- Email: [email protected]
Non-exchanging hydroxyl groups on the surface of cellulose fibrils: The role of interaction with water
Carbohydrate Research 434 (2016) 136e142
Contents lists available at ScienceDirect
Carbohydrate Research journal homepage: www.elsevier.com/locate/carres
Non-exchanging hydroxyl groups on the surface of cellulose fibrils: The role of interaction with water Erik L. Lindh a, b, c, Malin Bergenstråhle-Wohlert b, c, Camilla Terenzi a, b, 1, n Furo a n b, c, *, Istva Lennart Salme a b c
Division of Applied Physical Chemistry, Teknikringen 30, SE-10044 Stockholm, Sweden Wallenberg Wood Science Centre, KTH Royal Institute of Technology, Teknikringen 30, SE-10044 Stockholm, Sweden Innventia AB, Box 5604, SE-114 86 Stockholm, Sweden
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 June 2016 Received in revised form 6 September 2016 Accepted 7 September 2016 Available online 10 September 2016
The interaction of water with cellulose stages many unresolved questions. Here 2H MAS NMR and IR spectra recorded under carefully selected conditions in 1H-2H exchanged, and re-exchanged, cellulose samples are presented. It is shown here, by a quantitative and robust approach, that only two of the three available hydroxyl groups on the surface of cellulose fibrils are exchanging their hydrogen with the surrounding water molecules. This finding is additionally verified and explained by MD simulations which demonstrate that the 1HO(2) and 1HO(6) hydroxyl groups of the constituting glucose units act as hydrogen-bond donors to water, while the 1HO(3) groups behave exclusively as hydrogen-bond acceptors from water and donate hydrogen to their intra-chain neighbors O(5). We conclude that such a behavior makes the latter hydroxyl group unreactive to hydrogen exchange with water. © 2016 Elsevier Ltd. All rights reserved.
Keywords: Cellulose Deuterium exchange Hydroxyl groups 2 H MAS NMR FT-IR Molecular-dynamics simulation
1. Introduction Cellulose is the most voluminous renewable raw material on Earth. It has many traditional and emerging applications, all of which would be promoted by a better molecular-scale understanding of its properties [1]. Yet, such an understanding remains incomplete with significant white spots. One such spot is centered on the interaction of cellulose with water [2] e water that wets and significantly changes the macroscopic properties, such as the tensile strength, of the cellulose fibrils but does not dissolve the cellulose. The basic associative structural unit of a cellulose fiber is the partly crystalline fibril. The fibril diameter, in the order of several nanometers, together with the crystallinity both depend on the biological source and the subsequent processing [3]. Within fibrils of the dominant crystalline form in higher plants, namely cellulose Ib, the chains are ordered in flat sheets with hydrogen bonds
* Corresponding author. Innventia AB, Box 5604, SE-114 86 Stockholm, Sweden. n). E-mail address: [email protected] (L. Salme 1 Current address: Department of Chemical Engineering and Biotechnology, University of Cambridge, Pembroke Street, Cambridge CB2 3RA, UK. http://dx.doi.org/10.1016/j.carres.2016.09.006 0008-6215/© 2016 Elsevier Ltd. All rights reserved.
between neighboring chains. The sheets are in turn stacked on top of each other with no hydrogen bonds between adjacent layers. Hence, both hydrogen bonds and hydrophobic forces play a role in keeping the crystal intact in water [4]. There is a broad agreement concerning some basic features of the interaction of cellulose fibers with water: the water molecules seem to be able to (i) penetrate the fibers and (ii) interact with the surface of the constituting fibrils. Such interaction involves the hydroxyl groups residing on the noncrystalline surface of the fibrils [5]. The questions we investigate here, with the help of infrared (IR) and nuclear magnetic resonance (NMR) spectroscopies, as well as molecular dynamics (MD) simulations, are: (i) which hydroxyl groups interact with water and (ii) in which manner does this interaction occur? The hydrogen bond network within the fibrils has been accessed by neutron diffraction [6] and the relative stability of hydrogen bonds have been assessed [7e9]. Yet, even though various aspects of the water-cellulose interface have been comprehensively investigated by molecular simulations [10e12] the site-specific reactivity of hydroxyl groups on the surface of the fibrils remains, from a quantitative standpoint, incompletely characterized and, thus still poorly understood. The glucose unit within the cellulose chain possesses three hydroxyl groups, indexed as 1HO(2), 1HO(3), and 1HO(6) by their respective carbon position. The accessibility of those groups, to
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
various reactive chemicals, has been investigated in connection to the chemical modification of cellulose [13,14]. It was found that molecules like diethylamine or acetic anhydride preferentially react with the 1HO(2) and 1HO(6) moieties. The insight offered by MD simulations is that the 1HO(3)-O(5) hydrogen bond is rather strong at the surface of cellulose fibril, which may limit its accessibility [12,15]. Water molecules are smaller than, for instance, acetic anhydride and may interact in a different manner with the fibril surface. In addition, adsorbed water affects the structure of the cellulose-surface chains. Hence, the questions stated above are still persistent. One way of investigating these issues is to study, in the presence of heavy water, the site-selectivity of the exchange of the hydrogen nuclei (protons, 1H) in the different 1HO groups of cellulose to the deuterium nuclei (deuterons, 2H). Previously, IR features that appeared after proton-deuteron exchange have been explained by assuming particular exchanging 1HO populations [16,17]. But, even though the 1HO region of static FT-IR spectra have previously been successfully deconvoluted and compared to computed models of cellulose crystals [18,19], detailed decomposition of difference bands alone (or jointly with 2HO bands) has not been performed. Previous NMR results are similarly inconclusive on this point: 13C magic angle spinning (MAS) NMR with 2H recouping could not reveal significant differences between different hydroxyl groups [20]; the single previous attempt by 2H MAS NMR spectroscopy [21] did only consider the behavior of water, and the 2H quadrupoleecho spectrum of dry cellulose lacks chemical selectivity [22]. In the present paper, we provide conclusive evidence that only two of the three available hydroxyl groups are subject to proton exchange in the presence of adsorbed water. Our conclusions are derived from joint interpretation of FT-IR and 2H MAS NMR spectra, recorded in carefully prepared samples, together with state-of-theart MD simulations. 2. Results and discussion 2.1. NMR measurements The cellulose investigated here is nominally “microcrystalline” cellulose (MCC) and was derived from cotton linters. Available Xray or NMR-based methods for estimating the crystalline content in cellulose are not fully conclusive [23]. Yet, cotton-linter cellulose is typically indicated to be highly crystalline with fibril diameters in the order of 7e9 nm [24]. The presented sample-preparation procedure provided sufficiently long time for allowing hydrogendeuterium exchange to reach dynamic equilibrium in a controlled heavy-water atmosphere (approximately 93e95% relative humidity) after which the sample was vacuum dried. After this, the detected 2HO-band reports exclusively about the exchanged hydroxyl-groups. The NMR experiments were then performed on such dry samples, with the 2H signal (see Fig. 1, left inset) arising from exchanged deuterons in the hydroxyl groups. Such 2H MAS NMR spectrum is a collection of spinning side bands (SSBs) whose intensity roughly follows the spectral intensity of the corresponding non-spinning spectrum (data not shown here). Within each SSB, the narrowed peaks appear at their respective isotropic chemical shifts [25]. The chemical shift differences in carbohydrates are small [26] and therefore the individual spectral components within the SSB manifold of the 2H NMR spectrum in Fig. 1 could be resolved only by recording a spectral series with increasing evolution time in an inversion recovery experiment (see SI) [27]. Hence, the component intensities within the SSBs are modulated by their respective sitespecific longitudinal relaxation times (T1). A simultaneous fit of all SSBs permits one to separate the spectral components [27] and
137
Fig. 1. 2H MAS NMR spectra of MCC, first deuterated in heavy-water atmosphere and then vacuum dried. The 2H NMR signal arises from deuterons located in exchanged hydroxyl groups. Main plot: deconvoluted individual contributions to the SSB intensities of the inversion-recovery 2H MAS NMR experiment, assigned to 2HO(2) (black filled circles) and to 2HO(6) (blue filled triangles). Left inset: the measured 2H MAS spectrum plotted between 200 and 200 kHz. Right inset: evolution of one SSB in an inversion-recovery 2H MAS NMR experiment, with x-axis values in kHz (black dots: experimental data, red lines: global spectral fit). All intensities are in arbitrary units. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
to obtain their respective chemical shift and T1 value (see Table 1). The spectral fits illustrated in Fig. 1 were assumed to consist of two distinct hydroxyl groups that participate in the hydrogendeuterium exchange. With this assumption we obtain a plausible and consistent outcome: the band shapes of the deconvoluted spectral contributions are, as expected for quadrupole-broadened spectra, symmetric around the central frequency. In addition, the obtained chemical shifts for the two deuteroxyl groups are consistent with the existing literature data about 1H chemical shifts of glucose oligomers dissolved in a water/acetone solution: the peak at 5.9 ppm is then assigned to the 2HO(2) group and the peak at 5.4 ppm is assigned to the 2HO(6) group [26]. Small discrepancies between liquid 1H and solid 2H chemical-shift values are expected because of two effects: (i) differences in the local molecular environments between the solid and the liquid states [28,29], and (ii) second-order quadrupolar-shift effects for the 2H nuclei [30]. This latter effect is expected to be very similar for all the three hydroxyl groups and, therefore, does not change the relative chemical shift of those. Moreover, 13C CP-MAS longitudinal relaxation measurements of cellulose [31,32] have previously indicated that molecular motions for the C-1HO(2) groups are more restricted and, thus, slower than those for the C-1HO(6) groups. This is also reflected here by the smaller second moment (M2) and by the shorter T1 for the assigned 2 HO(6) population. If we allow for a third component, i.e. the 2 HO(3) sites, to contribute to our 2H MAS NMR spectra, the band shapes for the different sites become unphysically asymmetric (see Table 1 Spectral parameters extracted from the inversion-recovery 2H MAS NMR experiment, the presented values are averages over measurements of three different samples (the corresponding standard deviation is shown within brackets).
