Porous Fe@C Composites Derived from Silkworm Excrement for Effective Separation of Anisole Compounds.

Silkworm excrement is a very useful biomass waste, composed of layer-structured fats and proteins, which are great precursors for carbon composite materials. In this work, new porous composites derived from silkworm excrement were prepared for selective separation of flavor 4-methylanisole from the binary 4-methylanisole/4-anisaldehyde mixture. In particular, the silkworm excrement, possessing a unique nanosheet structure, is converted into a graphite-like carbon by a simple calcination strategy followed by a metal-ion-doping procedure. This Fe@C composite exhibits a special nano-spongy morphology, anchoring Fe3C/Fe5C2 on the carbon nanosheets. Density functional theory simulations showed that 4-methylanisole presents a stronger π-π interaction and attraction forces with sp2 carbon nanosheets in Fe@C composites than 4-anisaldehyde. The selective adsorption experiments further confirmed that the Fe@C composites exhibited a 4-methylanisole capacity of 7.3 mmol/g at 298 K and the highest selectivity of 17 for an equimolar 4-methylanisole/4-anisaldehyde mixture among the examined adsorbents including MOFs and commercial activated carbon materials, which demonstrates the potential of this low-cost and eco-friendly porous carbon material as a promising sustainable adsorbent.

Introduction

Anise compounds are popular flavor gradients un class="Chemical">sed as additives in the food industry and medical intermediates,[1] mainly containing anisaldehyde, anisole, anethole, estragole, and 4-methylanisole, and so forth.[2] Some of these species are high value-added flavor compounds, which are usually found in small amounts in natural plants or crop oils, whereas pure anisole has been explored as a very sustainable solvent for the hydroformylation process.[3] However, the complex composition and low concentration have posed difficulties for their industrial extraction and purification.[4,5] Separation of each species is even more challenging considering that some of them have similar sizes and chemical properties, which is normally carried out by the liquid–liquid extraction coupled with distillation methods.[6,7]

Certainly, more energy-efficient separation procesn class="Chemical">ses including adsorption and membrane separation are preferred technologies because of their ease of operation, high efficiency, and low energy consumption.[8] In adsorption, porous materials are the key factors determining the extraction and separation efficiency.[9] Because of this critical role, developing new porous adsorbents with superior capacity and fast kinetics with targeted species has gained considerable interest in the scientific community[10] as well as in industrial applications.[11,12] Among these, porous carbon materials and zeolites, because of their good chemical and thermal stability, food-grade safety, and cost-effective nature, have been widely used for purification of flavored chemicals.[13−15] However, improvements of adsorption capacity and selectivity for targeted flavor molecules are always needed. For example, Xiao[16] employed an anisole-modified hyper-crosslinked polystyrene resin and almost doubled the adsorption capacity of vanillin, that is, from 140.5 to 240.3 mg/g. Medellin-Castillo[17] studied ozone-modified activated carbon and found enhanced adsorption ability toward diethyl phthalate. Besides the adsorption capacity, improvement of adsorption selectivity has also been reported. Molecular imprinting and affinity separation have been used to recognize target molecules and realize great selectivity for larger biomolecules such as peptides and proteins.[18] However, separation of anisole or anisaldehyde from anise is still very challenging because ofo their similar molecular size, high boiling points, and diverse concentrations in industrial processes. As far as we are aware, there are rarely published reports about this separation.

Silkworm exn class="Chemical">crement (SE) is an abundant agricultural waste produced from the silk reeling industry in China and Southeast Asia.[19] Most of the waste is discarded and its improper disposal can cause severe environmental issues.[20] SE is mainly composed of cellulose and proteins with an ordered structure,[21] in which a high content of carbon with a unique nanosheet morphology has been found.[22] In this work, a “one-pot” calcination of the SE can produce a nano-spongy porous carbon structure loaded with iron ions. The obtained Fe@C composites demonstrated high adsorption selectivity toward 4-methylanisole from the 4-methylanisole/4-anisaldehyde (MA/AY) mixture, in which the synergistic effects of pore size and functional groups have played an import role, causing a strong interaction with MA through the possible sp2 C π–π stacking and weak polar interactions.

