Importance of Internal Porosity for Glucan Adsorption in Mesoporous

Jun 1, 2015 - (23, 24) We performed a control experiment using dense graphitic ..... thus only represents dead weight in the composite MCN-MSN materia...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/Langmuir

Importance of Internal Porosity for Glucan Adsorption in Mesoporous Carbon Materials Po-Wen Chung,† Alexandre Charmot,† Timothy Click,‡,§ Yuchun Lin,† YounJue Bae,† Jhih-Wei Chu,*,‡,§ and Alexander Katz*,† †

Department of Chemical and Biomolecular Engineering, University of California, Berkeley, California 94720, United States Institute of Bioinformatics and Systems Biology and §Department of Biological Science and Technology, National Chiao Tung University, Hsinchu 30068, Taiwan

‡

S Supporting Information *

ABSTRACT: To better understand the adsorption of longchain poly(1 → 4)-β-D-glucans on carbon surfaces as well as interactions responsible for this adsorption, we use a comparative study involving mesoporous carbon−silica composite materials that have been etched to varying degrees and all-atom molecular dynamics simulations. The materials synthesized as part of this etching study consist of an as-synthesized composite material (MCN-MSN), MCN-MSN-0.5 (composite materials consisting of 50% carbon by mass), MCN-MSN-0.3 (composite materials consisting of 70% carbon by mass), and MCN, in which silica etching was conducted using an aqueous ethanolic solution of either NaOH or HF. Data for the adsorption of long-chain glucans to these materials from concentrated aqueous HCl (37 wt %) solution demonstrate a direct relationship between the amount of β-glu adsorption and the magnitude of exposed carbon mesopore surface area, which systematically increases and is also accompanied by an increase in the mesopore size during silica etching. This demonstrates βglu adsorption as occurring on internal carbon mesopores rather than exclusively on the external carbon surface. These experimental data on adsorption were corroborated by molecular dynamics (MD) simulations of β-glu adsorption to a graphene bilayer separated by a distance of 3.2 nm, chosen to correspond to the carbon mesopore diameter of the experimental system. Simulation results using a variety of β-glu solvent systems demonstrate the rapid adsorption of a β-glu strand on the graphitic carbon surface via axial coupling and are consistent with experimentally observed trends in fast adsorption kinetics. Solventmediated effects such as small-scale hydrophobicity and preferential interactions with ions are shown to play important roles in modulating glucan adsorption to carbon surfaces, whereas experimental data on hydrophobically modified silica demonstrate that hydrophobicity in and of itself is insufficient to cause β-glu adsorption from concentrated aqueous HCl solution.

■

solution.23 A surprising aspect of this was the up to 7-fold-larger radius of gyration of the adsorbed glucan chains relative to the 1.6 nm pore radius of the MCN, as measured via gelpermeation chromatography (GPC) of desorbed glucans.23 Another surprising aspect was the rapid nature of the mass transport and adsorption, which occurred on a time scale that was faster than could be measured. These observations begged a central question that we focus on in this manuscript: is the observed long-chain β-glucan adsorption occurring on the internal or the external surface of the MCN? This question was motivated by our previously encountered difficulty of adsorbing glucans into confined pore spaces on silica, even when using the formation of strong covalent bonds as a driving force for glucan chemisorption, as well as the plethora of systems in which polysaccharide adsorption has been observed on external surfaces.23,24 We performed a control experiment using dense graphitic carbon nanopowders, which lack internal mesoporosity, and the lack of observed glucan adsorption in these materials suggested that it

INTRODUCTION There is a growing need for a deeper understanding of the interactions between long-chain poly(1 → 4)-β-D-glucans (βglu) and carbon surfaces, which drive the adsorption of the former onto the latter. These are thought to be crucial in a number of emerging catalytic systems for β-glu depolymerization to glucose via hydrolysis.1−20 Previously, we hypothesized that β-glu strand adsorption into internal pores places mechanical strain on those strands, which can be an important driving force for hydrolysis.21 This adsorption also places the βglu strand within an environment of immediate proximity to the surface, where a high density of weak-acid sites catalyzes glycosidic-bond hydrolysis.21,22 Although examples of β-glu adsorption and surface reaction on carbon catalysts continue to gain momentum, our fundamental understanding of long-chain β-glu adsorption on carbon and specifically whether such adsorption occurs on internal rather than solely external surface sites remains limited. We recently demonstrated the adsorption of long-chain (1 → 4)-β-D-glucans (β-glu) on mesoporous carbon nanoparticle (MCN) materials, in amounts corresponding to up to 30% mass uptake (relative to carbon mass), from a good cellulose solvent consisting of concentrated aqueous HCl (37 wt %) © XXXX American Chemical Society