Population fraction (%) Chemical shift (ppm) T1 (ms)a M2 (109 s 2)b a b
Longitudinal relaxation-time (T1). Second moment (M2).
Site A: 2HO(2)
Site B: 2HO(6)
60 (3) 5.9 (0.1) 640 (150) 3.9 (0.1)
40 (3) 5.4 (0.2) 50 (10) 3.3 (0.1)
138
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
SI). The intensity of the 2H MAS NMR spectrum also shows that the total amount of exchanged hydroxyl groups in the MCC sample is 23%. In addition, the 13C MAS NMR spectrum (see SI) yields that 42% of all glucose units are not in a crystalline environment [33]. If we assume that there is no bulk amorphous cellulose present and assign all non-crystalline sites to the fibril surface, this gives a lateral fibril size of nine cellulose chains, close to previous assessments [34]. These two estimated values (which indicate that 42% of all the hydroxyl groups are on the surface and that 23% of all hydroxyl groups are exchanged) are roughly consistent with having two out of the three available hydroxyl groups being exchangeable. 2.2. FT-IR measurements FT-IR spectra were recorded in two kinds of separate series of samples. In one arrangement the spectra were recorded while the dry cellulose was exposed to 2H2O vapor (at approximately 45% relative humidity), termed as “exchange experiment”. The FT-IR spectral signature of the resulting hydrogen-deuterium exchange is a decrease of the 1HO band occurring in parallel with the appearance of a 2HO band (Fig. 2a). As is known, the exchange influences only the hydroxyl groups in accessible regions such as the fibril surfaces [35]. In the other arrangement, the cellulose was first exposed to 2H2O vapor for a sufficiently long time and then dried. Thereafter, the FT-IR spectra were recorded over a long time period over which the sample was exposed to 1H2O at very low vapor pressure. This has led to a slow re-exchange of deuterium (residing solely on previously exchanged hydroxyls) by hydrogens and, thereby, this experimental series is termed as “re-exchange”. The spectral signature of re-exchange is a decrease of the 2HO bands and a corresponding increase of the 1HO band (Fig. 2b). The FT-IR spectra shown in Fig. 3 results from the following procedures: (i) customary background correction using spectra recorded in parallel ray paths with and without sample, (ii) baseline correction, and (iii) difference between the spectra at the beginning and at the end of either the exchange or the re-exchange process. In the difference spectra some systematic errors are cancelled and only the peaks that vary are needed to be included in the fitted line shapes (a large number of different peaks otherwise would contribute in the 1HO region between 3600 and 3000 cm 1 [16,19,36]). In the case of the hydrogen-deuterium exchange experiments, we do not present the 2HO band, because of the large contribution from 2H2O molecules; hence, all band shapes in Fig. 3 are
Fig. 3. Series of FT-IR difference bands in the 1HO and 2HO stretch regions (all shown as positive spectra). The 1HO-difference band recorded upon having kept dry cellulose in 2H2O vapor in the exchange experiments (main plot), and 1HO (upper left) and 2HO (upper right) difference bands recorded upon having kept deuterated and dry cellulose under low-pressure 1H2O vapor in the re-exchange experiments.
exclusively characteristic to the behavior of the hydroxyl groups. Care was taken to ascertain stable conditions in order to avoid spectral artifacts; still, significant baseline problems arise also because of resonant Mie scattering [37]. Current methods [37] used for accounting for the influence of resonant Mie scattering are not well suited for a situation where only a few distinct bands of the spectra exhibit change upon some physical process. The spectra obtained by conventional baseline correction (see Fig. 3) were of sufficient quality to permit fitting the 2HO and 1HO difference bands as sums of individual spectral lines. However, the results of the fits turned out to be somewhat dependent upon the particular baseline-correction procedure [38]. In Fig. 4 we illustrate this sensitivity when fitting the 1HO bands in Fig. 3 under the assumption of two contributing components. Fig. 4 indicates that there is no contribution to the IR spectra from the wavenumber region that is believed to be dominated by contributions from the 1HO(3) groups [16,17,19,36]. Here we point out that, inside cellulose Ib crystals [7,19], the vibrations of hydroxyl groups become to certain extent coupled. Hence, peaks arising in the wavenumber regions marked in Fig. 4 include some contributions from other hydroxyl groups, which complicates their assignment. However, at fibril surfaces such coupling among vibrations
Fig. 2. Baseline-corrected FT-IR spectra of the exchange experiment (a) and of the re-exchange experiment (b), where the blue and black lines represents the start and the final spectra of the experiments, respectively. The regions in-between the red triangles were used for deconvolution of the 1HO (3950-3650 cm 1) and 2HO (2100-1800 cm 1) contributions to the difference spectra. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
139
Fig. 4. The influence of the baseline-correction procedure of the FT-IR difference 1HO bands (shown in Fig. 3) on the results of the two-component fits. Inclusive (that is, sufficiently broad to include regions from separate assignments of localized vibration modes) literature wavenumber-regions of the different hydroxyl groups [16,19,36] are shown as colored areas.
must be relatively weak [39]. Moreover, even after having considered coupled modes, 1HO(3) is indicated to dominantly contribute in the wavenumber region marked in Fig. 4. Yet, when the fits were performed using three contributing spectral components, where the third component was forced to be in the 1HO(3) wavenumber range in Fig. 4, the resulting spectral intensity of that particular component remained (irrespective of the sample, exchange mode, and baseline correction method) within a few percent of the total peak integral (see SI). It is well known that the 1HO(3)-O(5) hydrogen bond is rather stable [7e9,15,17] and it has earlier been suggested that the 1HO(3) hydroxyl group exchanges only sparsely with water [17]. Yet, a widespread belief is that all accessible hydroxyl groups on the surface of cellulose are in hydrogen exchange with water [40,41]. The FT-IR and NMR results presented here jointly and unambiguously support the former qualitative conclusion and convey the message that the latter belief does not withstand scrutiny; there is no appreciable hydrogen-deuterium exchange for the 1HO(3) hydroxyl group, while there is clear evidence for exchange for the 1H/2HO(2) and 1H/2HO(6) groups. In contrast to previous qualitative indications [17], the approach adopted here is both more quantitative and more robust. Indeed, a result based on spectral assignments within a single experimental methodology (in the present case, by IR) is much more vulnerable to errors than that obtained by a consistent interpretation of two, from each other independent, experimental methods. Moreover, the underlying molecular mechanisms for the observed behavior are addressed with the MD simulations presented below.
Fig. 5. The results of the MD simulations, showing the average amount of different hydrogen bonds between the hydroxyl groups at a cellulose Ib crystalline surface and the surrounding water molecules.
simulations compared to 10 ns used in this study) and on an outdated force-field (GROMOS87). The force-field used in this study (GLYCAM06) is also considered a better choice for reproducing cellulose Ib structure [45]. Specifically, we obtain that the 1HO(3) hydroxyl group exhibits about as many hydrogen bonds to water molecules as the 1HO(2) or the 1HO(6) groups, but with the essential difference that the 1HO(3) groups solely act as acceptors. The 1HO(3) site is instead found to donate its hydrogens exclusively to O(5), see Fig. 5. Hence, the lack of exchange for 1 HO(3) is not a consequence of the lack of interaction between water molecules adjacent to the fibril surface and that hydroxyl group at the fibril surface. Instead, it is the mode of that interaction that differs between 1HO(3) on one hand (acceptor from water) and 1 HO(2) and 1HO(6) on the other hand (donors to water). Based on the used sample-preparation procedure, our finding reveals that hydrogen exchange at the 1HO(3) groups is blocked over the course of several hundred hours, which puts the rate of this exchange to <10 5 s 1. Compared to such extremely slow exchange kinetics, the time course of the IR spectra suggests that the other two rates (for 1 HO(2) and 1HO(6)) are at least two-three orders of magnitude higher. This latter feature will be investigated in more detail in future works.
