Structural features of resorcinol–formaldehyde resin chars and interfacial behavior of water co-adsorbed with low-molecular weight organics

Applied Surface Science 283 (2013) 683–693

Contents lists available at SciVerse ScienceDirect

Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc

Structural features of resorcinol–formaldehyde resin chars and interfacial behavior of water co-adsorbed with low-molecular weight organics Vladimir M. Gun’ko a,∗ , Viktor M. Bogatyrov a , Vladimir V. Turov a , Roman Leboda b , ˛ b , Iliya V. Urubkov c Jadwiga Skubiszewska-Zieba a

Chuiko Institute of Surface Chemistry, 17 General Naumov Street, 03164 Kyiv, Ukraine Faculty of Chemistry, Maria Curie-Skłodowska University, 20031 Lublin, Poland c Kurdyumov Institute of Metal Physics, 36 Vernadsky Boulevard, 03142 Kyiv, Ukraine b

a r t i c l e

i n f o

Article history: Received 2 May 2013 Received in revised form 23 June 2013 Accepted 28 June 2013 Available online 8 July 2013 PACS: 61.43.Gt (Powders porous materials) 68.08.−p (Liquid–solid interfaces) 68.43.−h (Chemisorption/physisorption:adsorbates on surfaces) 68.35.Md (Surface thermodynamics surface energies)

a b s t r a c t Products of resorcinol–formaldehyde resin carbonization (chars) are characterized by different morphology (particle shape and sizes) and texture (specific surface area, pore volume and pore size distribution) depending on water content during resin polymerization. At a low amount of water (Cw = 37.8 wt.%) during synthesis resulting in strongly cross-linked polymers, carbonization gives nonporous particles. An increase in the water content to 62.7 wt.% results in a nano/mesoporous char, but if Cw = 73.3 wt.%, a char is purely nanoporous. Despite these textural differences, the Raman spectra of all the chars are similar because of the similarity in the structure of their carbon sheets with a significant contribution of sp3 C atoms. However, the difference in the spatial organization of the carbon sheet stacks in the particles results in the significant differences in the textural and morphological characteristics and in the adsorption properties of chars with respect to water, methane, benzene, hydrogen, methylene chloride, and dimethylsulfoxide. © 2013 Elsevier B.V. All rights reserved.

Keywords: Resorcinol–formaldehyde resin Char Bound water 1 H NMR Dispersion medium effects Interfacial phenomena

1. Introduction Porous carbons as very effective adsorbents are widely used in industry, technology and medicine. Porous polymeric particles with phenol–formaldehyde resins, styrene divinylbenzene copolymers, etc. or different natural raw materials are main precursors of activated carbons (ACs) produced by activation of products of carbonization (chars) of these organic precursors [1–7]. Structural and textural characteristics of ACs depend on features of precursors and chars, activation conditions and a burn-off degree [8–11]. The morphological and textural characteristics, such as the porosity

∗ Corresponding author. Tel.: +380 44 4229627; fax: +380 44 4243567. E-mail addresses: vlad [email protected], [email protected] (V.M. Gun’ko). 0169-4332/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.apsusc.2013.06.165

(nano-, meso- and macroporosity), the pore size distribution (PSD) and particle shape of ACs can be similar to those of chars possessing, however, lower porosity and lower specific surface area than ACs [12–15]. This aspect is of importance for both structural and adsorption characteristics of the final ACs that affect the adsorbent efficiency of the materials [16–21]. Therefore, investigations of the factors governing the characteristics of both polymeric precursors and chars are of interest from both practical and theoretical points of view with respect to subsequent preparation of most effective ACs [1–7]. Resorcinol–formaldehyde resins can be used as porous polymers or precursors of chars to prepare porous carbon materials [22–30]. The properties of the final carbon materials are mainly determined by the structure of polymeric precursors which can be prepared by condensation of resorcinol with formaldehyde in

684

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

the aqueous media. The synthesis can be carried out in a neutral medium or with addition of basic or acidic catalysts affecting the polymer structure [31–33]. Adsorption properties of carbon adsorbents in respect to lowmolecular H-containing substances (water, methane, benzene, etc.) can be determined by adsorption methods, nuclear magnetic resonance (NMR), infrared (FTIR) and Raman spectroscopies, X-ray diffraction (XRD) and other methods. The NMR studies carried out on different materials showed that the amounts of bound adsorbates can be determined from intensity of proton signals of probes as a function of temperature. The 1 H NMR chemical shifts of adsorbates confined in pores depend strongly on the textural and structural characteristics of adsorbents. The confined space effects can be used to compare the characteristics of adsorbents degassed, e.g. before measurements of nitrogen adsorption isotherms, and adsorbents weakly or strongly hydrated before measurements using low-temperature 1 H NMR, cryoporometry, differential scanning calorimetry, DSC, thermoporometry [34,35]. These methods give quantitative information on the adsorbate location in pores of different sizes. Polycondensation of resorcinol and formaldehyde is occurred in the aqueous media where water can play a role of a solvent and a porogen. Therefore, it is of interest to compare the characteristics of chars of resorcinol–formaldehyde resins synthesized at different content of water. Therefore, the aim of this paper was to synthesize a set of chars of resorcinol–formaldehyde resins prepared at various amounts of water and to study the surface properties of the chars with respect to nonpolar and polar adsorbates. 2. Materials and methods 2.1. Materials Resorcinol (99.9%) and 37% aqueous solution of formaldehyde were used in synthesis of resorcinol–formaldehyde resin (RFR). The polymers were synthesized using 3 g of resorcinol, 1.67 g of formaldehyde and 2.84 g (RFR1), 7.84 g (RFR2) or 12.84 g (RFR3) of water. A mixture of resorcinol and formalin was stirred to form a transparent solution of ripe cherry color. Then 0 (RFR1), 5 (RFR2) or 10 (RFR3) ml of distilled water was added and the mixtures were heated slowly up to the boiling point. During a few minutes (0.5–5 min from the boiling start depending of the water content) the solution became turbid due to fast condensation and solidification of the reactants. This process occurs faster at a lower amount of water. The polymers were dried in air at 120 ◦ C for 1 h. Uniform polymers formed were of different colors from dark-brown (RFR1), brown (RFR2) to reddish-brown (RFR3). Chars (RFR1-C, RFR2-C, and RFR3-C) were prepared using the RFR powdered in a mortar and then pyrolized in a quartz reactor in the nitrogen atmosphere (nitrogen flow of 100 ml/min). The samples were heated to 800 ◦ C at a heating rate of 5 ◦ C/min and at 800 ◦ C for 2 h. 2.2. Characterization methods Thermogravimetric (TG) measurements were carried out in air using a Derivatograph C (MOM, Budapest) apparatus using 18–20 mg of samples placed in a ceramic crucible heated at a heating rate of 10 ◦ C/min. FTIR spectroscopy study was carried out in the 4000–400 cm−1 range (attenuated total reflectance, ATR, mode) using a FTIR Nicolet 8700A (ThermoScientific, USA) spectrophotometer equipped with a Diamond Smart Orbit ATR. SEM images were recorded using a JEOL JSM-6700F scanning electron microscope.