Experimental Section

Materials

SE was purchan class="Chemical">sed from a farmer’s market (Yizhou, China). MA (99.0%) and AY (99.0%) were supplied by Guangdong Hua Fang spice Co. Ltd. (Guangzhou, China). Analytical grade absolute ethanol (C2H5OH, 99.7%), FeCl2·4H2O (99.0%), and ZnCl2 (99.0%) were supplied by Aladdin Industrial Co. Ltd. (Shanghai, China). All the other materials were commercially available and were purchased at analytical grade and used directly without further purification. Chemical properties of MA and AY are shown in Table S1.

Synthesis of Fe@C Composites

SE was first washed in deionized (DI) n class="Chemical">water to remove ash from the raw materials and then dried at 130 °C for 8 h. Then, it was crushed into small particles with an average diameter of 1–2 mm. After that, the crushed SE particles (10.0 g) were immersed in a water solution of mixed ZnCl2 and FeCl2 for 12 h, and then freeze-dried to maintain the three-dimensional structure. The frozen Fe-doped SE was heated to 1200 °C at a rate of 5 °C/min and held at this temperature for 1.0 h under N2 atmosphere. After cooling down to room temperature, the final samples were washed using 10 mmol/L HCl solution to remove any loosely bonded species. More DI water was then used to wash the samples until pH 7. These samples were designated as Fe@C-1 and Fe@C-2, representing weight ratios of Fe/Zn at 1.0/1.0 and 2.0/1.0, respectively. For comparison, calcined SE without Fe loading was also prepared and named as SE-C.

Material Characterization

Fourier transform infrared (FTIR) spectra were recorded with a Bruker TENSOR II FTIR spectrophotometer in a range of 4000–400 cm–1. Powder X-ray diffraction (PXRD) measurements were performed on an X-ray spectrometer (Rigaku) with Cu Kα radiation (λ = 1.5406 Å). Raman spectra were carried out on a min class="Chemical">croscopic confocal Raman spectrometer (Renishaw, RW1000-inVia) with an excitation of laser light at 514.5 nm. Specific surface area and pore-size distribution were evaluated through nitrogen adsorption isotherms at 77 K on a Micromeritics ASAP 2460. Transmission electron microscopy (TEM) images were collected on a Cs-corrected ETEM (Titan ETEM G2 USA) at 300 kV acceleration voltages, whereas surface morphology was observed using a Hitachi S-3400N scanning electron microscope equipped with an energy dispersive X-ray spectrometer (Tokyo, Japan). Before the TEM measurement, Fe-containing carbon samples (Fe@C-1 and Fe@C-2) were washed with HCl to remove the Fe completely. Elemental compositions of the examined samples were analyzed by X-ray photoelectron spectroscopy (XPS, Thermo Fisher, USA).

Adsorption Equilibrium and Kinetics

For the adsorption experiments, single-component solutions were prepared by dissolving a certain amount of MA or AY in absolute ethanol and the batch adsorption was conducted using a thermostatic oscillator at 250 rpm. For the single-component kinetin class="Chemical">cs study, 50 mL of MA or AY solution (12.0 mmol/L) was transferred into a 100 mL conical flask, continuously shaken on a thermostatic oscillator, and incubated at 298 K for 20 min. Then, 50 mg of adsorbents were added into the flask, and the concentration of the solution components was determined at the given intervals of time until equilibrium was reached. The binary adsorption kinetics of MA and AY were performed following the same protocol at a 1:1 ratio of MA/AY. The equilibrium adsorption capacity qe (mmol/g), transient fractional adsorption capacity q (mmol/g), and selectivity (αMA/AY) were calculated according to eqs –3, respectively.where C0, Ce, and C (mmol/L) are the concentrations of the examined component at the initial condition, the equilibrium condition, and a time t (min) after adsorption started, respectively; V is the volume of the solution (L), and m is the weight of the dried adsorbents (g); C and C are the concentrations of MA and AY (mmol/g) at adsorption time t, respectively; q and q are the adsorbed amounts of MA and AY at time t, respectively.