2.3. MD simulations 3. Conclusions The interaction behavior of the different cellulose-surface hydroxyl groups when exposed to water was also studied by MD simulations. The fibrils of cellulose Ib are believed to dominantly expose the hydrophilic (1 -1 0) and (1 1 0) surfaces [11,42]. Those two surfaces are almost identical; they have similar dynamics at the celluloseewater interface and their most obvious difference lies within their inter-chain distances [43], possibly leading to minor conformational differences such as preferred conformation of the u torsion angle [44]. However, in order to explain the experimentally observed lack of exchange at one of the hydroxyl groups it has been judged sufficient to simulate the (1 -1 0) surface as a model. The simulation was analyzed in terms of hydrogen bonds (from geometric criteria) occurring between the surface hydroxyls themselves as well as with surrounding water molecules, with results presented in Fig. 5 (see SI for more details). The general direction of these findings is the same as has been indicated long time ago [15]. Yet, those old results were based on less statistics (0.5 ns
From the combined usage of MD simulations, NMR spectroscopy and FT-IR spectroscopy, it could be shown that at the surface of the cellulose fibrils only two out of the three accessible hydroxyl groups are in hydrogen exchange with water. This finding is here explained in terms of different modes of the interaction with water, which decisively differs across different hydroxyl groups (see Fig. 5). More specifically, MD simulations reveal that a required condition for hydrogen exchange to occur is the ability of the hydroxyl group to donate its hydrogen to a water molecule in their connecting hydrogen bond. This finding is clearly connected to the structural stability of cellulose. Having gained insight into this mode of interaction with water will prove to be useful when investigating the state of water in hydrated cellulose by the same spectroscopic methods. Moreover, having identified the hydroxyl populations is a prerequisite to achieve correct spectral assignments, and a pathway through which water dynamics can be correlated with the
140
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
dynamics of surface groups. In addition, the appropriate performance of the MD simulations discussed here confirms their credibility when trying to predict more complex dynamical features in fully hydrated cellulose fibrils. Further studies are needed to clarify more subtle effects leading, for example, to different amounts of exchanged 1HO(2) and 1HO(6) and differences in exchanged and reexchanged 1HO FT-IR band shapes. 4. Experimental section 4.1. Material 4.1.1. FT-IR Cotton linters, with a measured cellulose content of 99.9% and a NMR-measured [23] crystallinity index (CI) of about 70%, provided by Tumba Mill were used. The RH of ~95% was generated by a saturated salt solution of K2SO4 (99.0%, Merck) and 2H2O (99.9 atom% 2H, Sigma-Aldrich), and both chemicals were used as delivered. The industrial-grade N2 gas (purity of approximately 99.95%) was delivered by AGA Gas AB. 4.1.2. NMR Cotton-linters powder (MCC, S5504) was used and had a NMRmeasured [23] CI of about 58%. The saturated salt-solution, used for generating a RH of ~93%, was created from 2H2O (99.9 atom% 2H) and K2NO3 (99.0%). Tetramethylsilane (TMS, 99.95%) was used for chemical-shift referencing and to estimate the deuterium content of the deuterium-exchanged samples, the 2H MAS NMR intensity was related to that in a known sample of poly(methyl methacrylate)-d5 (PMMA, P1529-d5MMA). All chemicals were obtained from Sigma-Aldrich, except for PMMA which was purchased from Polymer Source Inc. 4.2. Experimental procedure 4.2.1. FT-IR FT-IR spectra were recorded on a FTS 6000 spectrometer (Digilab Inc., Randolph, MA, USA) equipped with a DTGS (deuterated triglycine sulphate) detector and a KBr beam splitter. Each spectrum was collected by 128 co-added scans of a spectral resolution of at least 4 cm 1. Thin cellulose papers were made by homogenizing the cotton linters with a homogenizer. The cellulose suspension was then filtered through a Durapore® PVDF-membrane filter (Millipore) with pore size 0.65 mm and dried overnight in lab atmosphere. The thin paper was cut and mounted in a homemade sample-holding box (SB) with IR-transparent windows and connections for N2 gas. The SB was built for three IR paths with Zink Selenide windows (Crystran, ZNSEP25-4): one for background measurements and two for sample measurements. The drying of the samples was conducted in a vacuum oven (set to maintain 50 C) from Binder (model VD53) connected to a vacuum pump (Leybold-Heraeus, Trivac D4A). A pressure of about 0.3 mbar was achieved inside the oven, which resulted in an estimated water-vapor pressure of maximum 7.6 mbar (assuming a surrounding laboratory atmosphere of maximum 80% RH and 25 C). Whenever the SB was handled outside of the vacuum oven, an external N2 tube was connected and a constant flow of N2 gas was always applied. In the experiment, termed “exchange experiment”, the samples were first dried for a couple of days in a vacuum oven and then deuterium-exchanged by a flow of moistened N2 gas, created by bubbling N2 gas through a closed container with 2H2O, by which the outlet could be directed through the SB by manually adjusted tube-valves. To be able to adjust the level of humidity entering the SB, the outlet from the primitive humidifier was mixed with dry N2
gas. The flows were then set to achieve a relative humidity of roughly 45% inside the SB. In the case of the second FT-IR experiment, termed “re-exchange experiment”, the samples were first dried for a couple of days in vacuum oven and subsequently deuterium-exchanged inside a desiccator. Inside the desiccator a saturated salt solution of K2SO4 in 2 H2O was used to maintain a relative humidity of roughly 95%. After the exchange (approximately five days), the samples were re-dried in the vacuum oven and the slow re-exchange, resulting from the low water-vapor pressure, was followed by FT-IR measurements. When the samples were moved between the SB and the desiccator, the transfer was carried out inside a plastic glove box (Air Science, Purair Flex-24) connected to N2 gas to lower the relative humidity down to <3% (1H2O). The spectra in the re-exchange experiment were an average of two subsequent background-corrected spectra, while in the case of the much faster exchange experiment only one spectrum could be recorded at each time point and the background spectra were collected at the end of the whole experiment. The backgroundcorrected spectra were then baseline corrected by subtracting a line generated by fitting a cubic polynomial to clusters of points located at lower and higher wavenumbers compared to the 2HO and 1HO peaks in each spectrum. These regions of wavenumbers (2100-1800 cm 1 and 3950-3650 cm 1) were chosen because of the absence of infrared-active vibrations. The baseline-corrected spectra were then subtracted by a reference spectrum acquired at the beginning of the experiment (exchange experiment: before the start of the 2H2O flow, re-exchange experiment: after approximately 1 h in vacuum oven - where most of the heavy-water molecules were removed). These difference spectra were then deconvoluted by Voigt-distributed line shapes [46,47], and the fitting procedure was carried out by using the trust-regionreflective algorithm in MATLAB™. The regions of each spectrum, used for deconvolution/integration, are shown with red triangles in Fig. 2. All the integrations were done by trapezoidal integration in MATLAB™ and the relative standard deviation (1%) of the calculated intensity differences were estimated from the variation of the total peak areas of measurements close in time at the end of the reexchange experiment. 4.2.2. NMR The inversion recovery 2H MAS NMR spectra were recorded on a Bruker Avance-HD 500 MHz spectrometer operating at a 2H-resonance frequency of 76.1 MHz and equipped with a 4 mm MAS probe. The used delays of the inversion-recovery experiment were logarithmically distributed between 0.2 ms and 6 s (minimum of 14 steps) and recorded with at least 2048 scans and a recycle delay of at least 5 s. The 90 pulse length was calibrated to 3 ms and the MAS spinning rate was 10 kHz. The experimental procedure, regarding chemical-shift calibration and the deconvolution of the spectra, is described elsewhere [27]. As an extra precaution the spinning sidebands at 10 and 10 kHz were excluded from the spectral deconvolution, to make sure that the contribution from any remaining water molecules in the sample was suppressed. To be able to quantitatively compare the width of the component-specific SSB band shapes, the second moment (M2) of the components of the NMR spectra were calculated as the intensity-weighted average of the squared frequencies at each component-specific SSB-maxima. The used equations are described in detail elsewhere [27]. The same hardware was used to record the 13C CP-MAS NMR spectra for determination of the CI of the samples, operating at the 13 C-resonance frequency of 125 MHz. To achieve maximum cross polarization (CP), the radio-frequency fields of the CP-transfer were set to match the Hartman-Hahn condition and the contact time was set to 1 ms. The 90 pulse lengths of 13C and 1H were 3.2 ms and
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
4.7 ms, respectively. During signal acquisition (32 scans) the 1H nuclei were heteronuclear decoupled by the TPPM pulse sequence and a recycle time of 96 s was used. A sample spinning speed of 10 kHz was used. To estimate the CI two Lorentzian-distributed lines were simultaneously fitted to each spectral peak of the C4c (crystalline), C4sa (surface and amorphous) and C235 carbons, the fit were carried out through the trust-region-reflective algorithm in MATLAB™. The CI was calculated by dividing the total area of the C4c by the combined area of C4c and C4sa [23,33], see SI. Due to the low water content of the investigated sample, the spectral resolution in the acquired 13C CP-MAS NMR spectra were relatively low and, therefore, it was enough to estimate the peak area of C4c and C4sa by reducing the number of peaks from eight [33] to four. The spectra used for estimating the number of exchanged hydroxyls were recorded with a Bruker Avance II 300 MHz spectrometer operating at a 2H-resonance frequency of 45.7 MHz and equipped with a 4 mm MAS probe. The 90 pulse length was calibrated to 2.3 ms, the recycle delay were set to 1 s (PMMA) and 4 s (MCC) and both spectra were collected with 2048 co-added scans. Spectra were baseline corrected by using a cubic polynomial in TopSpin™. The PMMA sample of known mass was used as an external reference, assuming total deuteration at the methyl and methylene groups (of the backbone of the polymer chains) and integrating the NMR spectrum between 200 and 200 kHz. This integral was then compared with the corresponding integral of the MCC sample, hence providing the number of 2H in the MCC sample of known initial dry mass (corrected for the extra mass attributed to 1 H-2H exchanged hydroxyl groups). All the 2H MAS NMR spectra were recorded at a temperature of 25 C.
4.2.3. MD simulations Molecular-dynamics simulations were performed with the GROMACS 5.0.6 package [48e53] on a molecular model of an interface between a cellulose Ib crystal [6] (8 4 chains of DP 8) and water. The cellulose chains were bonded to their own periodic images to mimic infinite chains. GLYCAM06 force field [54] was applied for cellulose. The cellulose/water interface was created by extending the simulation box in its y-direction, creating a surface parallel to the crystalline (1 -1 0) plane and subsequently filling the space with 2416 TIP3P water molecules [55], see Fig. 6. Other types of cellulose surfaces are likely to be present in a real system, but the
Fig. 6. The cellulose structure, equilibrated with water, as used in the MD calculations. The cellulose chains are aligned along the z-direction of the figure.