Received: March 31, 2015 Revised: May 28, 2015

A

DOI: 10.1021/acs.langmuir.5b01115 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir was internal rather than external porosity that was crucial because the external porosity of the nanopowder was actually greater than that of the MCN. However, there could be other reasons that the nanopowders were inefficient adsorbents, and an alternative scenario that cannot be ruled out based on previous data could still involve glucan adsorption primarily on the external MCN surface. Such a scenario would be reminiscent of the adsorption of homopolymers adsorbed on carbon nanotubes as random coils to form brushlike polymer layers on carbon surfaces. To support the observed coverages, given the external surface area of the MCN, roughly 13% of the adsorbed long-chain glucan would be adsorbed to the carbon support, whereas the remaining 87% would be protruding into the so-called brush layer. Related brush layers have previously been observed25,26 and date back to the times of ancient Egypt, where ink was synthesized by adsorbing carbohydrate/ glycoprotein consisting of gum Arabic on the external surface of carbon soot.27 Alternatively, the adsorption of glucan chains could involve some combination of the external surface and/or interior porosity, in which a glucan chain interacts by penetrating the interior porosity while also protruding out into solution, where it does not interact with the carbon surface. Apart from polymer adsorption on porous carbon materials, researchers have reported that ordered mesoporous carbon/composite materials can adsorb small dye molecules from aqueous solution, and the high quantity of dye adsorption is attributed to the internal porous structures.28−30 Direct proof for glucan adsorption on the internal surface of a mesoporous carbon material has remained elusive until now. Here, we investigate the adsorption of glucan on the surface of MCN materials using a combined comparative study of etched mesoporous carbon−silica composite materials as well as an allatom molecular dynamics (MD) simulation. The comparison involves composite materials that have been progressively silica etched in order to expose increasingly larger internal carbon mesopores for adsorption. We start this silica etching on a parent mesoporous carbon−silica composite material (MCNMSN), in which the dense silica template plugs virtually all of the carbon,31 except for the micropores native within the carbon wall, according to the nanocasting synthesis of the MCN.32 This synthesis proceeds via carbonization of organic precursor molecules on the walls of the silica. Etching with an aqueous ethanolic solution of NaOH was used to synthesize composite materials consisting of 50 and 70% carbon by mass. Nearly full etching of the silica template (97% of the material is carbon for MCN) was achieved using aqueous ethanolic HF solution. These materials form a basis of comparison, consisting of varying stages of etching, as illustrated in Scheme 1. They allow us to rigorously examine the effect of the carbon interior surface area on β-glu adsorption, which should scale directly with the carbon internal surface area if such adsorption is truly occurring on the surface of interior pores. Simulations of long-chain β-glu together with flat sheets of carbon surfaces were performed to elucidate the atomic interactions accompanying adsorption. Specifically, we analyze the behavior of a glucan chain consisting of 20 cellobiose repeat units in between a graphene bilayer that is separated by 3.2 nm (corresponding to the MCN internal mesopore diameter). This system is immersed in different solvents, water (known to be a poor cellulose solvent), water/LiCl (with 40.5 wt % LiCl used as a coarse model mimicking the 37 wt % aqueous HCl solution that was used to dissolve cellulose and yield dissolved longchain glucans), and DMF/LiCl (with 4.5 wt % LiCl, which is

Scheme 1. Schematic of Silica Etching Represented by the Direction of the Arrows Starting with an MCN-MSN Composite Materiala

a

The degree of silica etching progress is shown to progress sequentially in (a)−(d). The number in the name of the material, such as either 0.5 or 0.3, represents the silica weight fraction as measured using thermal gravimetric analysis (Figure S1 in Supporting Information).

considered to be a good cellulose solvent) to reveal solvent effects in the glucan interaction with carbon materials. Two starting structures consisting of compact and diagonal glucan chains were explored in simulations in these different solvents in order to investigate the dependence of the observed phenomena on initial configurations.