The concentrations of MA and AY were analyzed by a gas chromatography instrument, installed with a HP-INNOWAX column (30 m × 0.32 mm×0.25 μm) and a flame ionization detector (FID). Samples were injected into a 1.0 μL loop, at a N2 gas flow of 24 mL/min. The temperatures of the oven and FID were 180 and 300 °C, respectively. Under then class="Chemical">se conditions, linear regression of MA and AY concentrations (y1 and y2, mmol/L) and peak area (x1 and x2) was performed to obtain a standard curve: y1 = 189599x1 + 763620 (R2 = 0.999, 10.0–80.0 mmol/L) and y2 = 147165x2 – 623200 (R2 = 0.999, 10.0–80.0 mmol/L). All measurements were performed in triplicate and average values were reported.

Simulation Method

Calculations were performed via density functional theory (DFT) following the gradient-corrected hybrid density functional B3LYP method.[23] Chemical structures of MA and AY were optimized with the polarized continuum model and dielectric constant of n class="Chemical">ethanol condition.[24] All calculations were conducted with a 6-311g+(d) basis set as implemented in a Gaussian 09 package (revision C.01). The net charge on the atoms was assigned through the analysis of natural bonding orbitals. By using highest occupied molecular orbital (HOMO) and least unoccupied molecular orbital (LUMO) energy values for a molecule,[25] electronegativity (χ), chemical hardness (η), and softness (δ) can be calculated using eqs –6, respectivelywhere I = −EHOMO and A = −ELUMO are ionization potential and electron affinity.[26]

Results and Discussion

Structure, Surface Area, and Porosity

FTIR spectra of SE-C, n class="Chemical">Fe@C-1, and Fe@C-2 are Shown in Figure A, in which all have peaks at 1590 and 670 cm–1 associated with C=O stretching vibration and N–H bending vibration, respectively.[27] Also, a very broad peak near 3426 cm–1 and a sharp spike at 1015 cm–1 were shown in these samples, corresponding to stretching and bending vibrations of −OH groups, respectively;[28] however, their lower intensity of FTIR peaks indicates a significant decrease of O-containing groups from the original SE. Moreover, two narrow vibration peaks at 750 and 820 cm–1, representing bending vibration of σ=CH bonding from ortho- and meta-groups in a benzene ring, had disappeared in both Fe@C-1 and Fe@C-2 samples.[29] It means that the graphitization degrees of these Fe@C samples were increased after Fe-catalytic calcination.

Figure 1

(A) FTIR spectra, (B) XRD patterns, (C) Raman spectra, and (D) nitrogen adsorption isotherms of the as-prepared samples (SE-C, Fe@C-1, and Fe@C-2).

(A) FTIR spectra, (B) XRD patterns, (C) Raman spectra, and (D) nitrogen adsorption isotherms of the as-prepared samples (n class="Chemical">SE-C, Fe@C-1, and Fe@C-2).

The PXRD patterns of original SE, calcined n class="Chemical">SE-C, and Fe@C composites are shown in Figures B and S1. Original SE has diffraction peaks at 2θ = 15.2/24.6° and 26.79/29.32° indexed to CaCO3 and SiO2 species,[30] respectively, which might have come from silkworms’ food components.[31] After being calcined at a high temperature, the diffraction peak in all spectra appearing around 25.5° corresponded to the (002) plane of a partially graphitized structure.[32] After doping Fe species, new peaks at 2θ = 36.0/45.1 and 41.3/46.1° in Fe@C-1 and Fe@C-2 represent Fe3C and Fe5C2 structures, respectively,[33] and the peak intensity increases with the increased Fe loading amounts. We believe that these iron carbides are formed by the reduction of Fe ions during the high-temperature carbonization process, which also provide weak polar adsorption sites on the surface of Fe@C composites.[34,35]

The effect of n class="Chemical">Fe loading contents on the graphitization extent of SE-C was tested via Raman analysis, as shown in Figure C, which suggests that all samples exhibited two overlappings of the D peak and the G peak (ID/IG) as listed in Table S2. These values can determine their degree of graphitization.[36] The ID/IG values of Fe@C-2 (1.14) and Fe@C-1 (1.08) are obviously lower than that of the SE-C sample (1.77), confirming a higher degree of graphitization in the Fe@C samples because of the introduction of Fe species during the calcination.[37,38] This enriched graphitic carbon in the newly designed Fe@C composites may also have catalytic properties.[39]