141
surface parallel to the crystalline (1 -1 0) plane is considered one of the most ubiquitous, assuming either a rectangular or hexagonal fibril cross section. The plane is rich in hydroxyl groups and interacts strongly with water. The cellulose/water system was simulated with a 10 ns NPT simulation (with a time step of 2 fs) performed with temperature kept at 300 K with velocity rescaling [56] and an anisotropic pressure scaling (compressibility 1.0 e 6 bar 1 in the two crystal plane directions, 1.0 e 5 bar 1 in the direction normal to the interface) at 1 bar using Berendsen barostat [57]. Constraints were applied to all bonds using the LINCS algorithm. Long range electrostatics was accounted for with Particle Mesh Ewald summation (PME) [58,59] outside a cutoff of 1.0 nm. For the van der Waals interactions, a cutoff of 1.2 nm was used together with dispersion correction. Hydrogen bonds were analyzed with geometric criteria using a cutoff Donor (D)-Acceptor (A) distance of 0.3 nm and a D-AH angle of 30 . Numbers in Fig. 5 are averages over time (the last 5 ns of the simulation) and over protruding surface hydroxyl groups. The values are also normalized to the amount of protruding hydroxyl groups of each kind (64). The errors (see SI) were estimated by separating the simulation in ten time periods, and subsequently calculating the average number of each period and calculating the error from the spread of those average values. Acknowledgements This work was supported by the Wallenberg Wood Science Center. MBW is gratefully acknowledging SSF (ICA10-0086) for financial support while IF acknowledges support from the Swedish Research Council VR 2012-03244. All MD simulations were perr cluster at PDC under project number PDCformed at the Tegne 2015-28. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.carres.2016.09.006. References [1] S.J. Eichhorn, A. Dufresne, M. Aranguren, N.E. Marcovich, J.R. Capadona, S.J. Rowan, C. Weder, W. Thielemans, M. Roman, S. Renneckar, W. Gindl, S. Veigel, J. Keckes, H. Yano, K. Abe, M. Nogi, A.N. Nakagaito, A. Mangalam, J. Simonsen, A.S. Benight, A. Bismarck, L.A. Berglund, T. Peijs, J. Mater. Sci. 45 (2010) 1e33. €m, L. Stigsson, J. Mol. Liq. 156 (2010) 76e81. [2] B. Lindman, G. Karlstro [3] R.J. Moon, A. Martini, J. Nairn, J. Simonsen, J. Youngblood, Chem. Soc. Rev. 40 (2011) 3941e3994. [4] M. Bergenstråhle, J. Wohlert, M.E. Himmel, J.W. Brady, Carbohydr. Res. 345 (2010) 2060e2066. [5] R. Jeffries, J. Appl. Polym. Sci. 8 (1964) 1213e1220. [6] Y. Nishiyama, P. Langan, H. Chanzy, J. Am. Chem. Soc. 124 (2002) 9074e9082. [7] V. Agarwal, G.W. Huber, W.C. Conner, S.M. Auerbach, J. Chem. Phys. 135 (2011) 134506. [8] M. Bergenstråhle, L.A. Berglund, K. Mazeau, J. Phys. Chem. B 111 (2007) 9138e9145. [9] A. Watanabe, S. Morita, Y. Ozaki, Biomacromolecules 7 (2006) 3164e3170. [10] K. Mazeau, Carbohydr. Polym. 117 (2015) 585e591. [11] K. Mazeau, A. Rivet, Biomacromolecules 9 (2008) 1352e1354. [12] J.F. Matthews, C.E. Skopec, P.E. Mason, P. Zuccato, R.W. Torget, J. Sugiyama, M.E. Himmel, J.W. Brady, Carbohydr. Res. 341 (2006) 138e152. [13] C. Verlhac, J. Dedier, H. Chanzy, J. Polym. Sci. Part A Polym. Chem. 28 (1990) 1171e1177. [14] C.K. Lee, E.J. Kim, J.H. Jun, Bull. Korean Chem. Soc. 20 (1999) 1153e1158. [15] A.P. Heiner, L. Kuutti, O. Teleman, Carbohydr. Res. 306 (1998) 205e220. chal, H. Chanzy, J. Mol. Struct. 523 (2000) 183e196. [16] Y. Mare n, Cellulose 13 (2006) 131e145. [17] K. Hofstetter, B. Hinterstoisser, L. Salme [18] C.M. Lee, K. Kafle, Y.B. Park, S.H. Kim, Phys. Chem. Chem. Phys. 16 (2014) 10844e10853. [19] C.M. Lee, J.D. Kubicki, B. Fan, L. Zhong, M.C. Jarvis, S.H. Kim, J. Phys. Chem. B 119 (2015) 15138e15149. [20] C.H. Lemke, R.Y. Dong, C.A. Michal, W.Y. Hamad, Cellulose 19 (2012)
142
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
1619e1629. € bitz, J. Schaller, Carbohydr. Polym. 89 (2012) 640e647. [21] F.H. Larsen, M. Scho [22] D. Radloff, C. Boeffel, H.W. Spiess, Macromolecules 29 (1996) 1528e1534. [23] S. Park, J.O. Baker, M.E. Himmel, P.A. Parilla, D.K. Johnson, Biotechnol. Biofuels 3 (2010) 10. [24] H.P. Fink, D. Hofmann, B. Philipp, Cellulose 2 (1995) 51e70. [25] M.M. Maricq, J.S. Waugh, J. Chem. Phys. 70 (1979) 3300e3316. € m, J. Org. Chem. 68 (2003) 1671e1678. [26] S. Bekiroglu, L. Kenne, C. Sandstro , J. Magn. Reson. 268 (2016) 18e24. [27] E.L. Lindh, P. Stilbs, I. Furo [28] M.P. Bhate, J.C. Woodard, M.A. Mehta, J. Am. Chem. Soc. 131 (2009) 9579e9589. [29] C.R. Morcombe, K.W. Zilm, J. Magn. Reson. 162 (2003) 479e486. [30] A.E. Aliev, S.E. Mann, A.S. Rahman, P.F. McMillan, F. Cor a, D. Iuga, C.E. Hughes, K.D.M. Harris, J. Phys, Chem. A 115 (2011) 12201e12211. [31] A.N. Fernandes, L.H. Thomas, C.M. Altaner, P. Callow, V.T. Forsyth, D.C. Apperley, C.J. Kennedy, M.C. Jarvis, Proc. Natl. Acad. Sci. U. S. A. 108 (2011) 1195e1203. , Biomacromolecules 16 (2015) [32] C. Terenzi, K. Prakobna, L.A. Berglund, I. Furo 1506e1515. [33] P.T. Larsson, K. Wickholm, T. Iversen, Carbohydr. Res. 302 (1997) 19e25. [34] Y. Nishiyama, G.P. Johnson, A.D. French, V.T. Forsyth, P. Langan, Biomacromolecules 9 (2008) 3133e3140. [35] Y. Nishiyama, A. Isogai, T. Okano, M. Müller, H. Chanzy, Macromolecules 32 (1999) 2078e2081. [36] M. Schwanninger, J.C. Rodrigues, H. Pereira, B. Hinterstoisser, Vib. Spectrosc. 36 (2004) 23e40. [37] P. Bassan, A. Kohler, H. Martens, J. Lee, H.J. Byrne, P. Dumas, E. Gazi, M. Brown, N. Clarke, P. Gardner, Analyst 135 (2010) 268e277. [38] K.H. Liland, T. Almoy, B.H. Mevik, Appl. Spectrosc. 64 (2010) 1007e1016. [39] J.J. Cael, K.H. Gardner, J.L. Koenig, J. Blackwell, J. Chem. Phys. 62 (1975) 1145e1153. [40] E.T. Engelund, L.G. Thygesen, S. Svensson, C.A.S. Hill, Wood Sci. Technol. 47 (2013) 141e161. [41] D. Reishofer, S. Spirk, Adv. Polym. Sci. (2015) 1e22.
rez, D. Samain, in: H. Derek (Ed.), Adv. Carbohydr. Chem. Biochem., Ac[42] S. Pe ademic Press, 2010, pp. 25e116. [43] M. Bergenstråhle, J. Wohlert, P.T. Larsson, K. Mazeau, L.A. Berglund, J. Phys. Chem. B 112 (2008) 2590e2595. [44] J.F. Matthews, C.E. Skopec, P.E. Mason, P. Zuccato, R.W. Torget, J. Sugiyama, Carbohydr. Res. 341 (2005) 138e152. [45] J.F. Matthews, G.T. Beckham, M. Bergenstråhle-Wohlert, J.W. Brady, M.E. Himmel, M.F. Crowley, J. Chem. Theory Comput. 8 (2012) 735e748. [46] C. Jung, J. Mol. Recognit. 13 (2000) 325e351. [47] F. Schreier, J. Quant, Spectrosc. Ra. 112 (2011) 1010e1025. [48] S. Pronk, S. P all, R. Schulz, P. Larsson, P. Bjelkmar, R. Apostolov, M.R. Shirts, J.C. Smith, P.M. Kasson, D. van der Spoel, B. Hess, E. Lindahl, Bioinformatics 29 (2013) 845e854. [49] S. P all, M.J. Abraham, C. Kutzner, B. Hess, E. Lindahl, in: S. Markidis, E. Laure (Eds.), Solving Software Challenges for Exascale, Springer International Publishing, 2015, pp. 3e27. [50] B. Hess, C. Kutzner, D. van der Spoel, E. Lindahl, J. Chem. Theory Comput. 4 (2008) 435e447. [51] D. Van der Spoel, E. Lindahl, B. Hess, G. Groenhof, A.E. Mark, H.J.C. Berendsen, J. Comput. Chem. 26 (2005) 1701e1718. [52] E. Lindahl, B. Hess, D. van der Spoel, J. Mol. Model. 7 (2001) 306e317. [53] H.J.C. Berendsen, D. van der Spoel, R. van Drunen, Comp. Phys. Commun. 91 (1995) 43e56. lez-Outeirin ~ o, [54] K.N. Kirschner, A.B. Yongye, S.M. Tschampel, J. Gonza C.R. Daniels, B.L. Foley, R.J. Woods, J. Comput. Chem. 29 (2008) 622e655. [55] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L. Klein, J. Chem, Phys. 79 (1983) 926e935. [56] G. Bussi, D. Donadio, M. Parrinello, J. Chem. Phys. 126 (2007) 1. [57] H.J.C. Berendsen, J.P.M. Postma, W.F. van Gunsteren, A. DiNola, J.R. Haak, J. Chem. Phys. 81 (1984) 3684. [58] T. Darden, D. York, L. Pedersen, J. Chem. Phys. 98 (1993) 10089e10092. [59] U. Essmann, L. Perera, M.L. Berkowitz, T. Darden, H. Lee, L.G. Pedersen, J. Chem. Phys. 103 (1995) 8577e8593.
