.1. Materials. A sample of nanoporous carbon beads, NCB-8h,
was prepared at Murdoch University by carbonization of acidic, gel-type cation exchange resin based on a styrene−divinylbenzenecopolymer (Lanxess Lewatit, Germany) and subsequent activationwith CO2. In the carbonization step, ∼2 g of precursor was placed in aceramic crucible and carbonized at 650 °C for 1.5 h in a high-purityN2 steam (1.0 dm3/min, BOC, Australia) in a horizontal split-tubefurnace. Then, the carbonized beads were activated with CO2 at 900°C for 8 h in a high-purity CO2 steam (1.0 dm3/min, BOC, Australia)in a horizontal split-tube furnace. Norit GAC from CabotCorporation (USA) was used as an industry benchmark.
2.2. Raman, XRD, and SEM Measurements. Raman spectra ofthe NCB-8h sample were recorded using the WITec alpha 300RA +confocal Raman imaging system at the Centre for Microscopy,Characterization, and Analysis (CMCA), The University of WesternAustralia (UWA). We used the laser excitation wavelength of 532 nm.The Raman spectra were collected from 4 to 6 points chosenrandomly on the surface of NCB-8h sample. The nonpolarized Ramanscattering spectra of Norit were investigated in the spectral range of60−4500 cm−1 at Nicolaus Copernicus University (NCU). Ramanspectra were recorded in the backscattering geometry using Senterramicro-Raman system using green laser operating at 532 nm. The laserbeam was tightly focused on the sample surface through a 50×microscope objective. The position of the microscope objective withrespect to the sample was piezoelectrically controlled (XY position).To prevent any damage of the sample, an excitation power was fixedat 2 mW. The resolution was 4 cm−1, and CCD temperature of 223 K,laser spot of 2 μm in diameter, and total integration time of 100 s (50× 2 s) were used. X-ray powder scattering (XRD) patterns of NCB-8hsample was recorded using the Empyrean multipurpose researchdiffractometer using Cu Kα radiation (λ = 0.15406 nm) at CMCA,UWA. Bulk powder of the Norit sample was characterized by theXRD using a Philips X’Pert Pro diffractometer with Cu Kα1 radiationat NCU. Scanning electron microscopy (SEM) images of NCB-8hsample were recorded using the Verios XHR SEM system at theCMCA, UWA. SEM studies of Norit were performed with Quanta 3DFEG (EHT = 30 kV) instrument at NCU. Samples were placed ontocarbon tabs attached to aluminum SEM stubs. NCB-8h and Noritsamples were analyzed in the microscope without coating treatment.
2.3. N2 Adsorption Measurements. The N2 adsorption−desorption isotherms on Norit and NCB-8h samples were measuredat 77 K using the ASAP 2010 MicroPore System (Micromeritics,USA) at NCU. Before each measurement, carbon samples weredesorbed in vacuum at 383 K for 3 h. Adsorption and structuralparameters of Norit and NCB-8h samples were computed using theBrunauer−Emmett−Teller (BET) theory of multilayer adsorption,26
and general integral equation of adsorption implemented with localN2 adsorption isotherms computed from the classical nonlocal densityfunctional theory (NLDFT)27−29 (Figure S2 in the SupportingInformation).
2.4. Phenol Adsorption and Calorimetry Measurements.Phenol adsorption isotherms from water solutions (298 K, neutralpH) on NCB-8h and Norit samples were measured using UV−visspectrophotometer JASCO V-660.30 For each adsorption point, wecollected the equilibrium concentration after 24 and 48 h ofequilibration at 270 nm. The enthalpy of immersion in water andin phenol solutions was measured at 298 K using a Tian−Calvetisothermal microcalorimeter.18 The measurements for each carbonsample were repeated three times. The error was not larger than 0.5mJ/m2.
To gain insights into the mechanism of phenol adsorption fromdilute aqueous solutions, we computed the adsorption energydistribution (AED) function using the method recommended inrefs 31 and 32
where θt(c) denotes the total fractional coverage of a solute, F(ε12) isthe normalized distribution function, which characterizes the materialheterogeneity in terms of the adsorption energy difference, ε12 = ε1 −ε2, between the solute (phenol, ε1) and the solvent (water, ε2), x = c1/c1sol, where c1 and c1
sol are the equilibrium concentration and thesolubility of phenol in water, respectively, R is the universal gasconstant, and T denotes temperature. The home-made implementa-tion of the second-order Tikhonov regularization method with thenon-negative least-square algorithm was used for inversion of eq 1.33
Theoretical phenol-saturated capacity of studied carbon materialswas computed using the pore size distribution f(H) obtained from theNLDFT method from N2 (77 K) adsorption and the theoreticalphenol capacities, Ntheor.(H), simulated from constant pressure Gibbsensemble Monte Carlo (MC) simulations18
∫=N p p Npp
H f H H( / ) , ( )dexpt. 0 NLDFT0
i
kjjjjjj
y
{zzzzzz (2)
∫=N N H f H H( ) ( )dsat. theor. (3)
In the first step, we evaluated the pore size distribution function,f(H), from eq 2 using the second-order Tikhonov regularizationmethod with the non-negative least-square algorithm.33 In the secondstep, phenol-saturated capacity was computed using eq 3 and the poresize dependence of the MC simulated phenol-saturation capacity,Ntheor.(H).