N2 isotherms and the pore size distribution of n class="Chemical">SE-C and Fe@C composites are shown in Figures D and S2, whereas their porosity parameters are listed in Table . The SE-C sample showed a type IV isotherm with a clear hysteresis loop (H4) in its N2 isotherm at 77K, indicating a typical hierarchical structure with micro/mesopores,[40] and it has the highest N2 capacity and surface area (1560.2 m2/g), and a larger hysteresis loop among the three samples. The hierarchical structure of SE-C was composed of micropores and mesopores in its pore structure. After Fe doping, N2 uptake and the surface area, mesopore surface area, and pore volume of Fe@C-1 (1104.1 m2/g) and Fe@C-2 (904.3 m2/g) decreased as seen in Figure D and from Table . These changes could be caused by the shrinkage of pore structure, blockage of pore by Fe species during calcination, and possible density increase compared to the SE-C sample. Moreover, the pore size distribution of all these samples were calculated by employing DFT as shown in Figure S2, suggesting similar pore-size distribution at 5.0–6.0, 6.7–8.1, 12.0–15.0, and 20.0–25.0 Å.

Table 1

Physical Properties of SE-C, Fe@C-1, and Fe@C-2

sample	SBETa (m2/g)	Smica (m2/g)	Smesoa (m2/g)	Vta (cm3/g)	Vmica (cm3/g)
SE-C	1560.2	164.0	1396.2	0.96	0.06
Fe@C-1	1104.1	216.2	887.9	0.75	0.13
Fe@C-2	904.3	280.0	624.3	0.53	0.12
Fe-removed Fe@C-1	1396.1	211.0	1185.1	0.89	0.08
MIL-101	3163.3	56.7	3106.6	1.51	0.01
UiO-66	1292.4	1170.5	121.9	0.61	0.50
C-CA	760.4	668.1	92.2	0.38	0.31

SBET—BET surface area; Smic—microporous surface area; Smeso—mesoporous surface area; Vt—total volume; Vmic—microporous volume.

SBET—BET surface area; Smic—microporous surface area; Smeso—n class="Chemical">mesoporous surface area; Vt—total volume; Vmic—microporous volume.

Morphology Analysis

Scanning electron microscopy (n class="Chemical">SEM) and TEM images of SE, SE-C, and Fe@C composites are shown in Figures and S3. The original SE presents an interesting wrinkled structure with a smooth surface as shown in its SEM images (Figure S3A). Further examination by TEM (Figure A) shows that SE actually possesses very thin nanosheet structures, which are not commonly seen on carbon materials derived from biomass.[41,42] It is attributed to a special reconstruction of lamellar structures from mulberry leaf cellulose by silkworms through intestinal absorption.[43,44] After being calcinated with Fe species, SE was transformed into a 3D nano-spongy network packed with nanosheets (Figures S3B and 2B). This new Fe@C composite possesses many porous channels, providing a great diffusion pathway for molecules. Some nanosheets in the original SE were broken during high-temperature calcination and Fe/Zn activation, which were reconstructed into a spongy network through self-assembly. Figure C shows that Fe species are uniformly dispersed over the Fe@C-1 composites. Furthermore, a high-resolution TEM image shows that the clear lattices’ fringe spacing in Fe@C-1 is around 0.183, 0.201, 0.205 and 0.307 nm, among which 0.183 and 0.205 nm are assigned to the 312 and 510 planes of Fe5C2, 0.201 and 0.307 nm are assigned to the 211 and 210 planes of Fe3C, respectively.[45,46] These Fe–C lattices can be seen embedded into the carbon layer, whereas Fe–C bonds were in accordance with the PXRD results as shown in Figure B. The carbon layer with d = 0.345 nm can be easily assigned to (002) plane of the graphitic carbon,[47] confirming the catalytic role of Fe species in the formation of graphitic carbon in SE.

Figure 2

TEM images of (A) SE, (B) SE-C, (C) Fe@C-1, and (D) inset of Fe@C-1.

TEM images of (A) SE, (B) n class="Chemical">SE-C, (C) Fe@C-1, and (D) inset of Fe@C-1.