Contents lists available at ScienceDirect
Carbohydrate Research journal homepage: www.elsevier.com/locate/carres
Non-exchanging hydroxyl groups on the surface of cellulose fibrils: The role of interaction with water Erik L. Lindh a, b, c, Malin Bergenstråhle-Wohlert b, c, Camilla Terenzi a, b, 1, n Furo a n b, c, *, Istva Lennart Salme a b c
Division of Applied Physical Chemistry, Teknikringen 30, SE-10044 Stockholm, Sweden Wallenberg Wood Science Centre, KTH Royal Institute of Technology, Teknikringen 30, SE-10044 Stockholm, Sweden Innventia AB, Box 5604, SE-114 86 Stockholm, Sweden
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 June 2016 Received in revised form 6 September 2016 Accepted 7 September 2016 Available online 10 September 2016
The interaction of water with cellulose stages many unresolved questions. Here 2H MAS NMR and IR spectra recorded under carefully selected conditions in 1H-2H exchanged, and re-exchanged, cellulose samples are presented. It is shown here, by a quantitative and robust approach, that only two of the three available hydroxyl groups on the surface of cellulose fibrils are exchanging their hydrogen with the surrounding water molecules. This finding is additionally verified and explained by MD simulations which demonstrate that the 1HO(2) and 1HO(6) hydroxyl groups of the constituting glucose units act as hydrogen-bond donors to water, while the 1HO(3) groups behave exclusively as hydrogen-bond acceptors from water and donate hydrogen to their intra-chain neighbors O(5). We conclude that such a behavior makes the latter hydroxyl group unreactive to hydrogen exchange with water. © 2016 Elsevier Ltd. All rights reserved.
Keywords: Cellulose Deuterium exchange Hydroxyl groups 2 H MAS NMR FT-IR Molecular-dynamics simulation
1. Introduction Cellulose is the most voluminous renewable raw material on Earth. It has many traditional and emerging applications, all of which would be promoted by a better molecular-scale understanding of its properties [1]. Yet, such an understanding remains incomplete with significant white spots. One such spot is centered on the interaction of cellulose with water [2] e water that wets and significantly changes the macroscopic properties, such as the tensile strength, of the cellulose fibrils but does not dissolve the cellulose. The basic associative structural unit of a cellulose fiber is the partly crystalline fibril. The fibril diameter, in the order of several nanometers, together with the crystallinity both depend on the biological source and the subsequent processing [3]. Within fibrils of the dominant crystalline form in higher plants, namely cellulose Ib, the chains are ordered in flat sheets with hydrogen bonds
* Corresponding author. Innventia AB, Box 5604, SE-114 86 Stockholm, Sweden. n). E-mail address: [email protected] (L. Salme 1 Current address: Department of Chemical Engineering and Biotechnology, University of Cambridge, Pembroke Street, Cambridge CB2 3RA, UK. http://dx.doi.org/10.1016/j.carres.2016.09.006 0008-6215/© 2016 Elsevier Ltd. All rights reserved.
between neighboring chains. The sheets are in turn stacked on top of each other with no hydrogen bonds between adjacent layers. Hence, both hydrogen bonds and hydrophobic forces play a role in keeping the crystal intact in water [4]. There is a broad agreement concerning some basic features of the interaction of cellulose fibers with water: the water molecules seem to be able to (i) penetrate the fibers and (ii) interact with the surface of the constituting fibrils. Such interaction involves the hydroxyl groups residing on the noncrystalline surface of the fibrils [5]. The questions we investigate here, with the help of infrared (IR) and nuclear magnetic resonance (NMR) spectroscopies, as well as molecular dynamics (MD) simulations, are: (i) which hydroxyl groups interact with water and (ii) in which manner does this interaction occur? The hydrogen bond network within the fibrils has been accessed by neutron diffraction [6] and the relative stability of hydrogen bonds have been assessed [7e9]. Yet, even though various aspects of the water-cellulose interface have been comprehensively investigated by molecular simulations [10e12] the site-specific reactivity of hydroxyl groups on the surface of the fibrils remains, from a quantitative standpoint, incompletely characterized and, thus still poorly understood. The glucose unit within the cellulose chain possesses three hydroxyl groups, indexed as 1HO(2), 1HO(3), and 1HO(6) by their respective carbon position. The accessibility of those groups, to
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
various reactive chemicals, has been investigated in connection to the chemical modification of cellulose [13,14]. It was found that molecules like diethylamine or acetic anhydride preferentially react with the 1HO(2) and 1HO(6) moieties. The insight offered by MD simulations is that the 1HO(3)-O(5) hydrogen bond is rather strong at the surface of cellulose fibril, which may limit its accessibility [12,15]. Water molecules are smaller than, for instance, acetic anhydride and may interact in a different manner with the fibril surface. In addition, adsorbed water affects the structure of the cellulose-surface chains. Hence, the questions stated above are still persistent. One way of investigating these issues is to study, in the presence of heavy water, the site-selectivity of the exchange of the hydrogen nuclei (protons, 1H) in the different 1HO groups of cellulose to the deuterium nuclei (deuterons, 2H). Previously, IR features that appeared after proton-deuteron exchange have been explained by assuming particular exchanging 1HO populations [16,17]. But, even though the 1HO region of static FT-IR spectra have previously been successfully deconvoluted and compared to computed models of cellulose crystals [18,19], detailed decomposition of difference bands alone (or jointly with 2HO bands) has not been performed. Previous NMR results are similarly inconclusive on this point: 13C magic angle spinning (MAS) NMR with 2H recouping could not reveal significant differences between different hydroxyl groups [20]; the single previous attempt by 2H MAS NMR spectroscopy [21] did only consider the behavior of water, and the 2H quadrupoleecho spectrum of dry cellulose lacks chemical selectivity [22]. In the present paper, we provide conclusive evidence that only two of the three available hydroxyl groups are subject to proton exchange in the presence of adsorbed water. Our conclusions are derived from joint interpretation of FT-IR and 2H MAS NMR spectra, recorded in carefully prepared samples, together with state-of-theart MD simulations. 2. Results and discussion 2.1. NMR measurements The cellulose investigated here is nominally “microcrystalline” cellulose (MCC) and was derived from cotton linters. Available Xray or NMR-based methods for estimating the crystalline content in cellulose are not fully conclusive [23]. Yet, cotton-linter cellulose is typically indicated to be highly crystalline with fibril diameters in the order of 7e9 nm [24]. The presented sample-preparation procedure provided sufficiently long time for allowing hydrogendeuterium exchange to reach dynamic equilibrium in a controlled heavy-water atmosphere (approximately 93e95% relative humidity) after which the sample was vacuum dried. After this, the detected 2HO-band reports exclusively about the exchanged hydroxyl-groups. The NMR experiments were then performed on such dry samples, with the 2H signal (see Fig. 1, left inset) arising from exchanged deuterons in the hydroxyl groups. Such 2H MAS NMR spectrum is a collection of spinning side bands (SSBs) whose intensity roughly follows the spectral intensity of the corresponding non-spinning spectrum (data not shown here). Within each SSB, the narrowed peaks appear at their respective isotropic chemical shifts [25]. The chemical shift differences in carbohydrates are small [26] and therefore the individual spectral components within the SSB manifold of the 2H NMR spectrum in Fig. 1 could be resolved only by recording a spectral series with increasing evolution time in an inversion recovery experiment (see SI) [27]. Hence, the component intensities within the SSBs are modulated by their respective sitespecific longitudinal relaxation times (T1). A simultaneous fit of all SSBs permits one to separate the spectral components [27] and
137
Fig. 1. 2H MAS NMR spectra of MCC, first deuterated in heavy-water atmosphere and then vacuum dried. The 2H NMR signal arises from deuterons located in exchanged hydroxyl groups. Main plot: deconvoluted individual contributions to the SSB intensities of the inversion-recovery 2H MAS NMR experiment, assigned to 2HO(2) (black filled circles) and to 2HO(6) (blue filled triangles). Left inset: the measured 2H MAS spectrum plotted between 200 and 200 kHz. Right inset: evolution of one SSB in an inversion-recovery 2H MAS NMR experiment, with x-axis values in kHz (black dots: experimental data, red lines: global spectral fit). All intensities are in arbitrary units. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
to obtain their respective chemical shift and T1 value (see Table 1). The spectral fits illustrated in Fig. 1 were assumed to consist of two distinct hydroxyl groups that participate in the hydrogendeuterium exchange. With this assumption we obtain a plausible and consistent outcome: the band shapes of the deconvoluted spectral contributions are, as expected for quadrupole-broadened spectra, symmetric around the central frequency. In addition, the obtained chemical shifts for the two deuteroxyl groups are consistent with the existing literature data about 1H chemical shifts of glucose oligomers dissolved in a water/acetone solution: the peak at 5.9 ppm is then assigned to the 2HO(2) group and the peak at 5.4 ppm is assigned to the 2HO(6) group [26]. Small discrepancies between liquid 1H and solid 2H chemical-shift values are expected because of two effects: (i) differences in the local molecular environments between the solid and the liquid states [28,29], and (ii) second-order quadrupolar-shift effects for the 2H nuclei [30]. This latter effect is expected to be very similar for all the three hydroxyl groups and, therefore, does not change the relative chemical shift of those. Moreover, 13C CP-MAS longitudinal relaxation measurements of cellulose [31,32] have previously indicated that molecular motions for the C-1HO(2) groups are more restricted and, thus, slower than those for the C-1HO(6) groups. This is also reflected here by the smaller second moment (M2) and by the shorter T1 for the assigned 2 HO(6) population. If we allow for a third component, i.e. the 2 HO(3) sites, to contribute to our 2H MAS NMR spectra, the band shapes for the different sites become unphysically asymmetric (see Table 1 Spectral parameters extracted from the inversion-recovery 2H MAS NMR experiment, the presented values are averages over measurements of three different samples (the corresponding standard deviation is shown within brackets).
