In-Depth Study of the Microcrystalline Cellulose Amino-Functionalization Eciency

Microcrystalline cellulose (MCC) has unique properties and its use as reinforcement for polymer composites has been increasing. However, the intrinsic incompatibility with most polymers requires surface modication to improve chemical compatibility prior to its incorporation into a polymer. In this paper, an in-depth study of silanization of MCC using 3-aminopropyltriethoxysilane (APTES), at different concentration, was done. The grafting amount of APTES onto MCC was determined by different methods: from residual mass and from nitrogen content. Solid-state 13 C and 29 Si nuclear magnetic resonance (NMR), eld emission scanning electron microscopy with energy dispersive X-ray (SEM-EDX), spectroscopy plasma optical emission spectrometry (ICP), and a deep study of structural properties by X-ray diffraction were carried out. A better correlation for grafting amount of APTES onto MCC was observed for nitrogen content method than residual mass according the Pearson’s correlation. 13 C NMR revealed all the carbon structures from cellulose and from APTES molecules and 29 Si NMR revealed D, T and Q Si structures. The silane treatment did not alter the shape of MCC and all treated samples showed Si characteristic peak at ~ 1.75 kEv. From ICP was observed higher Si content before MCC addition than after, evidencing, once again, APTES grafting. The exposure to APTES in acidic medium caused several effects on the MCC, splitting larger I β crystallites in half and along the more reactive hydrophilic sides. The diameter of the smaller IV I crystallites was largely reduced by the treatment, especially when the silane concentration was 1:5 (m/v), above which the diameter increases again.


Introduction
Microcrystalline cellulose (MCC) is applied in several elds, including as reinforcement in composites, since it combines renewable source with good thermal and mechanical properties. However, cellulose-based llers and polymers are not chemically compatible in general. Thus, surface modi cation of cellulosic reinforcements is recommended in order to produce a stronger interface between them, improving the nal composite properties (Trache et al., 2016).
There are different types of silanes used for the chemical modi cation of cellulose. In particular, the 3aminopropyltriethoxysilane (APTES) has the amino functional groups and is recommended for polymers such as epoxy, polyethylene, polyvinyl chloride, butyl rubber and polyacrylate (Xie et al., 2010). The chemical groups at the surface of the reinforcement can be modi ed to chemically interact with other materials (Xie et al., 2010;Neves et al., 2020), obtaining tailor-made materials for speci c applications.
The surface groups of cellulose-based materials can be studied using different methods such as nuclear magnetic resonance, thermogravimetric analysis or elemental analysis (Abdelmouleh et al., 2004;Baer et al., 2013;Sun et al., 2019). X-ray diffraction is widely used to determine crystal type, size of crystallites and crystallinity index. The most widely known and easiest way to analyze cellulose crystallinity is to use the Segal`s method (Segal, et al., 1959). This method has relatively low accuracy but it is well-accepted for comparison purposes (Segal et al., 1959;French & Santiago, 2013;French, 2014French, , 2020. Posteriori methods, which rely on the preliminary knowledge of lattice or other parameters, are based on curve-tting procedures, such as the Debye or the Rietveld method (Thygesen et al., 2005;Driemeier, 2014). All the methods have drawbacks, and most are based on the identi cation of prominent peaks, which are not considered to overlap in the 10°-50° range of 2θ. Duchemin (2017) developed a free spreadsheet program that determines the shape, orientation, and crystallinity of cellulose Iβ using X-ray raw data. To summarize, a semi-empirical background is set and a small number of adjustable parameters are iterative-simulated until convergence. This software was successfully applied on raw, non-treated, microcrystalline cellulose (Duchemin, 2017) and on cellulose nanocrystals (Fernandes et al., 2020), but not for silane-modi ed MCC.
Thus, this study focuses on the determination of the APTES grafting index on the surface of MCC using two methods, followed by an in-depth structural characterization by X-ray diffraction, along with thermogravimetric and elemental analysis. Complementary techniques to evaluate the effect of the silanization included Fourier-transform infrared spectroscopy (FTIR), plasma optical emission spectrometry, solid-state 13 C and 29 Si nuclear magnetic resonance, and eld emission scanning electron microscopy with energy dispersive X-ray spectroscopy.

Surface modi cation of MCC
The modi cation of the MCC surface was reported in previous work  and is illustrated in Fig. 1 along with the characterization carried out at each step. Brie y, (i) the MCC is homogenized in an alcoholic solution (75% v/v) acidi ed at pH 4-5 using acetic acid and kept under stirring for 30 min at room temperature (rT); (ii) The silane undergoes hydrolysis and is also kept under stirring for 30 min at rT; (iii) APTES and MCC are stirred together for 2 h at (rT) for the condensation, followed by washing of the unreacted silane and centrifugatio; (iv) the MCC is grafted at 105°C for 15 min. The samples were named: MCC (raw microcrystalline cellulose) and MCC-Si (1:X) (m/v). The APTES relation choosen was 3, 4, 5 and 10 ml.