■

EXPERIMENTAL SECTION

Synthesis of Mesoporous Carbon Nanoparticles (MCN). The synthesis of MCN material uses a MCM-48-type mesoporous silica nanoparticle (MSN) material as the structure-directing template via a modified Stöber method.33 The synthesis is accomplished by mixing cetyltrimethylammonium bromide (CTAB; 1.0 g) and a triblock copolymer (Pluronic F127, EO106PO70EO106; 4.0 g) in 298 mL of H2O/NH3/EtOH solution NH4OH(aq) (2.8 wt % NH4OH in water)/EtOH = 2.5/1 (v/v)). Tetraethyl orthosilicate (TEOS; 3.6 g) is added to the solution at room temperature. After vigorous stirring for 1 min, the reaction mixture is kept under static conditions for 24 h at room temperature in order to achieve silica condensation. The resulting solid MSN product is isolated by centrifugation and washed with copious amounts of water, followed by drying at 70 °C in air. To synthesize the MCN material, the surface of MSN is first converted to an aluminosilicate form. This is performed by first calcining the assynthesized dry MSN product at 550 °C for a period of 2 h in air, which removes the surfactant. The calcined sample is mixed with distilled water to synthesize surface silanol groups, and then it is completely dried at 150 °C in air. The dried sample is slurried in an ethanol solution of anhydrous AlCl3 (Si/Al = 20) for a period of 1 h at room temperature. The ethanol solvent is then completely evaporated via rotary evaporation. The dried sample is calcined again at 550 °C for a period of 2 h in air. Mesoporous carbon nanoparticles, MCNs, are synthesized by using furfuryl alcohol (Aldrich) as the carbon source. One gram of aluminosilicate MCM-48 nanoparticles is impregnated with 0.91 mL of furfuryl alcohol. The resulting impregnated material is placed into a Schlenk reactor and is subjected to three freeze− vacuum−thaw de-gas cycles using liquid N2. After this, the mixture is kept under vacuum (under 1 mbar) at 35 °C for 1 h in order to B