Population fraction (%) Chemical shift (ppm) T1 (ms)a M2 (109 s 2)b a b
Longitudinal relaxation-time (T1). Second moment (M2).
Site A: 2HO(2)
Site B: 2HO(6)
60 (3) 5.9 (0.1) 640 (150) 3.9 (0.1)
40 (3) 5.4 (0.2) 50 (10) 3.3 (0.1)
138
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
SI). The intensity of the 2H MAS NMR spectrum also shows that the total amount of exchanged hydroxyl groups in the MCC sample is 23%. In addition, the 13C MAS NMR spectrum (see SI) yields that 42% of all glucose units are not in a crystalline environment [33]. If we assume that there is no bulk amorphous cellulose present and assign all non-crystalline sites to the fibril surface, this gives a lateral fibril size of nine cellulose chains, close to previous assessments [34]. These two estimated values (which indicate that 42% of all the hydroxyl groups are on the surface and that 23% of all hydroxyl groups are exchanged) are roughly consistent with having two out of the three available hydroxyl groups being exchangeable. 2.2. FT-IR measurements FT-IR spectra were recorded in two kinds of separate series of samples. In one arrangement the spectra were recorded while the dry cellulose was exposed to 2H2O vapor (at approximately 45% relative humidity), termed as “exchange experiment”. The FT-IR spectral signature of the resulting hydrogen-deuterium exchange is a decrease of the 1HO band occurring in parallel with the appearance of a 2HO band (Fig. 2a). As is known, the exchange influences only the hydroxyl groups in accessible regions such as the fibril surfaces [35]. In the other arrangement, the cellulose was first exposed to 2H2O vapor for a sufficiently long time and then dried. Thereafter, the FT-IR spectra were recorded over a long time period over which the sample was exposed to 1H2O at very low vapor pressure. This has led to a slow re-exchange of deuterium (residing solely on previously exchanged hydroxyls) by hydrogens and, thereby, this experimental series is termed as “re-exchange”. The spectral signature of re-exchange is a decrease of the 2HO bands and a corresponding increase of the 1HO band (Fig. 2b). The FT-IR spectra shown in Fig. 3 results from the following procedures: (i) customary background correction using spectra recorded in parallel ray paths with and without sample, (ii) baseline correction, and (iii) difference between the spectra at the beginning and at the end of either the exchange or the re-exchange process. In the difference spectra some systematic errors are cancelled and only the peaks that vary are needed to be included in the fitted line shapes (a large number of different peaks otherwise would contribute in the 1HO region between 3600 and 3000 cm 1 [16,19,36]). In the case of the hydrogen-deuterium exchange experiments, we do not present the 2HO band, because of the large contribution from 2H2O molecules; hence, all band shapes in Fig. 3 are
Fig. 3. Series of FT-IR difference bands in the 1HO and 2HO stretch regions (all shown as positive spectra). The 1HO-difference band recorded upon having kept dry cellulose in 2H2O vapor in the exchange experiments (main plot), and 1HO (upper left) and 2HO (upper right) difference bands recorded upon having kept deuterated and dry cellulose under low-pressure 1H2O vapor in the re-exchange experiments.
exclusively characteristic to the behavior of the hydroxyl groups. Care was taken to ascertain stable conditions in order to avoid spectral artifacts; still, significant baseline problems arise also because of resonant Mie scattering [37]. Current methods [37] used for accounting for the influence of resonant Mie scattering are not well suited for a situation where only a few distinct bands of the spectra exhibit change upon some physical process. The spectra obtained by conventional baseline correction (see Fig. 3) were of sufficient quality to permit fitting the 2HO and 1HO difference bands as sums of individual spectral lines. However, the results of the fits turned out to be somewhat dependent upon the particular baseline-correction procedure [38]. In Fig. 4 we illustrate this sensitivity when fitting the 1HO bands in Fig. 3 under the assumption of two contributing components. Fig. 4 indicates that there is no contribution to the IR spectra from the wavenumber region that is believed to be dominated by contributions from the 1HO(3) groups [16,17,19,36]. Here we point out that, inside cellulose Ib crystals [7,19], the vibrations of hydroxyl groups become to certain extent coupled. Hence, peaks arising in the wavenumber regions marked in Fig. 4 include some contributions from other hydroxyl groups, which complicates their assignment. However, at fibril surfaces such coupling among vibrations
Fig. 2. Baseline-corrected FT-IR spectra of the exchange experiment (a) and of the re-exchange experiment (b), where the blue and black lines represents the start and the final spectra of the experiments, respectively. The regions in-between the red triangles were used for deconvolution of the 1HO (3950-3650 cm 1) and 2HO (2100-1800 cm 1) contributions to the difference spectra. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
139
Fig. 4. The influence of the baseline-correction procedure of the FT-IR difference 1HO bands (shown in Fig. 3) on the results of the two-component fits. Inclusive (that is, sufficiently broad to include regions from separate assignments of localized vibration modes) literature wavenumber-regions of the different hydroxyl groups [16,19,36] are shown as colored areas.
must be relatively weak [39]. Moreover, even after having considered coupled modes, 1HO(3) is indicated to dominantly contribute in the wavenumber region marked in Fig. 4. Yet, when the fits were performed using three contributing spectral components, where the third component was forced to be in the 1HO(3) wavenumber range in Fig. 4, the resulting spectral intensity of that particular component remained (irrespective of the sample, exchange mode, and baseline correction method) within a few percent of the total peak integral (see SI). It is well known that the 1HO(3)-O(5) hydrogen bond is rather stable [7e9,15,17] and it has earlier been suggested that the 1HO(3) hydroxyl group exchanges only sparsely with water [17]. Yet, a widespread belief is that all accessible hydroxyl groups on the surface of cellulose are in hydrogen exchange with water [40,41]. The FT-IR and NMR results presented here jointly and unambiguously support the former qualitative conclusion and convey the message that the latter belief does not withstand scrutiny; there is no appreciable hydrogen-deuterium exchange for the 1HO(3) hydroxyl group, while there is clear evidence for exchange for the 1H/2HO(2) and 1H/2HO(6) groups. In contrast to previous qualitative indications [17], the approach adopted here is both more quantitative and more robust. Indeed, a result based on spectral assignments within a single experimental methodology (in the present case, by IR) is much more vulnerable to errors than that obtained by a consistent interpretation of two, from each other independent, experimental methods. Moreover, the underlying molecular mechanisms for the observed behavior are addressed with the MD simulations presented below.
Fig. 5. The results of the MD simulations, showing the average amount of different hydrogen bonds between the hydroxyl groups at a cellulose Ib crystalline surface and the surrounding water molecules.
simulations compared to 10 ns used in this study) and on an outdated force-field (GROMOS87). The force-field used in this study (GLYCAM06) is also considered a better choice for reproducing cellulose Ib structure [45]. Specifically, we obtain that the 1HO(3) hydroxyl group exhibits about as many hydrogen bonds to water molecules as the 1HO(2) or the 1HO(6) groups, but with the essential difference that the 1HO(3) groups solely act as acceptors. The 1HO(3) site is instead found to donate its hydrogens exclusively to O(5), see Fig. 5. Hence, the lack of exchange for 1 HO(3) is not a consequence of the lack of interaction between water molecules adjacent to the fibril surface and that hydroxyl group at the fibril surface. Instead, it is the mode of that interaction that differs between 1HO(3) on one hand (acceptor from water) and 1 HO(2) and 1HO(6) on the other hand (donors to water). Based on the used sample-preparation procedure, our finding reveals that hydrogen exchange at the 1HO(3) groups is blocked over the course of several hundred hours, which puts the rate of this exchange to <10 5 s 1. Compared to such extremely slow exchange kinetics, the time course of the IR spectra suggests that the other two rates (for 1 HO(2) and 1HO(6)) are at least two-three orders of magnitude higher. This latter feature will be investigated in more detail in future works.