Composition Analysis

In order to study the effects of n class="Chemical">Fe species on the surface properties of the SE-derived carbons, XPS elemental analysis was performed on SE-C and Fe@C-1 as shown in Figures , S4, and Table S3, in which the oxygen atomic percentage was decreased from 11.88% in SE-C to 3.96% in Fe@C-1 (Figure S4 and Table S3). Thus, we believe that through Fe catalysis and activation, larger amounts of oxygen in SE were removed during the calcination of Fe@C-1 compared to SE-C, making the surface of Fe@C-1 more hydrophobic. Meanwhile, 0.56% atomic percentage of Fe in the full survey spectrum of the Fe@C-1 sample indicates the successful grafting of iron on the carbon surface.

Figure 3

High-resolution C 1s XPS spectra of (A) SE-C and (B) Fe@C-1.

High-resolution C 1s XPS spectra of (A) SE-C and (B) n class="Chemical">Fe@C-1.

The high-resolution XPS spectra of C 1s from the SE-C and n class="Chemical">Fe@C-1 samples present peaks at binding energies of 284.8, 285.6, and 286.7 eV, which are assigned to sp2 C, sp3 C, and O=C–O bands, respectively, as shown in Figure A,B.[48] Clearly, the intensity of the sp3 C peak in Fe@C-1 was significantly lower than that in SE-C, as shown in Table S4, whereas the intensity of sp2 C was increased to 75.3%. The reduction of sp3 C in Fe@C-1 was due to its transformation into sp2 graphite carbon with Fe catalysis, which in turn can also enhance the hydrophobicity of Fe@C-1.[49,50] Moreover, the presence of Fe in the form of the Fe3C band (284.3 eV) would weaken the polarity of graphitic carbon as well.

Adsorption Kinetics

The results of adsorption kinetics of MA and AY on n class="Chemical">SE-C, Fe@C-1, and Fe@C-2 at 298 K are shown in Figure , in which the adsorption capacity of MA is much higher than that of AY on these samples. MA and AY have similar molecular structures, except for a methyl (−CH3) group in MA and an aldehyde group (−CHO) in AY. The faster adsorption rate by MA also suggests a stronger interaction between the adsorbents and MA than AY, and the order of capacity of MA follows Fe@C-1 > Fe@C-2 > SE-C, which is not consistent with the trend of their specific surface areas (Table ). These results demonstrate that the surface properties and surface areas are at least equally important in the adsorption of MA. Fe@C-1 composites exhibited not only the highest adsorption capacity, but also the highest adsorption density per surface area because of the synergistic effect of iron sites and carbon sites. However, excessive loading of Fe species may block the pores of porous carbon and lower the surface area as shown in the nitrogen isotherm (Figure D and Table ), leading to a slight decrease in adsorption capacity of MA.

Figure 4

Adsorption kinetic of MA and AY (A) on three materials and (B) on Fe@C-1 after Fe removal in a single component, and dependence of −ln(1 – Q/Qe) of MA (C) and AY (D) adsorption on time for the three materials at 298 K.

Adsorption kinetic of MA and AY (A) on three materials and (B) on Fe@C-1 after n class="Chemical">Fe removal in a single component, and dependence of −ln(1 – Q/Qe) of MA (C) and AY (D) adsorption on time for the three materials at 298 K.

To further understand the effects of n class="Chemical">Fe species on the adsorption efficiency of MA and AY on Fe@C composites, an HCl (1.0 mol/L) solution was used to completely remove Fe species on Fe@C-1 (Fe-removed Fe@C-1). The adsorption kinetic curves of MA and AY were measured again as shown in Figure B and porosity parameters of Fe-removed Fe@C-1 are listed in Table , in which the specific surface area increases greatly. However, the adsorption equilibrium capacity of MA on Fe-removed Fe@C-1 does not change much compared to that of Fe@C-1, whereas the kinetics is slower after Fe is removed, though it is still higher than the SE-C sample. Apparently, the Fe species played an important role in facilitating the diffusion and adsorption kinetics of MA on Fe@C composites. On the contrary, the adsorption uptake of AY on the Fe-removed Fe@C-1 samples increased majorly because of the increase of its surface area (SBET from 1273.9 to 1396.1 m2/g) rather than the Fe species. This further suggests that in the absence of Fe species, the surface area iss a major factor to determine the adsorption properties of AY.