DOI: 10.1021/acs.langmuir.5b01115 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir homogeneously distribute the furfuryl alcohol into the pores. After the Schlenk reactor is opened, it is maintained at a temperature of 100 °C for 6 h in air, during which time the polymerization of furfuryl alcohol occurs. Then the composite material is transferred to a quartz boat and subsequently maintained at 350 °C for 3 h under vacuum (1 mbar) for partial carbonization to occur. Afterward, the composite material is impregnated with 0.58 mL of furfuryl alcohol and then transferred to the Schlenk reactor. The same three freeze−vacuum−thaw de-gas cycles and aforementioned polymerization procedure are repeated. Again, the composite material is transferred to a quartz boat, and further carbonization is accomplished by heating the reactor to 900 °C for 2 h under vacuum (1 mbar).34 The carbon product is collected after dissolving the silica template of the final composite material with HF at room temperature for a period of 1 h (10 wt % HF in EtOH/ H2O solution HF (48%)/EtOH/H2O = 20 mL/40 mL/40 mL), followed by washing with copious amounts of water and ethanol. Synthesis of Partially Etched Mesoporous Carbon Nanoparticles. The partially etched MCN material is synthesized using 0.1 N NaOH ethanolic solution (water/ethanol = 1/1(v/v)) as an etching reagent. The composite MCN materials (200 mg) are added in 0.1 N NaOH ethanolic solution (80 mL) in a PP (polypropylene) bottle, and the mixture is held at 45 °C for 10 and 30 min. Materials are shown in Scheme 1. Sample (a) stands for the as-synthesized composite MCN material; sample (b) stands for the composite MCN material which is partially etched for 10 min; sample (c) stands for the composite MCN material which is partially etched for 30 min; and sample (d) stands for the MCN materials in which the silica template is completely removed using the aforementioned HF solution. Cellulose Hydrolysis. A modified protocol based on the original report by Miller et al. is used for the synthesis of shorter β-glu strands.35 This protocol consists of first dispersing 30 mg of cellulose (Avicel) in 10 mL of concentrated hydrochloric acid (37 wt % aqueous) at room temperature for 1 min. This is followed by the addition of 20 mL of cold concentrated hydrochloric acid (−20 °C) so as to reach a total volume of 30 mL. Complete dissolution is achieved at −20 °C after 15 min; afterward, the solution is warmed to 24 ± 1 °C using a water bath, to achieve further glucan hydrolysis, during a period of 2 h. During this time, shorter-chain glucans are synthesized from poly-β-glucans originally comprising the cellulose. This time frame is chosen on the basis of synthesizing a high yield of oligosaccharides relative to the glucose monomer. During the final 10 min, MCN is added and adsorption is allowed to occur under the conditions described below. The molecular weight distribution of (1 → 4)-β-glucan derived from hydrolyzate streams after 2 h of brief hydrolysis at room temperature is characterized using gel permeation chromatography and is shown in Figure S5. Adsorption of Glucans on MCN. Preweighed MCN material is placed in a suitable container with certain volumes of concentrated acid glucan hydrolyzate, as described above, and the resulting slurry mixture is vortexed for 10 min at 4 °C. Afterward, a Speedisk Column (J. T. Baker 8163−04, silica base) is employed for separation of solid MCN via filtration, and the filtered MCN after adsorption is subsequently washed with 3 mL of water to remove any traces of hydrochloric acid. In order to quantify glucose equivalent content in solution following adsorption, which in turn is used for completing material balances of adsorbed glucose equivalents, all glucans in the filtrate solution are converted to glucose. This is accomplished by allowing the collected filtrate to further hydrolyze for 46 h at room temperature, which by HPLC yields only glucose. Only adsorption data representing significant amounts of adsorption via material balance (consisting of differences of at least 10% between standard solution before and filtrate after adsorption) are used. The solution is subsequently diluted, and the adsorbate concentration in the filtrate is analyzed by HPLC using a refractive index detector (RID), and compared with the concentration in the standard solution. HPLC-RID analysis is performed using a Shimadzu HPLC equipped with a Biorad Aminex HPX-87H column at 323 K. Samples are eluted with a 0.01 N H2SO4 mobile phase at a flow rate of 0.6 mL min−1. Products are identified by comparison of retention times with reference compounds. Quantification of mass concentration is determined by

the integrated peak area of glucose or cellobiose using a six-point calibration curve. Characterization of Mesoporous Carbon Materials. The nitrogen adsorption isotherms are measured at liquid nitrogen temperature (77 K) using a Micromeritics ASAP 2020 volumetric adsorption analyzer. The Brunauer−Emmett−Teller (BET) equation is used to calculate the apparent surface area from adsorption data obtained at P/P0 between 0.05 and 0.2. The total volume of micropores and mesopores is calculated from the amount of nitrogen adsorbed at P/P0 = 0.95, assuming that nitrogen adsorption on the external surface is negligible compared to adsorption in internal pores. The pore size distributions (PSD) are calculated by analyzing the adsorption branch of the nitrogen sorption isotherm using the BarretJoyner-Halenda (BJH) method. Thermal gravimetric analysis (TGA) measurement is performed on a Netzsch 449C Jupiter TGA equipped with a QMS 403 Aëolos quadrapole mass spectrometer (MS). A carrier gas consisting of 20% oxygen and 80% argon is used, and volatiles are sent through a heated (300 °C) fused silica capillary column. Approximately 10 mg of sample is placed in an aluminumoxide crucible and is allowed to equilibrate at RT for 30 min prior to measurement. The temperature program consists of (i) heating to 40 °C and holding for 30 min and (ii) heating at 5 °C/min up to a final temperature of 800 °C. Results are analyzed using Proteus thermal analysis software. Molecular Simulation of Glucan Adsorption on Carbon Surfaces: System Setup. The behaviors of a 40-mer glucose chain in bulk and in between graphene bilayer are investigated via all-atom MD simulation. Two configurations are adopted to start the simulations in the bilayer systems for exploring the dependence of results on initial structures. The first is a compact conformation (Figure 3(a)) that is taken from a snapshot in an explicit water simulation at 300 K and 1 atm. The other is a straight conformation placed diagonally within the bilayer, Figure 3(a). The chain is solvated in either TIP3P water36 or N,N-dimethylacetamide (DMA). In addition to glucan chain in a pure solvent, simulations are also conducted with LiCl ions at 40.5 or 4.5 wt % in water and DMA, respectively, using the program Genion in Gromacs 4.6.5.37,38 The LiCl 40.5 wt % in water was calculated to represent the ionic strength of 37 wt % HCl in water that dissolves cellulose. DMA with 4.5 wt % LiCl is also a good solvent for desolving cellulose. The triclinic box dimensions were 26.1 nm × 8.0 nm × 6.2 nm for simulations in both solvents. To simulate the behaviors of the glucan chain with a characteristic surface of MCN in a confined geometry, the chain is placed in between two graphene sheets separated by 3.2 nm, with 3,820 carbon atoms per sheet. The graphene bilayer is set up using the CHARMM39 software, and the cellulose chain is centered between the bilayer. Additionally, a cellulose chain at an angle to both graphene layers (Figure 3(b)) is also employed as the initial configuration for MD simulations. Similar to the aforementioned cellulose chain, the bilayer/cellulose system is solvated in water or DMA with triclinic dimensions of 26.3 nm × 8.0 nm × 5.7 and 28.9 nm × 10.6 nm × 8.3; LiCl is included for a 40.5 wt % and 4.5 wt % within water and DMA, respectively.