2.3. MD simulations 3. Conclusions The interaction behavior of the different cellulose-surface hydroxyl groups when exposed to water was also studied by MD simulations. The fibrils of cellulose Ib are believed to dominantly expose the hydrophilic (1 -1 0) and (1 1 0) surfaces [11,42]. Those two surfaces are almost identical; they have similar dynamics at the celluloseewater interface and their most obvious difference lies within their inter-chain distances [43], possibly leading to minor conformational differences such as preferred conformation of the u torsion angle [44]. However, in order to explain the experimentally observed lack of exchange at one of the hydroxyl groups it has been judged sufficient to simulate the (1 -1 0) surface as a model. The simulation was analyzed in terms of hydrogen bonds (from geometric criteria) occurring between the surface hydroxyls themselves as well as with surrounding water molecules, with results presented in Fig. 5 (see SI for more details). The general direction of these findings is the same as has been indicated long time ago [15]. Yet, those old results were based on less statistics (0.5 ns
From the combined usage of MD simulations, NMR spectroscopy and FT-IR spectroscopy, it could be shown that at the surface of the cellulose fibrils only two out of the three accessible hydroxyl groups are in hydrogen exchange with water. This finding is here explained in terms of different modes of the interaction with water, which decisively differs across different hydroxyl groups (see Fig. 5). More specifically, MD simulations reveal that a required condition for hydrogen exchange to occur is the ability of the hydroxyl group to donate its hydrogen to a water molecule in their connecting hydrogen bond. This finding is clearly connected to the structural stability of cellulose. Having gained insight into this mode of interaction with water will prove to be useful when investigating the state of water in hydrated cellulose by the same spectroscopic methods. Moreover, having identified the hydroxyl populations is a prerequisite to achieve correct spectral assignments, and a pathway through which water dynamics can be correlated with the
140
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
dynamics of surface groups. In addition, the appropriate performance of the MD simulations discussed here confirms their credibility when trying to predict more complex dynamical features in fully hydrated cellulose fibrils. Further studies are needed to clarify more subtle effects leading, for example, to different amounts of exchanged 1HO(2) and 1HO(6) and differences in exchanged and reexchanged 1HO FT-IR band shapes. 4. Experimental section 4.1. Material 4.1.1. FT-IR Cotton linters, with a measured cellulose content of 99.9% and a NMR-measured [23] crystallinity index (CI) of about 70%, provided by Tumba Mill were used. The RH of ~95% was generated by a saturated salt solution of K2SO4 (99.0%, Merck) and 2H2O (99.9 atom% 2H, Sigma-Aldrich), and both chemicals were used as delivered. The industrial-grade N2 gas (purity of approximately 99.95%) was delivered by AGA Gas AB. 4.1.2. NMR Cotton-linters powder (MCC, S5504) was used and had a NMRmeasured [23] CI of about 58%. The saturated salt-solution, used for generating a RH of ~93%, was created from 2H2O (99.9 atom% 2H) and K2NO3 (99.0%). Tetramethylsilane (TMS, 99.95%) was used for chemical-shift referencing and to estimate the deuterium content of the deuterium-exchanged samples, the 2H MAS NMR intensity was related to that in a known sample of poly(methyl methacrylate)-d5 (PMMA, P1529-d5MMA). All chemicals were obtained from Sigma-Aldrich, except for PMMA which was purchased from Polymer Source Inc. 4.2. Experimental procedure 4.2.1. FT-IR FT-IR spectra were recorded on a FTS 6000 spectrometer (Digilab Inc., Randolph, MA, USA) equipped with a DTGS (deuterated triglycine sulphate) detector and a KBr beam splitter. Each spectrum was collected by 128 co-added scans of a spectral resolution of at least 4 cm 1. Thin cellulose papers were made by homogenizing the cotton linters with a homogenizer. The cellulose suspension was then filtered through a Durapore® PVDF-membrane filter (Millipore) with pore size 0.65 mm and dried overnight in lab atmosphere. The thin paper was cut and mounted in a homemade sample-holding box (SB) with IR-transparent windows and connections for N2 gas. The SB was built for three IR paths with Zink Selenide windows (Crystran, ZNSEP25-4): one for background measurements and two for sample measurements. The drying of the samples was conducted in a vacuum oven (set to maintain 50 C) from Binder (model VD53) connected to a vacuum pump (Leybold-Heraeus, Trivac D4A). A pressure of about 0.3 mbar was achieved inside the oven, which resulted in an estimated water-vapor pressure of maximum 7.6 mbar (assuming a surrounding laboratory atmosphere of maximum 80% RH and 25 C). Whenever the SB was handled outside of the vacuum oven, an external N2 tube was connected and a constant flow of N2 gas was always applied. In the experiment, termed “exchange experiment”, the samples were first dried for a couple of days in a vacuum oven and then deuterium-exchanged by a flow of moistened N2 gas, created by bubbling N2 gas through a closed container with 2H2O, by which the outlet could be directed through the SB by manually adjusted tube-valves. To be able to adjust the level of humidity entering the SB, the outlet from the primitive humidifier was mixed with dry N2
gas. The flows were then set to achieve a relative humidity of roughly 45% inside the SB. In the case of the second FT-IR experiment, termed “re-exchange experiment”, the samples were first dried for a couple of days in vacuum oven and subsequently deuterium-exchanged inside a desiccator. Inside the desiccator a saturated salt solution of K2SO4 in 2 H2O was used to maintain a relative humidity of roughly 95%. After the exchange (approximately five days), the samples were re-dried in the vacuum oven and the slow re-exchange, resulting from the low water-vapor pressure, was followed by FT-IR measurements. When the samples were moved between the SB and the desiccator, the transfer was carried out inside a plastic glove box (Air Science, Purair Flex-24) connected to N2 gas to lower the relative humidity down to <3% (1H2O). The spectra in the re-exchange experiment were an average of two subsequent background-corrected spectra, while in the case of the much faster exchange experiment only one spectrum could be recorded at each time point and the background spectra were collected at the end of the whole experiment. The backgroundcorrected spectra were then baseline corrected by subtracting a line generated by fitting a cubic polynomial to clusters of points located at lower and higher wavenumbers compared to the 2HO and 1HO peaks in each spectrum. These regions of wavenumbers (2100-1800 cm 1 and 3950-3650 cm 1) were chosen because of the absence of infrared-active vibrations. The baseline-corrected spectra were then subtracted by a reference spectrum acquired at the beginning of the experiment (exchange experiment: before the start of the 2H2O flow, re-exchange experiment: after approximately 1 h in vacuum oven - where most of the heavy-water molecules were removed). These difference spectra were then deconvoluted by Voigt-distributed line shapes [46,47], and the fitting procedure was carried out by using the trust-regionreflective algorithm in MATLAB™. The regions of each spectrum, used for deconvolution/integration, are shown with red triangles in Fig. 2. All the integrations were done by trapezoidal integration in MATLAB™ and the relative standard deviation (1%) of the calculated intensity differences were estimated from the variation of the total peak areas of measurements close in time at the end of the reexchange experiment. 4.2.2. NMR The inversion recovery 2H MAS NMR spectra were recorded on a Bruker Avance-HD 500 MHz spectrometer operating at a 2H-resonance frequency of 76.1 MHz and equipped with a 4 mm MAS probe. The used delays of the inversion-recovery experiment were logarithmically distributed between 0.2 ms and 6 s (minimum of 14 steps) and recorded with at least 2048 scans and a recycle delay of at least 5 s. The 90 pulse length was calibrated to 3 ms and the MAS spinning rate was 10 kHz. The experimental procedure, regarding chemical-shift calibration and the deconvolution of the spectra, is described elsewhere [27]. As an extra precaution the spinning sidebands at 10 and 10 kHz were excluded from the spectral deconvolution, to make sure that the contribution from any remaining water molecules in the sample was suppressed. To be able to quantitatively compare the width of the component-specific SSB band shapes, the second moment (M2) of the components of the NMR spectra were calculated as the intensity-weighted average of the squared frequencies at each component-specific SSB-maxima. The used equations are described in detail elsewhere [27]. The same hardware was used to record the 13C CP-MAS NMR spectra for determination of the CI of the samples, operating at the 13 C-resonance frequency of 125 MHz. To achieve maximum cross polarization (CP), the radio-frequency fields of the CP-transfer were set to match the Hartman-Hahn condition and the contact time was set to 1 ms. The 90 pulse lengths of 13C and 1H were 3.2 ms and
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
4.7 ms, respectively. During signal acquisition (32 scans) the 1H nuclei were heteronuclear decoupled by the TPPM pulse sequence and a recycle time of 96 s was used. A sample spinning speed of 10 kHz was used. To estimate the CI two Lorentzian-distributed lines were simultaneously fitted to each spectral peak of the C4c (crystalline), C4sa (surface and amorphous) and C235 carbons, the fit were carried out through the trust-region-reflective algorithm in MATLAB™. The CI was calculated by dividing the total area of the C4c by the combined area of C4c and C4sa [23,33], see SI. Due to the low water content of the investigated sample, the spectral resolution in the acquired 13C CP-MAS NMR spectra were relatively low and, therefore, it was enough to estimate the peak area of C4c and C4sa by reducing the number of peaks from eight [33] to four. The spectra used for estimating the number of exchanged hydroxyls were recorded with a Bruker Avance II 300 MHz spectrometer operating at a 2H-resonance frequency of 45.7 MHz and equipped with a 4 mm MAS probe. The 90 pulse length was calibrated to 2.3 ms, the recycle delay were set to 1 s (PMMA) and 4 s (MCC) and both spectra were collected with 2048 co-added scans. Spectra were baseline corrected by using a cubic polynomial in TopSpin™. The PMMA sample of known mass was used as an external reference, assuming total deuteration at the methyl and methylene groups (of the backbone of the polymer chains) and integrating the NMR spectrum between 200 and 200 kHz. This integral was then compared with the corresponding integral of the MCC sample, hence providing the number of 2H in the MCC sample of known initial dry mass (corrected for the extra mass attributed to 1 H-2H exchanged hydroxyl groups). All the 2H MAS NMR spectra were recorded at a temperature of 25 C.
4.2.3. MD simulations Molecular-dynamics simulations were performed with the GROMACS 5.0.6 package [48e53] on a molecular model of an interface between a cellulose Ib crystal [6] (8 4 chains of DP 8) and water. The cellulose chains were bonded to their own periodic images to mimic infinite chains. GLYCAM06 force field [54] was applied for cellulose. The cellulose/water interface was created by extending the simulation box in its y-direction, creating a surface parallel to the crystalline (1 -1 0) plane and subsequently filling the space with 2416 TIP3P water molecules [55], see Fig. 6. Other types of cellulose surfaces are likely to be present in a real system, but the
Fig. 6. The cellulose structure, equilibrated with water, as used in the MD calculations. The cellulose chains are aligned along the z-direction of the figure.