18 Note that the theoretical phenol-saturated capacitycomputed from eqs 2 and 3 corresponds to filling of carbonmicropores with pure phenol.2.5. Simulation Methodology. We used a model of Madagascar
graphite (6 layers, each wall has the dimensions x = z = 4.396 nm, y =2.0, 4.0 nm) reconstructed from experimental wide-angle X-rayscattering measurement and the hybrid reverse MC simulationmethod.34 The simulation box was constructed as follows. TwoMadagascar graphite crystals were placed in a box to form theopposite pore walls. The walls were movable to control the effectivepore widths (Heff, as shown in Figure 1). Next, two virtual repulsivewalls were placed at the top and the bottom of each slit-shaped pore(see violet planes in Figure 1) excepting the entrance to a pore.The boxes (x = z = 4.396 nm, y = 9.030 nm) containing
equilibrated water−phenol mixtures were placed above and below thepore box. Each box contained 120 phenol and 5103 water molecules,respectively (the concentration is equal to 10.90% mass). Those boxeswere prepared by the equilibration of the phenol−water mixtures
using the SPC/E water molecules35 and variable Langevin integrator/thermostat, together with the stochastic Andersen barostat (1 bar).The OPLSAA parameters for phenol were taken from the TINKERpackage.36 For the water−carbon interaction, the parametersproposed by Werder et al. were applied.37
Simulations of phenol adsorption (T = 323 K, NVT ensemble)were performed for each carbon pore for a single point constantphenol equilibrium concentration equal to 10.90% mass. TheOpenMM 7.01, a high-performance toolkit for molecular simulationswas applied (GPU-MD).38 The equilibration time was equal to 10.5ns. The simulation is performed in steps by consecutive equilibrationof the outside mixture containers with the pore. At each step, duringsimulation, phenol molecules fill the nanopore and the phenolpressure (or concentration) outside containers reduces down to theequilibrium concentration corresponding to the accumulated amountof phenol in the pore. On the next steps, phenol solution containerswere replaced by the new ones containing the same prepared 10.9%mixture, and the equilibration is repeated. The boxes are replaceduntil the number of molecules in the nanopore was constant, andadsorption was calculated, assuming that the molecule is inside ananopore if its center of mass is inside (Figure S3 in the SupportingInformation). The variable Langevin integrator with time step equalto 2 fs was used in all of the above simulations. Simulations wereperformed for the pores widths Heff (nm) = 0.8, 1.05, 1.3, 1.55, 1.8,2.05, 2.3, 2.8, 3.3, and 3.8.
To characterize the structural ordering of confined phenol andwater molecules, we computed the angular order parameter, S(α),from the following expression39,40
α α α= ⟨ ⟩ = −S P( ) (cos )
3 cos 122
2
(4)
where α denotes the angle between the graphitic pore wall normal andthe plane vector of the adsorbed molecule (Figure S4 in theSupporting Information). The bracket means that the orderparameters were averaged over all configurations recorded duringthe GPU-MD simulations. If all of the adsorbed molecules areperpendicular to the graphitic pore walls, S = 1.0. In contrast, S =−0.5 indicates that all adsorbed molecules are parallel to the graphiticpore walls. The isotropic distributions of adsorbed moleculesorientation in carbon micropores correspond to S = 0. Additionally,we computed the local phenol and water density profiles.41
3. RESULTS AND DISCUSSIONSFigures 2a,b and S1 in the Supporting Information displaySEM images of the surface of Norit and NCB-8h sample. Thesurface morphology of the Norit-activated carbon is quitedifferent compared with the NCB-8h carbon sample producedfrom a synthetic precursor. The surface of the Norit is veryrough with the heterogeneous structure of macro- and meso-pore entrances that are directly connected to the exterior(Figures 2a and S1 in the Supporting Information). Micro-pores are therefore predominantly connected to irregularmesopores, as in the classical tree-pore model of disorderedporous carbons.43 In contrast, the micropore entrances at thesurface of the NCB-8h spherical beads are directly connectedto the exterior, which is typical for ACFs produced frompolymeric precursors (so-called cylindrical pore model33). TheRaman spectral examination of both carbon materials displaytwo overlapping broad bands locating at 1350 and 1590 cm−1