The textural characteristics of carbon materials were determined using low-temperature nitrogen adsorption–desorption isotherms recorded using a Micromeritics ASAP 2405N analyzer or a Quantachrome Autosorb analyzer. The specific surface area (SBET ) was calculated according to the standard BET method [8]. The total pore volume was estimated from the volume of nitrogen adsorbed at relative pressure p/p0 ≈ 0.98–0.99. The pore size distributions (PSD) were calculated using nitrogen adsorption–desorption isotherms with modified Nguyen-Do (MND) [21,36] method and a slitshaped pore model, as well as nonlocal density functional theory (NLDFT) and quenched solid DFT (QSDFT) methods (Quantachrome software, an equilibrium model with slitshaped and cylindrical pores). Additionally, the nitrogen desorption data were used to compute the PSD (differential fV (x) ∼ dVp /dx and fS (x) ∼ dS/dx) using a self-consistent regularization (SCR) procedure under nonnegativity condition (fV (x) ≥ 0 at any pore half-width x) at a fixed regularization parameter ˛ = 0.01 with a complex pore model with slit-shaped and cylindrical pores and voids between spherical nanoparticles (10–100 nm in size) packed in random aggregates (SCV) for RFR2-C [37]. The differential PSDs (with respect to the pore volume fV (x) ∼ dV/dx, ʃfV (x)dx ∼ Vp ) were re-calculated to the incremental PSDs (IPSD, ˚V,i (x) = Vp ). The differential fS (x) functions were used to estimate the deviation (w = SBET /ʃfS (x)dx − 1) of the pore shape from the model [38]. The fV (x) and fS (x) functions were also used to calculate contributions of nanopores (Vnano and Snano at 0.2–0.35 < x < 1 nm), mesopores (Vmeso and Smeso at 1 < x < 25 nm), and macropores (Vmacro and Smacro at 25 < x < 100 nm) to the total pore volume and the specific surface area. The 1 H NMR spectra were measured using a high resolution Varian 400 Mercury spectrometer (magnetic field of 9.4 T) with probing 90◦ pulses of 3 ␮s duration. Temperature of a sensor was regulated by means of a Bruker VT-1000 device with the accuracy of ±1 K. Relative mean errors were smaller than ±10% for 1 H NMR signal intensity for overlapped signals and ±5% for single signals. The accuracy was improved using digital treatment of signals with compensation of phase distortion and zero line nonlinearity with the same intensity scale at different temperatures [34,35]. Before 1 H NMR measurements, water was adsorbed on a char sample from air at atmospheric pressure and 60% relative humidity. To adsorb methane onto dry char pre-heated at 400 K for 20 min to remove water traces, a NMR ampoule of 5 mm in diameter containing 200 mg of a char was connected (by a flexible hose) to a methane reservoir at pressure higher by 0.1 kP than atmospheric one. Therefore, additional portions of methane could be added to the ampoule during cooling of a sample. Hydrogen adsorption was studied in the analogous way with a char containing 1 wt.% of water and preliminarily saturated with methane at 293 K. Quantum chemical calculations were carried out using ab initio method with the 6-31G(d,p) basis set using the Gaussian 03 [39] and WinGAMESS 12 [40,41] program suits or semiempirical PM7 method (MOPAC 2012 [42]) to full geometry optimization of molecules or their clusters. Solvation of different fragments of resorcinol–formaldehyde polymer was studied using the SMD method implemented in WinGAMESS [40,41]. For visualization of the fields around a RFR fragment, TorchLite 10 program [43,44] was used. The calculated structures were also visualized using the Chemcraft program [45].

3. Results and discussion 3.1. Structural and textural characteristics of RFR and RFR-C The resorcinol–formaldehyde polymer is formed due to condensation of resorcinol and formaldehyde in the aqueous solution. The aromatic rings of resorcinol can be connected with the

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

685

Table 1 Weight loss of polymers during heating up to 1000 ◦ C in air (to 180 ◦ C) and then in the nitrogen atmosphere, and bulk density (b ) of chars. Sample

Total weight loss (wt.%)

Weight loss (wt.%) in air at 20–180 ◦ C

Weight loss (wt.%) in nitrogen at 180–800 ◦ C

RFR-C b (g/cm3 )

RFR1 RFR2 RFR3

44.1 55.2 54.1

7.4 18.2 23.0

36.7 37.2 31.1

0.64 0.33 0.20

Table 2 Textural characteristics of chars (MND and SCV/SCR methods). Sample

SBET (m2 /g)

Snano (m2 /g)

Sмeзо (m2 /g)

Smacro (m2 /g)

Vp (cm3 /g)

Vnano (cm3 /g)

Vmeso (cm3 /g)

Vmacro (cm3 /g)

w

RFR1-C RFR2-C RFR2-Ca RFR3-C

0.2 623 623 570

– 553 479 569

– 67 128 1

– 3 16 0

– 0.963 0.963 0.300

– 0.271 0.223 0.294

– 0.624 0.473 0.003

– 0.068 0.267 0.003

0.385 0.120 0.202

a

SCV/SCR method.