■

RESULTS AND DISCUSSION Nitrogen adsorption/desorption isotherms of all materials in Scheme 1 are measured at 77 K and are shown in Figure 1a. The adsorbed quantity of nitrogen in these isotherms is normalized on a carbon-mass basis, because the silica is known to be dense and thus only represents dead weight in the composite MCN-MSN materials.31 The uniform amount of microporosity observed in all materials (i.e., same y-intercept for all isotherms in Figure 1a) arises from the walls of the carbon.32 Upon etching the silica template, there is a synthesis of progressively more mesoporosity, as represented by larger uptakes at higher relative pressures in the isotherms of Figure 1a. The isotherms representing this onset of greater amounts of mesoporosity can be deconvoluted, by examining the BJH C

DOI: 10.1021/acs.langmuir.5b01115 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

Table 1. Surface Structural Properties and Adsorbed Glucans on a Consecutive Set of Composite MCN-MSN materials, Which Have Been Silica Etched to Varying Degrees, as Represented in Figure 1 surface structural properties

samples MCNMSN MCNMSN0.5 MCNMSN0.3 MCN a

SBET (m2/g of carbon)

ratio of adsorbent over hydroylsate (mg/mL)

adsorbed glucans on MCN (mg Glc equiv/g of material)

1273

20/3

a

1789

20/3

54

2235

20/3

99

2201

20/3

95

Nondetectable trace.

progresses, when starting with the same adsorbate solution and ratio of adsorbate solution to adsorbent (3 mL of adsorbate solution to 20 mg of adsorbent). These data also demonstrate a lack of detectable adsorption on the original MCN-MSN material, which is known to consist of micropores in the range of 5 Å − 8 Å.32 Presumably these micropores are too small to cause appreciable adsorption of β-glu strands from concentrated aqueous HCl solution. The data in Table 1 correlate the onset of mesoporosity in material MCN-MSN-0.5 in Figure 1b with measurable levels of glucan adsorption and demonstrate that glucan adsorption occurs within mesopores of the carbon material. Figure 2 offers a different way of examining experimental βglu adsorption data by correlating uptake to mesoporous Figure 1. (a) BET surface area and (b) BJH pore width distribution of a consecutive series of MCN-MSN composite materials, respectively. The adsorbed quantity in (a) as well as the pore volume in (b) are normalized on a per carbon mass basis.