Moreover, kinetics data of the three samples show that 60% adsorption uptake for both MA and AY ocn class="Chemical">curs within 30 min and the adsorption equilibrium is generally achieved after about 100 min. These adsorption kinetic profiles can be fitted to the linear driving force (LDF) model, expressed in eq .where qe (mmol/g) and q (mmol/g) are the adsorption capacities at equilibrium and adsorption time t (min), respectively, and k (1/min) is the adsorption rate constant. The fitting curves of the LDF model are shown in Figure C,D, and the fitting parameters are listed in Table . The R2 values of the model is above 0.99, indicating a high degree of agreement between the experimental data and the model. The slightly higher rate constant of MA than that of AY is consistent with the observed kinetic curves as described above. Moreover, Fe@C-2 and Fe@C-1 showed a higher rate constant for MA and a lower rate constant for AY compared to the bare SE-C sample, indicating that the Fe-doped surface with enriched sp2 carbon exhibited higher affinity of MA than that of AY.

Table 2

Adsorption Kinetics Parameters for MA and AY on SE-C, Fe@C-1, and Fe@C-2

	MA			AY
sample	qe1 (mmol/g)	k1 × 102 (min–1)	R12	qe2 (mmol/g)	k2 × 102 (min–1)	R22
SE-C	5.12	5.35	0.998	1.85	4.48	0.998
Fe@C-1	6.63	6.92	0.999	2.08	3.41	0.999
Fe@C-2	5.32	5.67	0.998	1.96	3.96	0.999

Adsorption Equilibrium

In order to compare the adsorption capacity of MA molecules on the new n class="Chemical">adsorbents, adsorption isotherms of MA were measured on SE-C, Fe@C composites, and several typical adsorbents including two MOFs and a commercial activated carbon XFNANO C-AC (Table ).[51] The experimental data were also fitted using Langmuir and Freundlich models given by eqs and 9,[52] whereas the results are shown in Figure A,B and the fitted parameters are listed in Table .where Qm and Qe (mmol/g) are the maximum and equilibrium adsorption capacity and KL (L/mmol) is a constant related to the adsorption energy; KF (mmol(1–1//g·L1/) is defined as the adsorption or distribution coefficient, and n is an indicator of the adsorption intensity.

Figure 5

Adsorption isotherms of MA on (A) SE-C and Fe@C, and (B) Fe@C-1, MIL-101, UiO-66 and commercial AC samples at 298 K and 12 h. Inset: Adsorption isotherms of MA on adsorbent at low concentrations.

Table 3

Langmuir and Freundlich Parameters for MA Adsorption on Various Porous Materials

	Langmuir			Freundlich
sample	KL	Qm (mmol/g)	R12	KF	nF	R22
SE-C	0.39	7.12	0.991	3.02	0.22	0.898
Fe@C-1	0.91	7.56	0.997	3.90	0.19	0.913
Fe@C-2	0.63	7.35	0.992	3.56	0.18	0.892
C-AC	0.78	4.25	0.998	2.19	0.18	0.908
UiO-66	0.53	6.82	0.992	3.66	0.16	0.883
MIL-101(Cr)	0.32	7.49	0.995	2.93	0.24	0.902

Adsorption isotherms of MA on (A) SE-C and n class="Chemical">Fe@C, and (B) Fe@C-1, MIL-101, UiO-66 and commercial AC samples at 298 K and 12 h. Inset: Adsorption isotherms of MA on adsorbent at low concentrations.

The Langmuir model provides a better fitting to the experimental data than the Freundlich model, indicating that adsorption of MA may be preferentially ocn class="Chemical">curring via monolayer formation over the surface of these adsorbents. Among the examined samples, Fe@C-1 has the highest KL value, indicating the highest adsorption affinity toward MA molecules. Fe@C-1 also exhibited much higher adsorption capacity (7.3 mmol/g) than MOFs, in which MIL-101 has mesoporous cages,[53] whereas UiO-66 possesses a microporous structure with hydrophilic metal sites (Zr4+) and a hydrophobic surface.[54] In particular, at the feeding concentration lower than 5.0 mmol/L of MA (the inset of Figure B), Fe@C-1 exhibited about 38% higher adsorption uptake than MIL-101 with SBET = 3163.3 m2/g and UiO-66 with SBET = 1292.4 m2/g (Table ), which strongly indicates that the high capacity of MA on Fe@C composites is dominated by the surface properties rather than by the surface areas of the examined adsorbents. In addition, the adsorption capacity per unit surface area of these samples, as depicted in Figure S5, in which Fe@C-1 has an adsorption capacity of 7.9 × 10–3 mmol/(g·m2), is about 3.6 and 1.8 times higher than that of MIl-101 and UiO-66, respectively.