methylene ( CH2 ) and ether ( CH2 O CH2 ) bridging groups [31]. The 3D RFR has two or three methylene bridges between each resorcinol ring. Polycondensation as well as morphological, structural and textural characteristics of polymers are most strongly affected by concentration of the reagents, especially water. These factors affect the morphological, textural and thermal features of not only the polymer but also the chars (Figs. 1–3). Nonporous RFR1-C (Fig. 1) looks as a monolith with a much smoother surface than that of nano/mesoporous RFR2-C (Fig. 2). However, nanoporous RFR3-C (Fig. 3) has a smoother surface than RFR2-C. The morphology of particles changes from smooth monolith (RFR1-C) to aggregated rough nanoparticles of 30–100 nm in size (RFR2-C) or larger smooth but weaker aggregated microparticles of 2–5 ␮m in size (RFR3-C). Thus, the morphological features of the chars are due to the difference in the water content on the polymer synthesis. Water at a relatively small content (∼38 wt.%) cannot play a role of a porogen preventing a high cross-linking degree of the polymer. Therefore, nonporous RFR1-C forms (Fig. 1). An increase in the water content to ∼63 wt.% provides the porogen effects and a certain re-arrangement of 3D RFR structure. The cross-linking degree of RFR2 and the bulk density of RFR2-C decrease (Table 1, b ), but the porosity increases (Table 2, Vp , SBET ). A subsequent increase in the water content (RFR3) results in a decrease in the mesoporosity of RFR3-C, but its nanoporosity and macroporosity increase, since the bulk density decreases. As

Fig. 1. SEM image of RFR1-C (magnification 103 ×).

mentioned above, the morphology and the texture of chars depend strongly on the water content during the RFR formation. RFR1-C corresponding to RFR prepared at a minimal water content (hydration degree h = 0.61 g of water per 1 g of resorcinol + formaldehyde) is monolithic (Fig. 1) and nonporous at SBET = 0.2 m2 /g. RFR2-C (h = 1.68 g/g) is nano/mesoporous but RFR3-C (h = 2.75 g/g) is practically purely nanoporous (Table 2) that are well seen from the shape of the nitrogen adsorption–desorption isotherms (Fig. 4), as well as the PSD (Fig. 5). Despite macropores at 25 < x < 100 nm giving a small contribution to the pore volume according to nitrogen adsorption (Table 2, RFR3-C), contribution of larger macropores at x > 100 nm is significant according to SEM images (Fig. 3) and the bulk density b decreases with increasing water content on RFR synthesis (Table 1). This result can be considered as a typical trend in the morphology and the texture of adsorbents since similar results were observed for cryogels for which water (ice) plays a role of the porogen, and their macroporosity strongly increases with increasing water content [46]. In the case of RFR3, h = 2.75 g/g that is enough to rearrange the polymers in more compacted structures (since RFR3-C includes mainly nanopores) than that in RFR2 (h = 1.68 g/g), since RFR2-C includes both nanopores and mesopores (Fig. 5 and Table 2). It should be noted that the deviation of the model of slitshaped pores (Table 2, w) is larger for RFR2-C than RFR3-C. Therefore, a more complex SCR/SCR method was also applied to RFR2-C (Table 2 and Fig. 5a). It gives a much smaller deviation w (Table 2, RFR2-C* ) than the slitshaped pore model. According to the SCR/SCR calculation, contribution of slitshaped pores into the pore volume (∼46%) of RFR2-C is slightly smaller than that of cylindrical pores (∼54%) but contribution of pores between relatively large nanoparticles (Fig. 2) is very low (∼0.01%). This calculation result is due to very poor filling of large macropores by nitrogen even at p/p0 ≈ 0.99. Theoretical calculations of the free energy of solvation of the RFR fragments in water (SMD/6-31G(d,p) method) give Gs,1 = −31 kJ/mol (Fig. 6a) and Gs,2 = −43 kJ/mol (Fig. 6b), i.e. the RFR hydration can be preferable than the formation of intermolecular bonds between neighboring macromolecules, since 2Gs,1 < Gs,2 . However, in the case of cross-linking of the neighboring polymers (Fig. 6c), water molecules are mainly displaced from intermolecular space. This occurs due to the hydrophobicity of aromatic rings interacting by dispersion interactions (Fig. 6d) and increasing confined space effects for water in more hydrophobic surroundings. Therefore, the spatial structure of RFR should depend strongly on the content of water during the polymer formation, as well as on the RFR cross-linking degree. Thus, an increase in the water content can decrease the cross-linking degree that affects

686

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

Fig. 2. SEM images of RFR2-C with magnification of (a) 103 ×, (b) 104 × and (c) 105 ×.

Fig. 3. SEM images of RFR3-C with magnification of (a) 103 ×, (b) 104 × and (c) 105 ×.

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

Fig. 4. Nitrogen adsorption–desorption isotherms for RFR2-C and RFR3-C.

Fig. 5. Pore size distributions calculated for (a) RFR2-C and (b) RFR3-C using NLDFT and QSDFT with an equilibrium model of slitshaped and cylindrical pores, MND with slitshaped pores, and SCV/SCR.