(Figure 2c). Such a feature is associated with the disordering ofthe samples (D band) and stretching vibration in the aromaticplanes (G band), respectively, and confirms that both carbonsamples consists of nanometer-scale graphitic crystallitesembedded in the disordered carbon matrix. For Norit andNCB-8h samples, the in-plane size of graphitic crystallites,2.1−2.2 nm, computed from the Ferrari−Robertson equation
Figure 1. Slit-shaped carbon nanopore model with the Madagascargraphite walls (left panel) and the simulation box before GPU-MDequilibration (right panel, two boxes with equilibrated phenolsolutions, virtual repulsive walls and a slit-shaped carbon nanopore).Note that the graphics collected in this figure and Figure 5 are createdusing the VMD program.42
are comparable (Table 1).44 It is not surprising because thecarbonization and activation of carbon precursors at 800−900°C should generate similar sizes of graphite nanocrystallites.To induce a further growth of graphitic crystallites,significantly higher temperatures are necessary.45 X-raydiffraction (XRD) patterns shown in Figure 2d are dominatedby intensities of (002) and (100) peaks. For both Norit andNCB-8h samples, the (002) diffraction peaks are broadenedand shifted to lower angles as compared to (002) diffractionpeak of graphite, indicating larger interlayer spacing betweengraphitic planes of nanocrystallites than those of graphite.46
Indeed, an average interlayer spacing (d002) computed forNorit and NCB-8h is 0.361 and 0.399 nm, respectively, and itis greater than graphite spacing of 0.335 nm (Table 1). Thesestructural characteristics are typical for turbostratic carbona-ceous materials.47 This observation justifies to some extent theuse of the simplistic carbon slit-shaped pore model employedfor the pore size distribution analysis and in computersimulations (Figure 1).48−50
Figure 3a presents, measured volumetrically, experimentalN2 adsorption−desorption isotherms at 77 K on Norit andNCB-8h. The isotherm measured for NCB-8h is typical formicroporous materials (type I isotherm following toIUPAC);51 however, for Norit, a mixture type between I and
IV is observed. The Norit sample is structurally heterogeneousactivated carbon with an NLDFT specific surface area of 894m2/g, and the total pore volume at p/p0 = 0.98 is 0.56 cm3/g(Table 1 and Figure 1a). The pore size distribution of Norithas a broad range of micropore ranges from ultra- to super-micropores (Figures 3b and S2 in the Supporting Informa-tion). Moreover, an increase in N2 adsorption in the classicalBET range (0.05−0.25p/p0) and the sorption hysteresis loopindicates the presence of narrow mesopores (Figure 3a). Thespecific surface area of mesopores 88 m2/g is calculated fromN2 adsorption isotherm and NLDFT method (Table 1 andFigure S2 in the Supporting Information). NBC-8h exhibited apredominant microporosity with an NLDFT specific surfacearea of 1186 m2/g, and the total pore volume at p/p0 = 0.98 is0.45 cm3/g (Table 1 and Figure 1a). The NLDFT pore sizedistribution function for NCB-8h is qualitatively similar to thatcomputed for Norit. However, for NCB-8h carbon, ultra-micropores of uniform pore size <0.6 nm are the predominantfraction of micropores, which is typical for carbon molecularsieves (Figures 3b and S2 in the Supporting Information).52,53
Figure 3c,d presents the experimental adsorption isotherms ofphenol measured from water solutions (298 K, neutral pH)and corresponding AED functions. A significant amount ofphenol is adsorbed by Norit and NCB-8h carbon samples,
Figure 2. SEM image of the (a) Norit-activated carbon and (b) NCB-8h sample activated using CO2 at 900 °C for 8 h. (c) Raman spectra and (d)XRD diffraction patterns recorded for NCB-8h and Norit carbon samples.
Table 1. Crystallographic (Interlayer Spacing d002), Crystallite (In-Plane Sizes of Graphitic Crystallites La), and NanoporeParameters Estimated from the Raman, XRD Scattering, and N2 Adsorption (77 K) Measurementsa
aThe total (Stot), mesopore (Smeso) and micropore (Smicro) specific surface areas were estimated from N2 adsorption isotherms and NLDFT method(Figure S2 in the Supporting Information). For comparison, the specific surface areas (SBET) and monolayer capacities (aBET) of carbon sampleswere evaluated from N2 adsorption isotherms and the BET method.