pore-size distribution data in Figure 1b. These distributions demonstrate that the full mesopore diameter of 3.2 nm grows only gradually during etching, within the series MCN-MSN-0.5, MCN-MSN-0.3, and MCN, which all show significant mesoporosity. The BJH pore-size distribution data in Figure 1b do not exhibit isosbestic points when comparing MCNMSN-0.3 and MCN. This is consistent with an absence of smaller mesopores becoming interconverted into larger ones upon progressive etching; instead, it appears that mesopores of varying sizes are synthesized during etching, with larger ones etched after smaller ones. In order to understand the correlation between textural characteristics (and particularly the importance of internal porosity) and function as glucan adsorbents, crystalline cellulose was dissolved in a good solvent consisting of concentrated aqueous HCl (37%), to synthesize a 1 g/L dilute solution, and adsorption of the dissolved glucans within this solution onto the materials of Scheme 1 was performed at room temperature. This solution consists of β-glu strands having a peak-average molecule weight of 3000 g/mol (see Supporting Information Figure S5) which corresponds to approximately 10 cellobiose repeat units and a radius of gyration in water of 2.9 nm-1.8-fold larger than the MCN pore radius in the fully etched material. Table 1 summarizes the results, and demonstrates larger amounts of glucan uptake as etching

Figure 2. Correlation between BET surface area and adsorbed- glucan coverage of the same consecutive set of MCN-MSN materials, which have been silica etched to varying degrees, as represented in Figure 1. The normalized surface area represents the BET surface area on a per carbon mass basis.

carbon surface area. These data demonstrate that the glucan adsorption capacity on a per weight of carbon basis is directly related (affine function) to the BET surface area−an accurate measure of surface area for a mesoporous material−on a carbon mass basis.40 To support this, we rule out glucan adsorption from concentrated aqueous acid solution on silica by demonstrating the lack of such adsorption on the silica template used, MSN. This was performed by conducting D

DOI: 10.1021/acs.langmuir.5b01115 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

Figure 3. Configuration of a 40-mer glucan chain. (a) Side view of the compact and diagonal initial structures. (b) Final configuration of adsorption between a graphene bilayer in different solvent systems.

Figure 4. Profiles of contact number between glucan and the bilayer during MD simulations in different solvent systems.

material, which was verified by low amounts of water adsorption relative to MSN (see the isotherm in Figure S4 of the Supporting Information), there was no observed glucan adsorption. This supports the notion that hydrophobicity in and of itself is not a guarantee of β-glu adsorption to a surface. Previously, we hypothesized that an additional driving force is CH−π interactions, based on structural data from single-crystal X-ray diffraction of glycoproteins,41 a result consistent with a recent DFT study of cellobiose adsorption on the carbon surface. A main conclusion of the latter study is an entropy increase upon cellobiose adsorption onto carbon,42 a signature of small-scale hydrophobicity effects.43 The magnitude of these

adsorption experiments at an even higher ratio of adsorbate solution to adsorbent (thus favoring higher amounts of β-glu adsorption) of 1.5 mL/2 mg of MSN. Previously, significant glucan adsorption on MCN (between 209 mg of glucose equiv/ g and 282 mg of glucose equiv/g) has been observed on MCN at similar adsorbate-to-adsorbent ratios.23 However, the result was a lack of measurable glucan uptake on MSN. Interestingly, a similar result was also obtained with a hydrophobically modified TMS-MSN silica, which represents MSN in which a fraction of silanols on the surface are capped with an equimolar mixture of trimethylsilyl chloride and hexamethyldisilazane at 60 °C. Even on this ostensibly hydrophobic TMS-MSN E

DOI: 10.1021/acs.langmuir.5b01115 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