687

the porosity of the RFR chars. This conclusion is in agreement with the experimental results (vide infra). The whole field surrounding a molecule can be broken down into four types of interactions, or fields, which correspond to (i) predominantly hydrophobic interactions, (ii) van der Waals attraction forces, or electrostatic fields dominated by (iii) electropositive or (iv) electronegative atoms or ions. In this visualization approach each atom or a group of atoms is considered as a “field point” of one of these four types, and together field points for a continuous field. For example, in a hydrophobic phenyl group, the field point is at the center of the benzene ring. This approach has proved to be useful in analysis of crystal structures and drug discovery applications [43]. There are certain areas of RFR as sites for ␲–␲ bond interactions, which correspond to the van der Waals (vdW) and hydrophobic fields around nonpolar fragments modeled using the TorchLite program [44]. These patches of RFR exhibit a hydrophobic character (Fig. 6d). The hydrophilic (positive and negative) fields are mainly located around O-containing groups (OH and C O C). The hydrophilic field has a longer range (∼r−2 for charged groups) than vdW and hydrophobic fields (∼r−6 ) around aromatic rings. Therefore, water molecules tend to form the hydrate shell out of the space with the dominating hydrophobic and vdW fields (compare Fig. 6c and d). Intensive removal of water and residual formaldehyde is observed during heating of the polymer in the 20–150 ◦ C range with the maximum at 85 ◦ C (Fig. 7a). In this range, the mass loss was 7–23 wt.% for different samples, and it is the maximal for RFR2. Above 250 ◦ C, the thermooxidizing destruction of hydrocarbon structures is observed that runs in a few stages, according to the shape of the DTG curve. The RFR pyrolysis is accompanied by the mass loss of 44–55 wt.% (Table 1) with the major portion related to polymer carbonization. Complete polymer decomposition and 100% sample mass loss is at 650–680 ◦ C for porous RFR2 and RFR3 but it is at 880 ◦ C for nonporous RFR1. An increase in heating temperature of char up to 1000 ◦ C in the dynamic regime does not lead to the complete thermooxidation destruction (Fig. 7b). The weight loss was 23, 31 and 34 wt.% at heating to 700 ◦ C for RFR1-C, RFR2-C, and RFR3-C, respectively. One can suppose that “soft” (incoherent) structures of carbon material undergo destruction as the first ones in the presence of oxygen from air. A large part of the chars is composed of densely packed carbon structures (Figs. 1–3) whose oxidation by oxygen from air is difficult under used conditions. The IR spectrum (Fig. 8) of RFR2-C (as a representative sample) contains absorption bands in the 3750–3550 cm−1 range characteristic of the O H stretching vibrations of the hydroxyl groups. Additionally, the absorption bands at 2900, 1750, 1600, 1250, 1000 and 700 cm−1 can be observed. The band at 1750 cm−1 can be assigned to the C O bonds. The bands at 1000 and 1250 cm−1 are due to the C O and C C bonds. The band at 1600 cm−1 corresponds to the aromatic structures, and the band at 2900 cm−1 is due to the C H stretching vibrations. According to the literature [47–50], the carbonyl, lactone, phenyl, ether and carboxylic groups can be formed on the graphene cluster peripheries that can explain a complex character of the IR spectrum observed (Fig. 8). XRD patterns (not shown here) suggest that the chars are composed of very small crystallites and amorphous fragments; i.e. the materials are practically amorphous. However, according to the Raman spectra (Fig. 9), intensity of the G band (corresponding to sp2 C atoms in the polyaromatic graphene structures) is higher than that of the D band (disordered structures with sp3 C atoms) [51,52]. Notice that the G/D intensity ratio for the studied samples is greater than that for chars of phenol–formaldehyde resins [15]. There is no only a decrease in the G band intensity but also a change in location of the band peak with decreasing crystallinity of the materials [53–60]. Decomposition of the main G and D bands of RFR1-C (as a representative sample) gives the main G1

688

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

Fig. 6. (a) Two resorcinol molecules bonded by a bridge C CH2 C (formed with attached formaldehyde molecule) (HF/6-31G(d,p) geometry); (b) two fragments interacting (HF/6-31G(d,p) geometry), (c) cross-linked polymer fragments (with mainly C CH2 C bridges and a few C CH2 O CH2 C bridges) surrounded by water molecules (PM7 geometry), and (d) fields around a polymer fragment without water molecules.

and D1 bands and additional lines G2 and D2 (Fig. 9b and Table 3). The band G2 can be assigned to amorphous graphite-like structures [53], which are out of graphene planes. There is no a unique assignment of the band D2 at ∼1100–1200 cm−1 [53–60]. It can be assigned to mixed sp2 –sp3 bonding or to the C C and C C stretching vibration modes of polyene-like structures [60], or structures with sp3 C atoms in defects out of the planes. The Raman spectra (Fig. 9) show that there are no linear correlation between the water content in the reaction medium, affecting the RFR and RFR-C morphology, and the band decomposition results (Table 3). This is due to several factors such as the crosslinking degree affecting ordering of structures in the polymers and the chars, different contributions of pores of different types affecting sheet–sheet interactions in chars. The studied structural and textural features of porous RFR2-C and RFR3-C samples affect the interfacial phenomena in the adsorption layer of water and other adsorbates. 3.2. Interfacial phenomena studied by low-temperature 1 H NMR spectroscopy Water was observed in the 1 H NMR spectra of weakly hydrated (h = 0.05 g/g) RFR2-C (Fig. 10a, dashed lines) as a single signal with

intensity stable at T > 250 K and decreased at very low temperature due to partial freezing of water. The chemical shift of water (ıH ) is observed in the 1–2 ppm range and increases with decreasing temperature. These chemical shifts are much smaller than that of liquid water (4–5 ppm). According to the literature analyzed in the monograph [34], the upfield shift of 1 H NMR signal of substances adsorbed onto carbons depends on the shielding effects of ring currents and local magnetic anisotropy of carbon sheets, especially in nanopores where molecules are more strongly affected by the magnetic shielding of the ring current of both pore walls. The total effect of the polyaromatic system is equivalent to the current flowing on the outer contour (the second Kirchhoff law for the electric circuits), since the currents flow in the opposite directions on the boundary rings. Independently of location of molecules adsorbed on the carbon sheets in pores, the shielding effect is determined by the size, distance of adsorbed molecule to the surface, wall thickness (number of graphene layers) and the distance to the opposite wall (i.e. pore width). Therefore, it is maximal in narrow slitshaped nanopores characteristic for chars RFR2-C and RFR3-C (Fig. 5 and Table 2). If the chemical shift ıH for the bulk water is equal to 5 ppm [34], that the shielding effect for water adsorbed onto RFR2C results in the upfield shift at −ıH = 4–7 ppm. Similar shielding effects are also observed for adsorbed methane (Fig. 10).

Table 3 Decomposition results for the main bands of Raman spectra of chars. Sample

D1

D2

G1

G2

FWHMD1

FWHMG1

AD1 /AG1

RFR1-C RFR2-C RFR3-C

1341 1342 1340

1186 1215 1094

1597 1590 1589

1527 1517 1520

175 170 188

58 65 68

3.32 2.58 2.83

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

689

Fig. 7. TG and DTG thermograms of pyrolysis of (a) resorcinol–formaldehyde polymers and (b) chars.