which indicates that carbon ultramicropores and narrowsupermicropores are accessible to phenol molecules, and theblockage of micropore entrances due to solvent effects can beneglected. The broad AED function of Norit (Figure 3d)indicates that phenol molecules are adsorbed on energeticallyheterogeneous structure of micro- and meso-pores, which isconsistent with SEM observations. The calculated parametersshow that the adsorption energy of phenol on Norit is larger by12.7 ± 6.5 kJ/mol than that of water at the neutral pH (Table2). On the other hand, a bimodal AED function of NBC-8h(Figure 3d) can be explained by adsorption of phenolmolecules in fine-tuned ultra- and super-micropores, which isconsistent with the SEM observations and NLDFT pore sizeanalysis. The adsorption energy of phenol on NCB-8h isstronger by 13.3 ± 1.9 and 25.7 ± 1.1 kJ/mol than that ofwater at neutral pH (Table 2). The higher value of adsorptionenergy is comparable with the experimental molar enthalpychange 22 kJ/mol measured for phenol−water systems on AR-BPL porous carbon (SBET = 1088 m2/g).54
Figure 4a,b displays measured calorimetrically (T = 298 K)enthalpy of immersion of Norit and NCB-8h in phenolsolutions and pure water, respectively. For both carbonsamples, the variation of enthalpy of immersion in phenolsolutions is quite similar. At low pore loadings, a small heat of40 J/g is released because the adsorption process is dominated
by the weak and nonspecific interactions of water moleculeswith the carbon surface (Figure 4a). With further adsorption ofphenol molecules, the enthalpy of immersion is monotonicallyincreasing because of the enhanced phenol−carbon inter-actions in ultra- and super-micropores (e.g., π−π dispersionforces). Following Bradley,55 the experimental enthalpy ofimmersion of nonporous carbon black in water correlates withthe carbon surface oxygen level [O]t (atom %)
Δ [ ] = + [ ]H (H O) mJ/m 40 11.6 Oi 22
t (5)
The enthalpy of immersion of NCB-8h and Norit samples inpure water 44−45 mJ/m2 (Figure 4b) is comparable with 40mJ/m2 (e.g., very small [O]t in eq 5), indicating a negligibleamount of oxygen-containing functional groups on bothcarbon samples. This confirms that Norit and NCB-8h carbonsamples are hydrophobic with a strong adsorption affinitytoward phenol.Figure 4c,d shows the pore width-variation of the excess of
phenol adsorption computed using eqs 2 and 3. It isremarkable that experimentally measured phenol adsorptioncapacity from aqueous solutions on Norit and NBC-8h is justslightly overestimated by the theory (Table 2). Because thetheoretical phenol capacities are computed with theassumption of the micropore filling with pure phenol (eqs 2and 3), there is evidence that coadsorption of water in Norit
Figure 3. (a) Experimental N2 adsorption isotherms (77 K) on Norit and NCB-8h samples. (b) Pore size distributions computed from N2adsorption isotherms and NLDFT method. (c) Experimental (expt.) and theoretical (theor.) phenol adsorption isotherms from aqueous solutionsat neutral pH (298 K). (d) AED functions for phenol on Norit and NCB-8h carbon samples.
Table 2. Characteristics of the Adsorption Energy Distribution Functions for NCB-8h and Norit Carbon Samples Evaluatedfrom Phenol Adsorption Isotherms Measured from Water Solutions at Neutral pH (298 K)a
and NCB-8h samples is negligible. Figure 4c,d also reveals theimportance of the pore size distribution on the mechanism ofphenol adsorption. For Norit sample, only 100 mg/g of phenolis adsorbed in carbon ultramicropores, whereas 200 mg/g ofphenol is adsorbed in carbon ultramicropores of NCB-8hsample. NCB-8h and Norit supermicropores are then furtherfilled by phenol; however, both theoretical and experimentalsaturation phenol capacity of NCB-8h sample is 50 mg/ghigher than Norit. The superior phenol-adsorption perform-ance for the NCB-8h sample clearly demonstrates that thepurification and clean-up of water from phenol can beimproved by nanopore engineering and tuning the carbonsample wettability.Although it could be argued that phenol molecules are
solidified in carbon micropores and the coadsorption of wateris negligible, the GPU-MD simulations are crucial forvisualization and characterization of phenol−water phasesadsorbed in carbon micropores. Figure 5 shows the results ofthe phenol adsorption from water solutions in model graphiticnanopores simulated from the GPU-MD method. We noticethat the surface density of adsorbed phenol increases up to themaximum at 1.8 nm (Figure 5c), followed by a fast decrease upto 2.25 nm. For pore sizes >2.25 nm, the surface density ofadsorbed phenol reaches an asymptotic value. This indicatesthat the mechanism of the phenol adsorption in mesoporesand on the flat graphite surface is the same. While this is notsurprising per se, it is remarkable that the monolayerconfinement stabilizes the two-dimensional phenol crystalswith a peculiar pattern of hydrophilic−hydrophobic stripes(Figure 5a). Static self-assembly between adsorbed phenolmolecules does explain the formation of hydrophobic−hydrophilic stripes. To the best of our knowledge, this is the
first observation of the self-assembled phenol crystals stabilizedby the monolayer confinement. From the variation of the localdensity profile and order parameter in 0.8 nm supermicropore,we concluded that phenol molecules are compressed in themiddle of the micropore and their preferential orientations areparallel to the pore walls (Figure 6). Furthermore, we foundthat phenol crystal defects are filled with water singlemolecules only (Figure 7a). The water molecules are adsorbedcloser to the pore walls compared with phenol, and theirpreferential orientations are parallel to the pore walls (Figures6 and 7a). A peculiar configuration of water molecules in theadsorbed layer can be explained by constraints that areimposed by the two-dimensional phenol crystal (Figure 7a).Two contact layers of phenol are found in a wider 1.0 nmmicropore (Figure 6). There is, however, a fraction of phenolmolecules that are adsorbed close to the pore center.Interestingly, small water clusters (e.g., monomers, dimers,trimers, and small chain-like water clusters) are preferentiallycoadsorbing at phenol contact layers including spaces close tothe pore walls and the pore center. That is why, the water localdensity profile is somehow broader as compared to the phenoldensity profile (Figure 6). Snapshots of the GPU-MDsimulations reveal that small water clusters are isolated (Figure7b). They do not exclude the pore volume for phenol but theyfill the defects between adsorbed phenol molecules (Figure 6).Inside 1.8 nm micropore, the water clusters with different sizesand morphologies are found around the pore center (Figures 5and 6). Two layers of phenol are adsorbed on pore walls, asshown on the local density profile (Figure 6). As expected,water molecules have random orientations with an exception ofwater coadsorbing in phenol layers. The water clusters aregrowing in 1.8 nm supermicropores; however, they do not
Figure 4. (a) Experimental enthalpy of NCB-8h and Norit immersion in aqueous solutions of phenol at 298 K. (b) Experimental enthalpy of NCB-8h and Norit immersion in water per surface area measured at 298 K. (c,d) Theoretical excess of phenol adsorption computed using themethodology proposed by Kowalczyk et al.18
form a percolating network. In 2.0 nm supermicropores theliquid-like isotropic water is filling the spaces close to the porecenter, whereas phenol molecules are adsorbed close to thepore walls with preferential flat orientations (Figures 5−7b).Moreover, we observe the formation of the percolated clusterof water molecules spanning the whole space of the pore(Figures 5 and 7b). In narrow mesopores (pore size 2.25 nm),the surface density of adsorbed phenol reaches an asymptoticvalue corresponding to the adsorption of phenol from watersolutions on two independent graphitic surfaces (Figure 5).Taking into account the GPU-MD simulation results, weconcluded that the maximum enhancement of the surfacedensity of phenol in 1.8 nm supermicropore is 23% greater ascompared to an open graphite surface (Figure 5). The highestselectivity of phenol over water is theoretically predicted innarrow supermicropores, which is an important factor fordesign and pore engineering of novel carbon adsorbents forproduction of ultrapure water.