Selective Separation of Binary MA/AY Mixture

Multicomponent adsorption is important to evaluate the real performance of new adsorbents.[55,56] In partin class="Chemical">cular, the effects of adsorption time, adsorbate concentration, and molar ratio in the multicomponent mixture on the adsorption selectivity are examined in this work, as shown in Figure . The selectivity changes on SE-C and Fe@C-1 show a clear decreasing trend with the adsorption time, whereas the examined MOFs and C-AC adsorbents show much lesser change in their selectivity with the adsorption time. However, these MOFs and C-AC have a much lower selectivity ranging from 4 to 2, indicating again that the surface property is more important than the surface area affecting the selectivity of binary adsorption of MA and AY. Fe@C-1 exhibits the highest selectivity (17.3) for MA/AY at the beginning of adsorption, which is 5.8–2.7 times higher than the examined MOFs and C-AC. This could be mainly attributed to the unique nano-spongy morphology and high porosity (Figure B,C), which effectively accelerated the MA diffusion (Table ) in Fe@C composites. This leads to the preoccupation of adsorption sites by MA rather than AY, resulting in a higher MA selectivity at the beginning of the adsorption process. Then, MA selectivity decreases with the adsorption time, suggesting that more AY molecules are getting adsorbed onto the adsorbent surface so that the selectivity dropped. The selectivity of MA/AY on the Fe-removed Fe@C-1 sample is lower than Fe@C-1 (Figure A), which further suggests the important role that the Fe species has played in the adsorption of MA.

Figure 6

Effects of (A) adsorption time (molar ratio of MA/AY = 1:1 and concentration of MA and AY = 12.0 mmol/L), (B) binary mixture ratio (adsorption time = 1 min and concentration of MA and AY = 12.0 mmol/L), (C) mixture concentration (adsorption time = 1 min and molar ratio of MA/AY = 1:1) on selective of MA over AY in binary system on SE-C, Fe@C, and other porous materials at 298 K.

Effects of (A) adsorption time (molar ratio of MA/AY = 1:1 and concentration of MA and AY = 12.0 mmol/L), (B) binary mixture ratio (adsorption time = 1 min and concentration of MA and AY = 12.0 mmol/L), (C) mixture concentration (adsorption time = 1 min and molar ratio of MA/AY = 1:1) on n class="Chemical">selective of MA over AY in binary system on SE-C, Fe@C, and other porous materials at 298 K.

The effects of n class="Chemical">adsorbate concentration and the molar ratio of MA/AY on selectivity were also studied under the condition of 298 K and 60 s adsorption time, as shown in Figure B,C. The results clearly show that the selectivity on Fe@C-1 increases with the molar ratio between MA and AY, giving selectivity as high as 43.7 when the molar ratio is 9:1 in the mixture, as shown in Figure B. However, the selectivity on Fe@C-1 decreases with the absolute concentration of MA and AY in an equimolar mixture solution, changing from 17.2 at 12 mmol/L to 5.9 at 96 mmol/L of MA and AY, as shown in Figure C. The decrease in selectivity is possibly due to the quick saturation of limited adsorption sites or a concentration-dependent diffusion rate at higher concentrations.

Electronic Structure of Adsorbate Molecules and Adsorption

DFT simulations were performed to calculate the molen class="Chemical">cular properties including dipole moment (Figure S6), charge distribution (Figure ), and frontier molecular orbital (FMO) energies (Figure ). The data obtained from the calculation for MA and AY are listed in Table S5.

Figure 7

Molecular electrostatic potential for (A) MA and (B) AY.

Figure 8

MA population of frontier orbitals: (A) LUMO and (B) HOMO, and AY population of frontier orbitals: (C) LUMO and (D) HOMO.