Large changes in the chemical shift of water with temperature can be caused by changes in the association degree of water confined in different pores and frozen, therefore, at different temperatures, its translocation into narrower pores possessing much higher adsorption potential, and location of the molecules directly at pore walls. If the temperature dependence of the chemical shift is assumed to be linked to the diffusion of the molecules toward pores with a higher adsorption potential (i.e. to narrower pores)

Fig. 9. (a) Raman spectra (normalized to unit) of chars, and (b) decomposition of the Raman spectrum of RFR1-C.

that one may expect the −ıH value stability or its small increase due to a decrease in the distance between the opposite walls in narrower pores. Thus, the most probable reason for a decrease in −ıH with decreasing temperature is changes in location of mobile water molecules in pores and the effects of formed ice on the topology of free space in pores, as well as the pore size because of a larger volume of ice in comparison with liquid water [34]. As follows from the data (Fig. 10), water completely freezes in pores of RFR2-C at relatively low temperatures. A decrease in the freezing point of water located in pores can be used to estimate the size distribution of unfrozen water structures filling these pores using the Gibbs–Thomson (GT) equation for the freezing point depression of liquid confined in pores [34,61] Tm = Tm (R) − Tm,∞ =

Fig. 8. IR spectrum (ATR) of RFR2-C.

2sl Tm,∞ kGT = R Hf R

(1)

where Tm (R) is the melting temperature of frozen liquid in pores of radius R, Tm,∞ the bulk melting temperature,  the density of the solid,  sl the energy of solid–liquid interaction, and Hf the bulk enthalpy of fusion. The size distribution of water clusters bound in pores of RFR2-C (Fig. 11) includes two types of structures corresponding two types of pores. The PSD based on the cryoporometry data is in agreement with the PSD calculated using the nitrogen adsorption isotherm. However, a small content of adsorbed water causes filling only

690

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

Fig. 11. The size distribution of adsorbed water clusters in pores of RFR2-C (NMR cryoporometry) and SCV/SCR PSD for this carbon.

Fig. 10. 1 H NMR spectra (taken at different temperatures) of (a) water (dashed lines, h = 0.05 g/g) and methane (solid lines), (b) a mixture of water (h = 0.01 g/g) and methane, and (c) a mixture of water (h = 0.01 g/g), methane and hydrogen adsorbed on RFR2-C.

narrow pores, since pores at x > 10 nm are practically free of water (Fig. 11). Comparing intensity of water and methane signals (ICH4 /IH2O = 5.2/1.4, Fig. 10), it is possible to estimate the amount of methane adsorbed in the presence of 1% pre-adsorbed water. Taking into account twice as large number of protons in the methane molecule and slightly smaller molecular mass, the adsorption of methane can be estimated as 1.7 wt.%. Hydrogen adsorption is insignificant (Fig. 10b and c). Therefore, hydrogen signal at ıH = 4 ppm is observed only at low temperatures when water signal intensity drops down due to freezing. Comparison of intensity of hydrogen signals with other adsorbates shows that its amount does not exceed hundredth parts of percent. Unusual effect appears for methane adsorbed onto RFR2-C in the presence of a fixed amount of water (Fig. 10). Since methane does not freeze in the temperature range used, one may expect a stable 1 H NMR signal intensity of methane with temperature. Typically, a certain increase in signal intensity of adsorbates could be caused by changes in the population of nuclear levels (Curie law [62]). However, a decrease in signal intensity of methane is observed with decreasing temperature. This may be due to an increase in intensity of relaxation processes. Acceleration of nuclear spin relaxation with increasing temperature for conducting materials (such as carbons studied here according to electrophysical data [63]) can be connected with an increase in conductivity that can affect the relaxation processes. When adsorbed benzene concentration is small (smaller than the volume of nanopores, Fig. 12a), two signals with the chemical shifts of 1 and −1 ppm (shielding effect −6 to −8 ppm) are registered at 290 K. With decreasing temperature, both signals demonstrate the downfield shift and they are broadened in such way that one signal with the chemical shift of 4 ppm is observed at 200 K. Thus, benzene, as well as water and methane, can be located not only at the pore walls but also in a central part of narrow pores (forming small clusters there) or broader pores (forming domain there) where the surface shielding effects decrease with pore size increasing. A significant decrease in benzene signal intensity with decreasing temperature is due to partial freezing of benzene. A large value of the shielding effect indicates that benzene molecules responsible for these signals are located in narrow slitshaped pores. One can expect that at a given concentration, benzene will adsorb mainly in nanopores (Figs. 5 and 11). With increasing benzene concentration (Fig. 12a) intensive signal of adsorbed benzene appears at ıH ≈ 7.5 ppm. It slightly changes with decreasing temperature. The shielding effect of the surface is small that is due to a nano/mesoporous character of RFR2C discussed above. When the whole pore space is filled with benzene (Fig. 12b), intensity of the downfield signal increases and the

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

691

Fig. 12. 1 H NMR spectra (recorded at different concentrations and temperatures) of water co-adsorbed with (a and b) benzene and (b) methylene chloride onto RFR2-C.

up-field signal is observed only at relatively high temperatures. Most of benzene filling the pores is weakly bound because it freezes at temperatures close to the freezing point. In this case, a relatively narrow signal of unfrozen benzene forming small clusters is observed in the spectra. Analogous situation can be observed with filling pore space with liquid methylene chloride whose freezing point is about 180 K (Fig. 12b). There is registered one asymmetric or two overlapping signals at ıH ≈ 5 ppm. A weak signal at ıH ≈ 0 ppm corresponds to adsorbate located in nanopores and narrow mesopores that causes the upfield shift. To elucidated the nanoporosity effects on the interfacial behavior of adsorbates, water (h = 0.05 g/g), benzene and methane were adsorbed onto RFR3-C (Fig. 13) which has mainly nanopores (Fig. 5 and Table 2). Strong shielding effects in nanopores [34,64] cause the appearance of water signal at −6 ppm, i.e. ıH ≈ −11 ppm, whose width increases with temperature. Methane is poorly adsorbed onto RFR3-C (Fig. 13) in comparison with ACs characterized by more developed porosity [34,64]. Adsorbed methane is observed as a low-intensive narrow signal at 0 ppm; i.e. it cannot penetrate into nanopores where the upfield shift should be observed.