4. CONCLUSIONS
In this work, we investigated the mechanism of phenoladsorption from aqueous solutions (293 K, neutral pH) on tworepresentative samples of activated carbons, namely, nano-porous carbon beads NCB-8h and GAC Norit from Cabot. Forboth carbon samples, we found the low value of the enthalpy ofimmersion in pure water 44−45 mJ/m2, which indicate a
strongly hydrophobic character of carbon micropores. Thesuper-sieving effect in phenol adsorption from aqueoussolutions at neutral pH in Norit and NCB-8h carbon samplesis confirmed based on the combined theoretical andexperimental results. However, from the recent developmentof the theory of super-sieving effect in phenol adsorption fromaqueous solutions, we found that the distribution of the phenoladsorption capacity in Norit and NCB-8h is very different. ForNorit, only 100 mg/g of phenol is adsorbed in ultramicropores(pore size < 0.6 nm), whereas 200 mg/g of phenol is adsorbedin ultramicropores of NCB-8h. Both theoretical and exper-imental saturation phenol capacity of NCB-8h is 50 mg/ghigher than Norit.To gain insights into microscopic mechanism of super-
sieving effect in phenol adsorption from aqueous solutions, weimplemented and simulated the phenol adsorption fromneutral water solutions in model slit-shaped graphitic carbonnanopores using MD on a graphics processing units. GPU-MDsimulations revealed that adsorption and structural ordering ofphenol−water mixtures in hydrophobic carbon microporesdepend strongly on the pore size. In 0.8 nm supermicropore, astatic self-assembly of phenol molecules into two-dimensionalcrystal with a peculiar pattern of hydrophilic−hydrophobicstripes is stabilized by the monolayer confinement at ambienttemperatures. Irregular water clusters (e.g., monomers, dimers,and trimers) and short linear water chains are found in 1.8 nm
Figure 5. Surface density of phenol adsorbed from aqueous solutions at 323 K and neutral pH computed from GPU-MD simulations. Middlepanels show the equilibrium snapshots of phenol−water mixtures adsorbed in 0.8 (a), 1.0 (b), 1.8 (c), and 2.0 nm (d) graphitic micropores at thegiven conditions (water molecules are brighter). The structure of water monomers, clusters, cavities, and percolating cluster is highlighted in thebottom panels.
supermicropore. However, the simulated snapshots showedthat the coadsorbing water is below the percolation threshold.Additionally, in 0.8 nm supermicropore, we predicted thehighest surface density of phenol that is 23% greater ascompared to that on a graphite surface. Further growth ofwater clusters results in the formation of the percolating watercluster spanning the entire pore space of 2.0 nm super-micropore. In narrow mesopores (pore size 2.25 nm), thesurface density of adsorbed phenol reaches an asymptotic valuethat corresponds to adsorption on two independent graphitesurfaces.The super-sieving effect in phenol adsorption from aqueous
solutions in hydrophobic carbon micropores is explained byanalyzing the structure of coadsorbed water. We found thatsmall clusters (e.g., monomers, dimers, and trimers) and shortlinear water chains do not exclude pore volume for phenoladsorption significantly. Interestingly, the percolating watercluster is found in 2.0 nm supermicropores. In mesopores, the
liquid-like water excludes the significant part of the porevolume for adsorption of phenol.Our findings show that GPU-MD simulations combined
with the pore size distribution analysis can be used for theory-informed synthesis of advanced materials for water purificationand clean-up from nonelectrolyte micropollution. Narrowsupermicropores and ultramicropores are desirable forproduction of ultrapure water because of the highest selectivityof phenol over the water (e.g., formation of two-dimensionalphenol crystals). The highest uptake of phenol per surface areaof pore walls is found in 1.8 nm supermicropores. Theoptimization of adsorption properties can be achieved bytuning the hydrophobicity and pore size in the ultramicroporerange as confirmed by the superior phenol-adsorptionperformance for NCB-8h samples.