The electrostatic potential of MA and AY was mapped into the constant electron density surface, as shown in Figure . The electrostatic potential was increan class="Chemical">sed in the color order of red (−0.05 eV) < orange < yellow < green < blue (+0.05 eV), indicating that electron flows from the blue to the red region, and the highest electron density exists in the red region.[57] In the case of MA, electrons were transferred from alkyl and alkoxy groups to the benzene ring (arrows in Figure A), which increased the charge density over the benzene ring (fluorescent yellow). As for AY, electrons were transferred from the alkoxyl group to the aldehyde group, and formed the highest charge density shown by the red color (arrows in Figure B). This variation in electron redistribution results in a lower electron density in the benzene ring of AY than that of MA, which could be a major reason for the different adsorption behaviors in experiments. The Fe@C composites possess a high amount of sp2 C, and its graphitic surface will preferentially form π–π stacking interactions with the benzene ring of MA or AY. However, AY possesses a strong electron-donating (alkoxyl) group and a strong electron-withdrawing (aldehyde) group attached to the benzene ring.[58,59] The decreased electron density on the benzene ring leads to increased polarity of the molecule and thus weakens the affinity of AY with the graphitic surface of the SE-derived carbon composites. These factors result in a decreased adsorption capacity and slower kinetics of AY, thus a high selectivity toward MA in the mixture adsorption.

Moreover, the dipole moments and FMO energies of the two adsorbates were also caln class="Chemical">culated and are shown in Figure . Based on the simulation results, the dipole moments of MA and AY exhibited a great difference, that is, 1.655 and 5.627 D (Table S5), respectively. Based on their FMO treatment, HOMO energies as −6.150/–6.731 eV and LUMO energies as −0.545/–2.022 eV for MA and AY, respectively, were calculated. Through eqs –6, we can see that the electro negativity (χ) and chemical hardness (δ) for MA are 3.348 and 2.803 eV, and 4.377 and 2.355 eV for AY, respectively. AY shows a much smaller energy gap between HOMO and LUMO than MA, indicative of an easy polarization. It was also consistent with their respective dipole moment data, that is, 1.655 and 5.627 for MA and AY. According to the principle of similarity compatibility, compared to AY, MA would prefer to bond with a weak-polar surface of Fe@C-1 (sp2 C with weak-polar Fe–C bond). Meanwhile, the χ and δ of these two molecules exhibited a similar trend. MA possesses a lower χ than AY, and thus shows a mild hardness of the basic property. According to the hard and soft acid and base theory,[24] Fe metal on Fe@C composites belongs to a soft acid, which prefers to form stronger bonds with a soft basic center (MA) rather with a hard basic center (AY). Thus, MA as a weak-polar molecule, would be preferentially adsorbed through the π–π stacking interactions with the sp2 graphitic carbon surface and weak polar Fe species on Fe@C composites.[60] On the contrary, AY was confirmed to possess the polarity in its molecule, and thus would not be easily attracted by weakly polar and graphitic carbon on the surface of Fe–C composites. These factors collectively contribute to the enhanced adsorption capacity and selectivity of MA on the currently designed Fe@C composites compared to other MOFs and porous carbon materials, which shows the potential of SE-derived Fe@C composites for the separation applications of anisole compounds.

Conclusions

New composite adsorbents have been generated from the n class="Chemical">SE, which is a useful biomass waste commonly found in China and South Asia. A calcination process was used to prepare high-graphitic porous nano-spongy carbon materials (SE-C). A metal-doping procedure was also used to modify the surface of SE-C to create Fe@C composites for anisole separation applications. The microscopy imaging and composition analysis show that Fe3C and Fe5C2 species exist on Fe@C composites. Fe activation and catalysis resulted in carbon surface of Fe–C composites with a high degree of graphitization having the ratio ID/IG = 1.08/1.14. In particular, Fe@C composites have demonstrated high adsorption capacity of 7.3 mmol/g 4-methylanisole and fast adsorption kinetics, compared to highly porous MOF MIL-101 and UiO-66. Selectivity of MA/AY on Fe@C-1 reached up to 43.7 and 17.3 at 9:1 and 5:5 of binary MA/AY solution mixtures, respectively, which is 5.8–2.7 times higher than the examined MOFs and a commercial activated carbon. Our DFT simulation of MA and AY molecules provides some explanation for the observed adsorption properties by analyzing the electronic structures and electron distribution. We believe that the outstanding selective adsorption of MA was attributed to intensified π–π interaction from the benzene ring and weak polar alkoxyl from MA with sp2 C and Fe species on the surface of Fe@C composites.