Fig. 13. 1 H NMR spectra at different temperatures of adsorbates interacting with RFR3-C: water at h = (a) 0.05 (solid lines) and 0.2 (dashed lines), (b) 1.0 (solid) and 0.15 (dashed), and (c) 0.03 g/g in different medium (a) CH4 , (b) air (solid) and C6 H6 (dashed), and (c) DMSO.

692

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693

At water content h = 0.2 g/g (smaller than the pore volume, Table 3), there is signal of bulk water at 5 ppm which is frozen close to 273 K. Signal of water located in pores demonstrates the downfield shift in comparison with sample at h = 0.05 g/g (Fig. 13a). Consequently, a portion of water at h = 0.2 g/g is located in broader pores or out of pores in voids between particles (Fig. 2). At h = 1 g/g (it is greater than the pore volume), the behavior of water in nanopores changes insignificantly; however, signal of bulk water at 5 ppm increases. The 1 H NMR spectra of RFR3-C at h = 0.15 g/g in the benzene dispersion medium (Fig. 13b) show that benzene weakly penetrates into narrow pores since the main signal at 7.2 ppm corresponding to liquid benzene is observed only at 280 K. DMSO (containing about 3 wt.% of water) adsorbed onto dried RFR3-C can penetrate into pores since signal of CH3 groups at 2.5 ppm is observed at low temperatures at T < Tm = 292 K (Fig. 13c). Signal of water dissolved in DMSO is observed at T > 240 K at 3.5 ppm. With decreasing temperature and freezing a portion of DMSO, the chemical shift of water increases because of the formation of strongly associated water domains between freezing DMSO structures [34,65]. 4. Conclusion Chars with the specific surface area ≈600 m2 /g and the total pore volume 0.3–1 cm3 /g were prepared using resorcinol–formaldehyde resin synthesized at different content of water. The pore size distribution in chars depends on features of the synthesis of the polymer, mainly the water content. Thus, changes in the water content in the reaction mixture with resorcinol and formaldehyde results in the formation of the polymers and chars with very different textural and morphological characteristics. At a minimal content of water (∼38 wt.% in the initial solution), polymers are maximum cross-linked that results in the formation of practically nonporous char with the bulk density of 0.64 g/cm3 . At the water content of ∼63 wt.%, char is nano/mesoporous. At the water content of ∼73 wt.%, char is composed of purely nanoporous microparticles which are weakly aggregated since the bulk density is low (0.2 g/cm3 ) due to large voids (macropores) between microparticles. From the 1 H NMR spectra of water co-adsorbed with methane, benzene, hydrogen, methylene chloride, and DMSO one can conclude that nanopores can be poorly accessible in chars for organic co-adsorbates if water (even in a small amount) was pre-adsorbed and blocked the entrance into inner nanopores of microparticles. The adsorbed water forms polyassociates (clusters, domains) whose minimal size is larger than that of nanopores, i.e. water fills mesopores in which the formation of clusters and domains is energetically more advantageous than in nanopores, especially due to interactions of water with O-containing functionalities. Acknowledgments The work was supported by the EC Seventh Framework Programme (FP7/2007–2013), Marie Curie International Research Staff Exchange Scheme (grant no. 230790, COMPOSITUM). References [1] M. Smisek, S. Cerny, Active Carbon, Elsevier, Amsterdam, 1970. [2] R.C. Bansal, J.B. Donnet, F. Stoeckli, Active Carbon, Marcel Dekker, New York, 1988. [3] B. McEnaney, T.J. Mays, F. Rodriguez-Reinoso (Eds.), Fundamental Aspects of Active Carbons, Special issue, Carbon 36 (10) (1998). [4] D.O. Cooney, Activated Charcoal in Medical Applications, Marcel Dekker, New York, 1995.