■ ASSOCIATED CONTENT
*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.lang-muir.8b02832.
High magnification SEM image of the surface of theNorit-activated carbon, fitting of the experimental N2adsorption isotherms (77 K) by NLDFT method,NLDFT cumulative pore size distributions, variation ofthe phenol adsorption with GPU-MD run, schematicdiagram showing a definition of the angle between thegraphitic pore wall normal and the plane vector of theadsorbed phenol, and variation of the local densityprofiles and order parameter computed from GPU-MDsimulations (PDF)
Figure 6. Variation of the local density (left panels) and order parameter (right panels) of phenol and water adsorbed in 0.8, 1.0, 1.8, and 2.0 nmgraphitic micropores at 323 K and neutral pH computed from GPU-MD simulations.
Figure 7. Snapshots of the equilibrium phenol−water mixturesadsorbed in 0.8 (a) and 2.0 (b) nm supermicropores obtained fromGPU-MD simulations. Monomer (a) and percolating water clusterspanning the entire pore space (b) is highlighted by rectangular boxes.
ORCIDPiotr Kowalczyk: 0000-0002-7523-0906Artur P. Terzyk: 0000-0003-0622-1771Katsumi Kaneko: 0000-0002-2378-478XPaul A. Webley: 0000-0003-3598-3767Alexander V. Neimark: 0000-0002-3443-0389NotesThe authors declare no competing financial interest.
■ ACKNOWLEDGMENTS
P.K. acknowledges the financial support from the MurdochUniversity start-up grant: nanopore controlled syntheticcarbons for interfacial separations and catalysis (11701). Thework was partially supported by Grant-in-Aid for ScientificResearch (B) (17H03039) and the JST OPERA project. AVNacknowledges support by the NSF grant No 1834339.
■ REFERENCES(1) Marsh, T. H.; Rodríguez-Reinoso, F. Activated Carbon; Elsevier:Amsterdam, 2006.(2) Dąbrowski, A.; Podkoscielny, P.; Hubicki, Z.; Barczak, M.Adsorption of phenolic compounds by activated carbon-a criticalreview. Chemosphere 2005, 58, 1049−1070.(3) Moreno-Castilla, C. Adsorption of organic molecules fromaqueous solutions on carbon materials. Carbon 2004, 42, 83−94.(4) Lin, S.-H.; Juang, R.-S. Adsorption of phenol and its derivativesfrom water using synthetic resins and low-cost natural adsorbents: areview. J. Environ. Manage. 2009, 90, 1336−1349.(5) Bansal, R. C.; Goyal, M. Activated Carbon Adsorption; CRCPress: Boca Raton, 2005.(6) Derbyshire, F.; Jagtoyen, M.; Andrews, R.; Rao, A.; Martin-Gullo n, I.; Grulke, E. A. Carbon materials in environmentalapplications. In Chemistry and Physics of Carbon; Radovic, L. R.,Ed.; Marcel Dekker: New York; 2001, Vol. 27, pp 1−66.(7) Terzyk, A. P. Further insights into the role of carbon surfacefunctionalities in the mechanism of phenol adsorption. J. ColloidInterface Sci. 2003, 268, 301−329.(8) Salame, I. I.; Bandosz, T. J. Role of surface chemistry inadsorption of phenol on activated carbons. J. Colloid Interface Sci.2003, 264, 307−312.(9) Stoeckli, F.; Lo pez-Ramo n, M. V.; Moreno-Castilla, C.Adsorption of Phenolic Compounds from Aqueous Solutions, byActivated Carbons, Described by the Dubinin−Astakhov Equation.Langmuir 2001, 17, 3301−3306.(10) Laszlo, K.; Podkoscielny, P.; Dąbrowski, A. Heterogeneity ofpolymer-based active carbons in adsorption of aqueous solutions ofphenol and 2,3,4-trichlorophenol. Langmuir 2003, 19, 5287−5294.(11) Radovic, L. R.; Moreno-Castilla, C.; Rivera-Utrilla, J. Carbonmaterials as adsorbents in aqueous solutions. In Chemistry and Physicsof Carbon; Radovic, L. R., Ed.; Marcel Dekker: New York, 2001; Vol.27, pp 227−405.(12) Bandosz, T. J. Activated Carbon Surfaces in EnvironmentalRemediation; Elsevier: Oxford, 2006; Vol. 7.(13) Fernandez, E.; Hugi-Cleary, D.; Lopez-Ramon, M. V.; Stoeckli,F. Adsorption of phenol from dilute and concentrated aqueoussolutions by activated carbons. Langmuir 2003, 19, 9719−9723.(14) Terzyk, A. P. The impact of carbon surface chemicalcomposition on the adsorption of phenol determined at the realoxic and anoxic conditions. Appl. Surf. Sci. 2007, 253, 5752−5755.(15) Hamdaoui, O.; Naffrechoux, E. Modelling of adsorptionisotherms of phenol and chlorophenols onto granular activated carbon