The addition of a high concentration of LiCl ions in water did not alter the strong adsorption of glucan to graphene sheets in MD simulations, although it took a slightly longer time for the contact number with surfaces to reach plateau values in Figure 4. In bulk water, interactions of ions with the chain, particularly the Li−O couplings (Figure S3), lead to a lower number of intramolecular hydrogen bonds along the chain (Figure S4). In between the bilayer, ion−glucan couplings have magnitudes similar to those in the bulk simulation (Figure S3). Residues close to the bilayer rapidly attach due to the strong driving force of the axial face of glucose to the carbon surfaces. The chain structure observed in the MD simulation as shown in Figure 3b is thus determined by the specific spatial arrangement of the bilayer. It is thus expected that in a real material the configuration of chain adsorption will depend on the surface geometry of the internal carbon pores, and ions in a pure aqueous solvent system would not play a significant role in affecting the strong tendency of attachment to carbon surfaces. Compared to water as a solvent, DMA is a polar organic with a larger molecular size. For the simulation of glucan in bulk DMA in the absence of ions, the number of intrachain hydrogen bonds is greater than that in bulk water (Figure S4), suggesting weaker interactions of the polysaccharide chain with the DMA solvent compared to those in water. However, the adsorption of glucan to graphene layers in pure DMA is not as strong as that in pure water. Although glucan was retained in between the carbon surfaces in the DMA simulation starting from a compact geometry, the contact number is lower than that in water after 300 ns of simulation time, in Figure 4. In the DMA simulation with the diagonal configuration as the initial structure, glucan eventually left the confined space and the contact number dropped to zero in Figure 4. Adsorption to graphene sheets is thus weaker in DMA. The distinct behaviors of glucan adsorption in water and DMA suggest that in addition to CH−π interactions solvent-mediated forces such as smallscale hydrophobicity can play a significant role in adsorption to carbon materials. Indeed, solvent-mediated effects through their specific structures and interactions with solutes and surfaces, together known as small-scale hydrophobicity, are known to be an essential factor in causing the insolubility of glucan in aqueous solutions.43,44 Adding LiCl to DMA leads to significant changes in solvation and gives rise to a good solvent for dissolving cellulose because of the solvent-mediated preferential interactions of ions with the glucan.45 The altered nature of solvation with LiCl in DMA also resulted in distinct behaviors of adsorption to carbon surfaces in MD simulations as compared to those in pure DMA. Persistent adsorption of glucan onto the graphene bilayer was observed in the DMA/LiCl system, in Figure 4. Although the adsorption is weaker than that in water as judged by the contact number, attachment to carbon surfaces is consistent in simulations starting from both initial structures (compact and diagonal). Glucose units close to the carbon surface also tended to adhere to it in the beginning as in simulations with water, but to a smaller overall extent. Subsequent relaxation (100−400 ns) led to only a slight variation of the total contact number, in Figure 4. Compared to glucan simulation in bulk DMA with LiCl, the bilayer does not affect the high local concentration of ions near glucan, and ions stick closely to glucan while adsorbing to the bilayer in DMA. The numbers of intrachain hydrogen bonds are thus significantly lower than those observed in pure DMA. The adsorption of glucan to carbon surfaces can thus reduce the number of ions ordered by the

effects relative to CH−π interactions requires quantitative characterization of the CH−π interactions between an adsorbed β-glu strand and the carbon surface. The fact that all data in Figure 2 lie on the same line further supports the insignificance of the residual silica template and any difference in carbon functionalization as a result of either ethanolic HF or NaOH etching, from the perspective of impacting β-glu adsorption. Indeed, all materials in Figure 2 exhibit an effective area of 3 nm2 of carbon per glucose fragment, under the conditions of the β-glu adsorption experiment. The correlation shown in Figure 2 between the carbon BET surface area and amount of glucan adsorbed on a carbon mass basis further cements the importance of carbon internal mesoporosity as the cause of the observed β-glu adsorption. In addition, we also followed β-glu adsorption via MALDITOF-MS of materials after adsorption (washed with a small amount of water to displace any hydrolyzate between particles). These data demonstrate glucan chain lengths of up to 20-mer of glucose repeat units on the adsorbent for all three materials showing glucan uptake. To gain molecular-level insight into glucan adsorption on a graphitic carbon surface, all-atom molecular dynamics simulations were performed using a 40-mer (20 cellobiose repeat units) glucan that is initially sandwiched between a graphene bilayer, which consists of two sheets separated by 3.2 nm, the same distance as the MCN pore diameter. Two initial chain configurations shown in Figure 3 were investigated, consisting of either a compact (which is taken from a snapshot in an explicit water simulation at 300 K and 1 atm) or diagonal conformation (which consists of a straight chain placed diagonally within the graphene bilayer). The chain was solvated in either TIP3P water,36 water with LiCl (40.5 wt % so as to have similar ionic strength to 37% HCl), or dimethylacetamide (DMA) with LiCl (4.5% to correspond to a known experimental good-solvent system for crystalline cellulose).24 For comparison, control simulations of glucan in the bulk of the stated solvent systems were also performed. Figure 4 shows the temporal evolution of the contact number that the glucan chain made with the graphene bilayer surface for each snapshot in the MD simulation, the profiles of which also indicate the simulation times of different trajectories. A repeatedly observed behavior in all cases is the very rapid (