[5] C. Liang, Z. Li, S. Dai, Mesoporous carbon materials: synthesis and modification, Angewandte Chemie 47 (2008) 3696–3717. [6] Y. Li, Z.-Y. Fu, B.-L. Su, Hierarchically structured porous materials for energy conversion and storage, Advanced Functional Materials 22 (2012) 4634–4667. [7] Y. Zhai, Y. Dou, D. Zhao, P.F. Fulvio, R.T. Mayes, S. Dai, Carbon materials for chemical capacitive energy storage, Advanced Materials 23 (2011) 4828–4850. [8] S.J. Gregg, K.S.W. Sing, H.F. Stoeckli (Eds.), Characterization of Porous Solids, Soc. Chem. Industry, London, 1979. [9] S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and Porosity, 2nd ed., Academic Press, London, 1982; Studies in Surface Science and Catalysis, in: F. Rodriguez-Reinoso, B. McEnaney, J. Rouquerol, K. Unger (Eds.), Characterisation of Porous Solids VI, vol. 144, Elsevier Science, Amsterdam, 2002. [10] M. Inagaki, Pores in carbon materials-importance of their control, New Carbon Materials 24 (2009) 193–232. [11] A. Stein, Z. Wang, M.A. Fierke, Functionalization of porous carbon materials with designed pore architecture, Advanced Materials 21 (2009) 265–293. [12] V.M. Gun’ko, V.V. Turov, O.P. Kozynchenko, D. Palijczuk, R. Szmigielski, S.V. Kerus, P.P. Gorbik, Characteristics of adsorption phase with water/organic mixtures at a surface of activated carbons possessing intraparticle and textural porosities, Applied Surface Science 254 (2008) 3220–3231. [13] V.M. Gun’ko, V.V. Turov, O.P. Kozynchenko, V.G. Nikolaev, S.R. Tennison, S.T. Meikle, E.A. Snezhkova, A.S. Sidorenko, F.Ehrburger-Dolle, I. Morfin, D.O. Klymchuk, S.V. Mikhalovsky, Activation and structural and adsorption features of activated carbons with highly developed micro-, meso- and macroporosity, Adsorption 17 (2011) 453–460. [14] V.M. Gun’ko, S.T. Meikle, O.P. Kozynchenko, S.R. Tennison, F. Ehrburger-Dolle, I. Morfin, S.V. Mikhalovsky, Comparative characterization of carbon and polymer adsorbents by SAXS and nitrogen adsorption methods, Journal of Physical Chemistry C 115 (2011) 10727–10735. [15] V.M. Gun’ko, V.M. Bogatyrov, O.I. Oranska, M.V. Galaburda, E.V. Polshin, I.V. ˛ Urubkov, R. Leboda, J. Skubiszewska-Zieba, B. Charmas, Effect of nanosilica on characteristics of carbonizates of phenol–formaldehyde resin – Fe(acac)3 , Applied Surface Science 264 (2013) 707–712. [16] M.T. Izquierdo, A. Marthnez de Yuso, R. Valenciano, B. Rubio, M.R. Pino, Influence of activated carbon characteristics on toluene and hexane adsorption: application of surface response methodology, Applied Surface Science 264 (2013) 335–343. [17] G. Xiao, L. Long, Efficient removal of aniline by a water-compatible microporous and mesoporous hyper-cross-linked resin and XAD-4 resin: a comparative study, Applied Surface Science 258 (2012) 6465–6471. [18] Y. Kong, T. Qiu, J. Qiu, Fabrication of novel micro–nano carbonous composites based on self-made hollow activated carbon fibers, Applied Surface Science 265 (2013) 352–357. [19] R. Pietrzak, P. Nowicki, H. Wachowska, The influence of oxidation with nitric acid on the preparation and properties of active carbon enriched in nitrogen, Applied Surface Science 255 (2009) 3586–3593. ˛ [20] K. László, P. Podko´scielny, A. Dabrowski, Heterogeneity of activated carbons with different surface chemistry in adsorption of phenol from aqueous solutions, Applied Surface Science 252 (2006) 5752–5762. [21] P. Kowalczyk, V.M. Gun’ko, A.P. Terzyk, P.A. Gauden, H. Rong, Z. Ryu, D.D. Do, The comparative characterization of structural heterogeneity of mesoporous activated carbon fibers (ACF), Applied Surface Science 206 (2003) 67–77. [22] T. Horikawa, J. Hayashi, K. Muroyama, Size control and characterization of spherical carbon aerogel particles from resorcinol–formaldehyde resin, Carbon 42 (2004) 169–175. [23] T. Horikawa, Y. Ono, J. Hayashi, K. Muroyama, Influence of surface-active agents on pore characteristics of the generated spherical resorcinol–formaldehyde based carbon aerogels, Carbon 42 (2004) 2683–2689. [24] A. Léonard, N. Job, S. Blacher, J.-P. Pirard, M. Crine, W. Jomaa, Suitability of convective air drying for the production of porous resorcinol–formaldehyde and carbon xerogels, Carbon 43 (2005) 1808–1811. [25] L. Zhang, H. Liu, M. Wang, L. Chen, Structure and electrochemical properties of resorcinol–formaldehyde polymer-based carbon for electric double-layer capacitors, Carbon 45 (2007) 1439–1445. [26] J. Xu, A. Wang, T. Zhang, A two-step synthesis of ordered mesoporous resorcinol–formaldehyde polymer and carbon, Carbon 50 (2012) 1807–1816. [27] A.H. Moreno, A. Arenillas, E.G. Calvo, J.M. Bermúdez, J.A. Menéndez, Carbonisation of resorcinol–formaldehyde organic xerogels: effect of temperature, particle size and heating rate on the porosity of carbon xerogels, Journal of Analytical and Applied Pyrolysis 100 (2013) 111–116. [28] A. Szczurek, G. Amaral-Labat, V. Fierro, A. Pizzi, E. Masson, A. Celzard, Porosity of resorcinol–formaldehyde organic and carbon aerogels exchanged and dried with supercritical organic solvents, Materials Chemistry and Physics 129 (2011) 1221–1232. [29] S. Tanaka, T. Yasuda, Y. Katayama, Y. Miyake, Pervaporation dehydration performance of microporous carbon membranes prepared from resorcinol/formaldehyde polymer, Journal of Membrane Science 379 (2011) 52–59. [30] A.M. Elkhatat, S.A. Al-Muhtaseb, Advances in tailoring resorcinol–formaldehyde organic and carbon gels, Advanced materials 23 (2011) 2887–2903. [31] S.A. Al-Muhtaseb, J.A. Ritter, Resorcinol–formaldehyde organic and carbon gels, Advanced Materials 15 (2003) 101–114. [32] R.W. Pekala, Organic aerogels from the polycondensation of resorcinol with formaldehyde, Journal of Materials Science 24 (1989) 3221–3227.