Part I. Two-parameter models and equations allowing determinationof thermodynamic parameters. J. Hazard. Mater. 2007, 147, 381−394.(16) Fierro, V.; Torne-Fernandez, V.; Montane, D.; Celzard, A.Adsorption of phenol onto activated carbons having different texturaland surface properties. Microporous Mesoporous Mater. 2008, 111,276−284.(17) Caturla, F.; Martín-Martínez, J. M.; Molina-Sabio, M.;Rodriguez-Reinoso, F.; Torregrosa, R. Adsorption of substitutedphenols on activated carbon. J. Colloid Interface Sci. 1988, 124, 528−534.(18) Kowalczyk, P.; Deditius, A.; Ela, W. P.; Wisniewski, M.;Gauden, P. A.; Terzyk, A. P.; Furmaniak, S.; Włoch, J.; Kaneko, K.;Neimark, A. V. Super-sieving effect in phenol adsorption fromaqueous solutions on nanoporous carbon beads. Carbon 2018, 135,12−20.(19) Wisniewski, M.; Furmaniak, S.; Terzyk, A. P.; Gauden, P. A.;Kowalczyk, P. Properties of Phenol Confined in Realistic CarbonMicropore Model: Experiment and Simulation. J. Phys. Chem. C 2015,119, 19987−19995.(20) Lorenc-Grabowska, E. Effect of micropore size distribution onphenol adsorption on steam activated carbons. Adsorption 2016, 22,599−607.(21) Stoeckli, F.; Hugi-Cleary, D. On the mechanisms of phenoladsorption by carbons. Russ. Chem. Bull. 2001, 50, 2060−2063.(22) Lorenc-Grabowska, E.; Rutkowski, P. High basicity adsorbentsfrom solid residue of cellulose and synthetic polymer co-pyrolysis forphenol removal: Kinetics and mechanism. Appl. Surf. Sci. 2014, 316,435−442.(23) Kiraly, Z.; Dekany, I.; Klumpp, E.; Lewandowski, H.; Narres, H.D.; Schwuger, M. J. Selective sorption of phenol and relatedcompounds from aqueous solutions onto graphitized carbon black.Adsorption and flow microcalorimetric studies. Langmuir 1996, 12,423−430.(24) Boehm, H. P. Some aspects of the surface chemistry of carbonblacks and other carbons. Carbon 1994, 32, 759−769.(25) Kelemen, S. R.; Freund, H. XPS characterization of glassy-carbon surfaces oxidized by O2, CO2, and HNO3. Energy Fuels 1988,2, 111−118.(26) Brunauer, S.; Emmett, P. H.; Teller, E. Adsorption of gases inmultimolecular layers. J. Am. Chem. Soc. 1938, 60, 309−319.(27) Kowalczyk, P.; Gauden, P. A.; Terzyk, A. P.; Neimark, A. V.Screening of carbonaceous nanoporous materials for capture of nerveagents. Phys. Chem. Chem. Phys. 2013, 15, 291−298.(28) Ravikovitch, P. I.; Vishnyakov, A.; Russo, R.; Neimark, A. V.Unified Approach to Pore Size Characterization of MicroporousCarbonaceous Materials from N2, Ar, and CO2 Adsorption Isotherms.Langmuir 2000, 16, 2311−2320.(29) Kowalczyk, P.; Gun’ko, V. M.; Terzyk, A. P.; Gauden, P. A.;Rong, H.; Ryu, Z.; Do, D. D. The comparative characterization ofstructural heterogeneity of mesoporous activated carbon fibers(ACFs). Appl. Surf. Sci. 2003, 206, 67−77.(30) Pacholczyk, A.; Terzyk, A. P.; Wisniewski, M.; Gauden, P. A.;Wesołowski, R. P.; Furmaniak, S.; Szczes, A.; Chibowski, E.; Kruszka,B. Phenol Adsorption on Closed Carbon Nanotubes. J. ColloidInterface Sci. 2011, 361, 288−292.(31) Heuchel, M.; Braeuer, P.; Von Szombathely, M.; Messow, U.;Einicke, W. D.; Jaroniec, M. Evaluation of the energy distributionfunction from liquid/solid adsorption measurements. Langmuir 1993,9, 2547−2554.(32) Jaroniec, M.; Madey, R. Physical Adsorption on HeterogeneousSolids; Elsevier: Amsterdam, 1988.(33) Kowalczyk, P.; Tanaka, H.; Kanoh, H.; Kaneko, K. AdsorptionEnergy Distribution from the Aranovich−Donohue Lattice DensityFunctional Theory. Langmuir 2004, 20, 2324−2332.(34) Kowalczyk, P.; Gauden, P. A.; Furmaniak, S.; Terzyk, A. P.;Wisniewski, M.; Ilnicka, A.; Łukaszewicz, J.; Burian, A.; Włoch, J.;Neimark, A. V. Morphologically disordered pore model for character-ization of micro-mesoporous carbons. Carbon 2017, 111, 358−370.