141
surface parallel to the crystalline (1 -1 0) plane is considered one of the most ubiquitous, assuming either a rectangular or hexagonal fibril cross section. The plane is rich in hydroxyl groups and interacts strongly with water. The cellulose/water system was simulated with a 10 ns NPT simulation (with a time step of 2 fs) performed with temperature kept at 300 K with velocity rescaling [56] and an anisotropic pressure scaling (compressibility 1.0 e 6 bar 1 in the two crystal plane directions, 1.0 e 5 bar 1 in the direction normal to the interface) at 1 bar using Berendsen barostat [57]. Constraints were applied to all bonds using the LINCS algorithm. Long range electrostatics was accounted for with Particle Mesh Ewald summation (PME) [58,59] outside a cutoff of 1.0 nm. For the van der Waals interactions, a cutoff of 1.2 nm was used together with dispersion correction. Hydrogen bonds were analyzed with geometric criteria using a cutoff Donor (D)-Acceptor (A) distance of 0.3 nm and a D-AH angle of 30 . Numbers in Fig. 5 are averages over time (the last 5 ns of the simulation) and over protruding surface hydroxyl groups. The values are also normalized to the amount of protruding hydroxyl groups of each kind (64). The errors (see SI) were estimated by separating the simulation in ten time periods, and subsequently calculating the average number of each period and calculating the error from the spread of those average values. Acknowledgements This work was supported by the Wallenberg Wood Science Center. MBW is gratefully acknowledging SSF (ICA10-0086) for financial support while IF acknowledges support from the Swedish Research Council VR 2012-03244. All MD simulations were perr cluster at PDC under project number PDCformed at the Tegne 2015-28. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.carres.2016.09.006. References [1] S.J. Eichhorn, A. Dufresne, M. Aranguren, N.E. Marcovich, J.R. Capadona, S.J. Rowan, C. Weder, W. Thielemans, M. Roman, S. Renneckar, W. Gindl, S. Veigel, J. Keckes, H. Yano, K. Abe, M. Nogi, A.N. Nakagaito, A. Mangalam, J. Simonsen, A.S. Benight, A. Bismarck, L.A. Berglund, T. Peijs, J. Mater. Sci. 45 (2010) 1e33. €m, L. Stigsson, J. Mol. Liq. 156 (2010) 76e81. [2] B. Lindman, G. Karlstro [3] R.J. Moon, A. Martini, J. Nairn, J. Simonsen, J. Youngblood, Chem. Soc. Rev. 40 (2011) 3941e3994. [4] M. Bergenstråhle, J. Wohlert, M.E. Himmel, J.W. Brady, Carbohydr. Res. 345 (2010) 2060e2066. [5] R. Jeffries, J. Appl. Polym. Sci. 8 (1964) 1213e1220. [6] Y. Nishiyama, P. Langan, H. Chanzy, J. Am. Chem. Soc. 124 (2002) 9074e9082. [7] V. Agarwal, G.W. Huber, W.C. Conner, S.M. Auerbach, J. Chem. Phys. 135 (2011) 134506. [8] M. Bergenstråhle, L.A. Berglund, K. Mazeau, J. Phys. Chem. B 111 (2007) 9138e9145. [9] A. Watanabe, S. Morita, Y. Ozaki, Biomacromolecules 7 (2006) 3164e3170. [10] K. Mazeau, Carbohydr. Polym. 117 (2015) 585e591. [11] K. Mazeau, A. Rivet, Biomacromolecules 9 (2008) 1352e1354. [12] J.F. Matthews, C.E. Skopec, P.E. Mason, P. Zuccato, R.W. Torget, J. Sugiyama, M.E. Himmel, J.W. Brady, Carbohydr. Res. 341 (2006) 138e152. [13] C. Verlhac, J. Dedier, H. Chanzy, J. Polym. Sci. Part A Polym. Chem. 28 (1990) 1171e1177. [14] C.K. Lee, E.J. Kim, J.H. Jun, Bull. Korean Chem. Soc. 20 (1999) 1153e1158. [15] A.P. Heiner, L. Kuutti, O. Teleman, Carbohydr. Res. 306 (1998) 205e220. chal, H. Chanzy, J. Mol. Struct. 523 (2000) 183e196. [16] Y. Mare n, Cellulose 13 (2006) 131e145. [17] K. Hofstetter, B. Hinterstoisser, L. Salme [18] C.M. Lee, K. Kafle, Y.B. Park, S.H. Kim, Phys. Chem. Chem. Phys. 16 (2014) 10844e10853. [19] C.M. Lee, J.D. Kubicki, B. Fan, L. Zhong, M.C. Jarvis, S.H. Kim, J. Phys. Chem. B 119 (2015) 15138e15149. [20] C.H. Lemke, R.Y. Dong, C.A. Michal, W.Y. Hamad, Cellulose 19 (2012)
142
E.L. Lindh et al. / Carbohydrate Research 434 (2016) 136e142
1619e1629. € bitz, J. Schaller, Carbohydr. Polym. 89 (2012) 640e647. [21] F.H. Larsen, M. Scho [22] D. Radloff, C. Boeffel, H.W. Spiess, Macromolecules 29 (1996) 1528e1534. [23] S. Park, J.O. Baker, M.E. Himmel, P.A. Parilla, D.K. Johnson, Biotechnol. Biofuels 3 (2010) 10. [24] H.P. Fink, D. Hofmann, B. Philipp, Cellulose 2 (1995) 51e70. [25] M.M. Maricq, J.S. Waugh, J. Chem. Phys. 70 (1979) 3300e3316. € m, J. Org. Chem. 68 (2003) 1671e1678. [26] S. Bekiroglu, L. Kenne, C. Sandstro , J. Magn. Reson. 268 (2016) 18e24. [27] E.L. Lindh, P. Stilbs, I. Furo [28] M.P. Bhate, J.C. Woodard, M.A. Mehta, J. Am. Chem. Soc. 131 (2009) 9579e9589. [29] C.R. Morcombe, K.W. Zilm, J. Magn. Reson. 162 (2003) 479e486. [30] A.E. Aliev, S.E. Mann, A.S. Rahman, P.F. McMillan, F. Cor a, D. Iuga, C.E. Hughes, K.D.M. Harris, J. Phys, Chem. A 115 (2011) 12201e12211. [31] A.N. Fernandes, L.H. Thomas, C.M. Altaner, P. Callow, V.T. Forsyth, D.C. Apperley, C.J. Kennedy, M.C. Jarvis, Proc. Natl. Acad. Sci. U. S. A. 108 (2011) 1195e1203. , Biomacromolecules 16 (2015) [32] C. Terenzi, K. Prakobna, L.A. Berglund, I. Furo 1506e1515. [33] P.T. Larsson, K. Wickholm, T. Iversen, Carbohydr. Res. 302 (1997) 19e25. [34] Y. Nishiyama, G.P. Johnson, A.D. French, V.T. Forsyth, P. Langan, Biomacromolecules 9 (2008) 3133e3140. [35] Y. Nishiyama, A. Isogai, T. Okano, M. Müller, H. Chanzy, Macromolecules 32 (1999) 2078e2081. [36] M. Schwanninger, J.C. Rodrigues, H. Pereira, B. Hinterstoisser, Vib. Spectrosc. 36 (2004) 23e40. [37] P. Bassan, A. Kohler, H. Martens, J. Lee, H.J. Byrne, P. Dumas, E. Gazi, M. Brown, N. Clarke, P. Gardner, Analyst 135 (2010) 268e277. [38] K.H. Liland, T. Almoy, B.H. Mevik, Appl. Spectrosc. 64 (2010) 1007e1016. [39] J.J. Cael, K.H. Gardner, J.L. Koenig, J. Blackwell, J. Chem. Phys. 62 (1975) 1145e1153. [40] E.T. Engelund, L.G. Thygesen, S. Svensson, C.A.S. Hill, Wood Sci. Technol. 47 (2013) 141e161. [41] D. Reishofer, S. Spirk, Adv. Polym. Sci. (2015) 1e22.
rez, D. Samain, in: H. Derek (Ed.), Adv. Carbohydr. Chem. Biochem., Ac[42] S. Pe ademic Press, 2010, pp. 25e116. [43] M. Bergenstråhle, J. Wohlert, P.T. Larsson, K. Mazeau, L.A. Berglund, J. Phys. Chem. B 112 (2008) 2590e2595. [44] J.F. Matthews, C.E. Skopec, P.E. Mason, P. Zuccato, R.W. Torget, J. Sugiyama, Carbohydr. Res. 341 (2005) 138e152. [45] J.F. Matthews, G.T. Beckham, M. Bergenstråhle-Wohlert, J.W. Brady, M.E. Himmel, M.F. Crowley, J. Chem. Theory Comput. 8 (2012) 735e748. [46] C. Jung, J. Mol. Recognit. 13 (2000) 325e351. [47] F. Schreier, J. Quant, Spectrosc. Ra. 112 (2011) 1010e1025. [48] S. Pronk, S. P all, R. Schulz, P. Larsson, P. Bjelkmar, R. Apostolov, M.R. Shirts, J.C. Smith, P.M. Kasson, D. van der Spoel, B. Hess, E. Lindahl, Bioinformatics 29 (2013) 845e854. [49] S. P all, M.J. Abraham, C. Kutzner, B. Hess, E. Lindahl, in: S. Markidis, E. Laure (Eds.), Solving Software Challenges for Exascale, Springer International Publishing, 2015, pp. 3e27. [50] B. Hess, C. Kutzner, D. van der Spoel, E. Lindahl, J. Chem. Theory Comput. 4 (2008) 435e447. [51] D. Van der Spoel, E. Lindahl, B. Hess, G. Groenhof, A.E. Mark, H.J.C. Berendsen, J. Comput. Chem. 26 (2005) 1701e1718. [52] E. Lindahl, B. Hess, D. van der Spoel, J. Mol. Model. 7 (2001) 306e317. [53] H.J.C. Berendsen, D. van der Spoel, R. van Drunen, Comp. Phys. Commun. 91 (1995) 43e56. lez-Outeirin ~ o, [54] K.N. Kirschner, A.B. Yongye, S.M. Tschampel, J. Gonza C.R. Daniels, B.L. Foley, R.J. Woods, J. Comput. Chem. 29 (2008) 622e655. [55] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L. Klein, J. Chem, Phys. 79 (1983) 926e935. [56] G. Bussi, D. Donadio, M. Parrinello, J. Chem. Phys. 126 (2007) 1. [57] H.J.C. Berendsen, J.P.M. Postma, W.F. van Gunsteren, A. DiNola, J.R. Haak, J. Chem. Phys. 81 (1984) 3684. [58] T. Darden, D. York, L. Pedersen, J. Chem. Phys. 98 (1993) 10089e10092. [59] U. Essmann, L. Perera, M.L. Berkowitz, T. Darden, H. Lee, L.G. Pedersen, J. Chem. Phys. 103 (1995) 8577e8593.