V.M. Gun’ko et al. / Applied Surface Science 283 (2013) 683–693 [33] N. Job, F. Panariello, M. Crine, J.-P. Pirard, A. Léonard, Rheological determination of the sol–gel transition during the aqueous synthesis of resorcinol–formaldehyde resins, Colloids and Surfaces A: Physicochemical and Engineering Aspects 293 (2007) 224–228. [34] V.M. Gun’ko, V.V. Turov, Nuclear Magnetic Resonance Studies of Interfacial Phenomena, CRC Press, Boca Raton, 2013. [35] V.M. Gun’ko, V.V. Turov, V.M. Bogatyrev, V.I. Zarko, R. Leboda, E.V. Goncharuk, A.A. Novza, A.V. Turov, A.A. Chuiko, Unusual properties of water at hydrophilic/hydrophobic interfaces, Advances in Colloid and Interface Science 118 (2005) 125–172. [36] V.M. Gun’ko, D.D. Do, Characterization of pore structure of carbon adsorbents using regularization procedure, Colloids and Surfaces A: Physicochemical and Engineering Aspects 193 (2001) 71–83. [37] V.M. Gun’ko, Consideration of the multicomponent nature of adsorbents during analysis of their structural and energy parameters, Theoretical and Experimental Chemistry 36 (2000) 319–324. [38] V.M. Gun’ko, S.V. Mikhalovsky, Evaluation of slitlike porosity of carbon adsorbents, Carbon 42 (2004) 843–849. [39] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman, J.A. Montgomery Jr., T. Vreven, K.N. Kudin, J.C. Burant, J.M. Millam, S.S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G.A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J.E. Knox, H.P. Hratchian, J.B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev, A.J. Austin, R. Cammi, C. Pomelli, J.W. Ochterski, P.Y. Ayala, K. Morokuma, G.A. Voth, P. Salvador, J.J. Dannenberg, V.G. Zakrzewski, S. Dapprich, A.D. Daniels, M.C. Strain, O. Farkas, D.K. Malick, A.D. Rabuck, K. Raghavachari, J.B. Foresman, J.V. Ortiz, Q. Cui, A.G. Baboul, S. Clifford, J. Cioslowski, B.B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R.L. Martin, D.J. Fox, T. Keith, M.A. Al-Laham, C.Y. Peng, A. Nanayakkara, M. Challacombe, P.M.W. Gill, B. Johnson, W. Chen, M.W. Wong, C. Gonzalez, J.A. Pople, Gaussian 03, in: Revision E.01, Gaussian, Inc., Wallingford, CT, 2004. [40] M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S. Gordon, J.J. Jensen, S. Koseki, N. Matsunaga, K.A. Nguyen, S. Su, T.L. Windus, M. Dupuis, J.A. Montgomery, General atomic and molecular electronic structure system, Journal of Computational Chemistry 14 (1993) 1347–1363. [41] M.S. Gordon, M.W. Schmidt, in: C.E. Dykstra, G. Frenking, K.S. Kim, G.E. Scuseria (Eds.), Theory and Applications of Computational Chemistry the First Forty Years, Elsevier, Amsterdam, 2005, pp. 1167–1189. [42] J.J.P. Stewart, MOPAC, Stewart Computational Chemistry, Colorado Springs, CO, USA, 2012 http://openmopac.net/ (06.01.13). [43] T. Cheeseright, M. Mackey, S. Rose, A. Vinter, Molecular field extrema as descriptors of biological activity: definition and validation, Journal of Chemical Information and Modeling 46 (2006) 665–676. [44] TorchLite 10, www.cresset-group.com (accessed 04.03.13). [45] G.A. Zhurko, D.A. Zhurko, Chemcraft (version 1.7, build 375), http://www. chemcraftprog.com [46] V.M. Gun’ko, I.N. Savina, S.V. Mikhalovsky, Cryogels: morphological, structural and adsorption characterisation, Advances in Colloid and Interface Science 187–188 (187) (2013) 1–46.

693

[47] H.P. Bochem, Infra-red investigation of carbon surface, Advanced Catalysis and Related Subjects 16 (1966) 179–274. [48] O. Ito, H. Seki, M. Iino, Diffuse reflectance spectra in near i.r. region of coals; a new index for degrees of coalification and carbonization, Fuel 67 (1988) 573–578. [49] K.R. Ko, C.G. Kim, S.K. Ryu, Functional groups of carbon fibers and activated carbon fibers treated with ozone, Materials 23rd Biennial Conference on Carbon (Carbon 97), Penn. State; 18–23 July, 1977, pp. 264–265. [50] B.J. Meldrum, C.H. Rochester, The study in situ of oxidation of the surface of active carbon in oxygen and carbon oxide by IR spectroscopy, Journal of the Chemical Society, Faraday Transactions 86 (1990) 861–865. [51] F. Tuinstra, J.L. Koenig, Raman spectrum of graphite, Journal of Chemical Physics 53 (1970) 1126–1130. [52] C.S. Sharma, D.K. Upadhyay, A. Sharma, Controlling the morphology of resorcinolformaldehyde based carbon xerogels by sol concentration, shearing and surfactants, Industrial & Engineering Chemistry Research 48 (2009) 8030–8036. [53] T.-H. Ko, W.-S. Kuo, Y.-H. Chang, Raman study of the microstructure changes of phenolic resin during pyrolysis, Polymer Composites 21 (2000) 745–750. [54] G.A. Zickler, B. Smarsly, N. Gierlinger, H. Peterlik, O. Paris, A reconsideration of the relationship between the crystallite size L-a of carbons determined by X-ray diffraction and Raman spectroscopy, Carbon 44 (2006) 3239–3246. [55] A. Manivannan, M. Chirila, N.C. Giles, M.S. Seehra, Microstructure, dangling bonds and impurities in activated carbons, Carbon 37 (1999) 1741–1747. [56] B.M. Vogelaar, A.D. van Langeveld, S. Eijsbouts, J.A. Moulijn, Fuel 86 (2007) 1122–1129. ´ [57] A. Wesełucha-Birczynska, K. Babeł, K. Jurewicz, Carbonaceous materials for hydrogen storage investigated by 2D Raman correlation spectroscopy, Vibrational Spectroscopy 60 (2012) 206–211. [58] C. Pardanaud, C. Martin, P. Roubin, G. Giacometti, C. Hopf, T. Schwarz-Selinger, W. Jacob, Raman spectroscopy investigation of the H content of heated hard amorphous carbon layers, Diamond and Related Materials 34 (2013) 100–104. [59] N. Shimodaira, A. Masui, Raman spectroscopic investigations of activated carbon materials, Journal of Applied Physics 92 (2002) 902–909. [60] B. Dippel, H. Jander, J. Heintzenberg, NIR FT Raman spectroscopic study of flame soot, Physical Chemistry Chemical Physics 1 (1999) 4707–4712. [61] J.H. Strange, J. Mitchell, J.B.W. Webber, Pore surface exploration by NMR, Magnetic Resonance Imaging 21 (2003) 221–226. [62] A. Abragam, The Principles of Nuclear Magnetism, Oxford University Press, Oxford, UK, 1961. [63] S.N. Makhno, V.M. Bogatyrov, E.I. Oranska, V.M. Gun’ko, P.P. Gorbyk, R. Leboda, ˛ Electrophysical properties of carbon-inorganic nanoJ. Skubiszewska-Zieba, composites C/Mx Oy /SiO2 , Surface 4 (2012) 186–192 (in Russian). 1 [64] V.V. Turov, R. Leboda, H NMR spectroscopy of adsorbed molecules and free surface energy of carbon adsorbents, Chemistry and Physics of Carbon 27 (2000) 67–124. [65] V.M. Gun’ko, V.V. Turov, V.N. Barvinchenko, A.A. Turova, A.A. Rugal, V.I. Zarko, R. Leboda, Nonuniformity of starch/nanosilica composites and interfacial behaviour of water and organic compounds, Applied Surface Science 256 (2010) 5275–5280.