(35) Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. The missingterm in effective pair potentials. J. Phys. Chem. 1987, 91, 6269−6271.(36) Lagardere, L.; Jolly, L.-H.; Lipparini, F.; Aviat, F.; Stamm, B.;Jing, Z. F.; Harger, M.; Torabifard, H.; Cisneros, G. A.; Schnieders,M. J.; Gresh, N.; Maday, Y.; Ren, P. Y.; Ponder, J. W.; Piquemal, J.-P.Tinker-HP: a massively parallel molecular dynamics package formultiscale simulations of large complex systems with advanced pointdipole polarizable force fields. Chem. Sci. 2018, 9, 956−972.(37) Werder, T.; Walther, J. H.; Jaffe, R. L.; Halicioglu, T.;Koumoutsakos, P. On the Water−Carbon Interaction for Use inMolecular Dynamics Simulations of Graphite and Carbon Nanotubes.J. Phys. Chem. B 2003, 107, 1345−1352.(38) Friedrichs, M. S.; Eastman, P.; Vaidyanathan, V.; Houston, M.;Legrand, S.; Beberg, A. L.; Ensign, D. L.; Bruns, C. M.; Pande, V. S.Accelerating molecular dynamic simulation on graphics processingunits. J. Comput. Chem. 2009, 30, 864−872.(39) Allen, M. P.; Tildesley, D. J. Computer Simulation of Liquids;Clarendon: Oxford, UK; 1987.(40) Gauden, P. A.; Furmaniak, S.; Włoch, J.; Terzyk, A. P.;Zielinski, W.; Kowalczyk, P.; Kurzawa, J. The influence of geometricheterogeneity of closed carbon nanotube bundles on benzeneadsorption from the gaseous phase-Monte Carlo simulations.Adsorption 2016, 22, 639−651.(41) Kowalczyk, P.; Tanaka, H.; Kaneko, K.; Terzyk, A. P.; Do, D.D. Grand canonical Monte Carlo simulation study of methaneadsorption at an open graphite surface and in slitlike carbon pores at273 K. Langmuir 2005, 21, 5639−5646.(42) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual moleculardynamics. J. Mol. Graphics 1996, 14, 33−38.(43) Mochida, I.; Korai, Y.; Shirahama, M.; Kawano, S.; Hada, T.;Seo, Y.; Yoshikawa, M.; Yasutake, A. Removal of SOx and NOx overactivated carbon fibers. Carbon 2000, 38, 227−239.(44) Ferrari, A. C.; Robertson, J. Interpretation of Raman spectra ofdisordered and amorphous carbon. Phys. Rev. B: Condens. MatterMater. Phys. 2000, 61, 14095−14107.(45) Wang, S.; Morelos-Gomez, A.; Lei, Z.; Terrones, M.; Takeuchi,K.; Sugimoto, W.; Endo, M.; Kaneko, K. Correlation in structure andproperties of highly-porous graphene monoliths studied with athermal treatment method. Carbon 2016, 96, 174−183.(46) Biscoe, J.; Warren, B. E. An X-Ray Study of Carbon Black. J.Appl. Phys. 1942, 13, 364−371.(47) Burian, A.; Ratuszna, A.; Dore, J. C. Radial distribution functionanalysis of the structure of activated carbons. Carbon 1998, 36, 1613−1621.(48) Steele, W. A. The physical interaction of gases with crystallinesolids. Surf. Sci. 1973, 36, 317−352.(49) Lastoskie, C.; Gubbins, K. E.; Quirke, N. Pore sizeheterogeneity and the carbon slit pore: a density functional theorymodel. Langmuir 1993, 9, 2693−2702.(50) Kowalczyk, P.; Tanaka, H.; Hołyst, R.; Kaneko, K.; Ohmori, T.;Miyamoto, J. Storage of hydrogen at 303 K in graphite slitlike poresfrom grand canonical Monte Carlo simulation. J. Phys. Chem. B 2005,109, 17174−17183.(51) Thommes, M.; Kaneko, K.; Neimark, A. V.; Olivier, J. P.;Rodriguez-Reinoso, R.; Rouquerol, J.; Sing, K. S. W. Physisorption ofgases, with special reference to the evaluation of surface area and poresize distribution (IUPAC Technical Report). Pure Appl. Chem. 2015,87, 1051−1069.(52) Rungta, M.; Wenz, G. B.; Zhang, C.; Xu, L.; Qiu, W.; Adams, J.S.; Koros, W. J. Carbon molecular sieve structure development andmembrane performance relationships. Carbon 2017, 115, 237−248.(53) Kowalczyk, P.; Terzyk, A. P.; Gauden, P. A.; Furmaniak, S.;Wisniewski, M.; Burian, A.; Hawelek, L.; Kaneko, K.; Neimark, A. V.Carbon Molecular Sieves: Reconstruction of Atomistic StructuralModels with Experimental Constraints. J. Phys. Chem. C 2014, 118,12996−13007.(54) Barton, S. S.; Evans, M. J. B.; MacDonald, J. A. F. Chemicalheterogeneity on the carbon surface and adsorption from binaryaqueous solution. Pol. J. Chem. 1997, 71, 651−656.
(55) Bradley, R. H. Recent developments in the physical adsorptionof toxic organic vapours by activated carbons. Adsorpt. Sci. Technol.2011, 29, 1−28.