Characterization
The amount of silane grafted onto the MCC surface was determined by thermogravimetric analysis (TG) and elemental analysis (CHN). Complementary analysis perfomed included: solid-state 13 C and 29 Si Nuclear Magnetic Resonance (NMR), Plasma Optical Emission Spectrometry (ICP-OES), Field Emission Scanning Electron Microscopy (FEG-SEM) coupled by Energy Dispersive X-ray Spectroscopy (EDS) and X-ray Diffraction (XRD).
The thermogravimetric analysis (TG) was performed on all samples (modi ed or not) using Shimadzu TGA-50 equipment in an N 2 atmosphere with a 50 mL.min − 1 purge rate, using a heating rate of 10°C min − 1 from 25 to 700°C, and ca. 10 mg of sample mass. The amount of APTES grafted onto the MCC surface was quantitatively determined based on the TG residue (char) according to Eqs. where W N is the content (mass%) of nitrogen from the elemental analysis and M N is the molar mass of N (14.01 g/mol).
Solid-state 13 C measurements were performed using resonance frequency of 125 MHz, spinning rate 10 kHz, cross-polarization (CP) contact time 7 ms, recycle delay 5 s, SPINAL-64 1H decoupling and acquisition time 43 ms. The chemical shift was referenced against hexametilbenzene (38.3 ppm). 29 Si NMR measurements were performed using resonance frequency of 99.3 MHz and chemical shift was referenced against talc (-90 ppm). Samples with lower silane content (MCC-Si (1:3) and (1:4)) were analyzed for 29 Si at longer exposure times, once they have the lowest APTES content. Both analyzes were performed on an Agilent DD2 500 MHz spectrometer (California, US), equipped with a 4mm dual broadband CP-MAS probe. The morphology and composition of the MCC-Si samples were characterized using Tescan brand equipment -model FEG Mira 3 (Czech Republic) by SEM and EDX. All samples were previously coated with Au and the acceleration voltage used was 15 kV.
Quanti cation of silicon content in the aqueous solution was performed by ICP after the APTES hydrolysis (step (ii), Fig. 1) and after the condensation (step (iii), Fig. 1). The analysis was performed in a SMEWW 3120 B in ICP (ThermoSienti c -ICAP 7000 SERIES) equipment. The samples were submitted to a digestion in SMEWW 3030-E. The Si analysis had the limit of quanti cation of 0.01 mg/L. FTIR-ATR was performed with the help of a Nicolet device, model IS10 Thermo Scienti c. Each spectrum was the accumulation of 32 scans in the 4000 − 400 cm − 1 wavenumber range. The data treatment was performed with Orange 3.27 (Demšar et al., 2013). Three regions were treated separately: the -OH region between 3600 and 3050 cm − 1 , the -CH region between 3000 and 2700 cm − 1 and the region between 1800 and 650 cm − 1 . The regions were baseline subtracted (linear baseline for the two rst regions and rubber band correction for the last one) before being area normalized. A second derivative transformation was performed in order to determine peak positions with a Savitsky-Golay lter (99 points window, 2nd order polynomial approximation).
XRD measurements were performed with a Shimadzu XRD-6000 diffractometer at 40 kV and 30 mA with a CuK α beam (λ = 0.1542 nm). The intensities were measured in the 2°<2θ < 30° range (0.05°/4 s). The results were interpreted with the help of a routine reported in (Duchemin, 2017;Kheli et al., 2018;Leboucher et al., 2020). Brie y, a Rietveld analysis was performed with the hypothesis that the diffraction signal was constituted of three components: cellulose I β with large crystallites and relatively long-range order, a form of native cellulose with small crystallites (usually below 2 nm in lateral cross-section) and a background signal with a constant value over the whole angular range. This bimodal size distribution is justi ed by the coexistence of proto brils aggregates, and it has been successfully used to explain anisotropic recrystallization, The concept of multimodal size distributions has recently been reported by another team using a simulation matching approach (Rosén et al., 2020). In contrast, the approach used here is based on Rietveld re nement and it enables more parameters to be tted from the experimental data. The larger crystallites were modelled after the I β coordinates originally published by Nishiyama et al., (2003). The tting parameters were intensity, lattice parameters, March-Dollase (0 0 l) texture, and crystallite dimensions and shape. The line shape was described with a Pearson VII function and a shape parameter of 1.88. The geometry of these crystallites was assumed to be that of a superellipsoid, as described in (Rosén et al., 2020). The smaller crystallites were modelled using the IV I unit cell assuming the absence of a preferred orientation, and only two parameters were re ned, crystallite diameter (the crystallites were considered spherical) and relative intensity. Since the silanization process occurred in an acidic medium (pH: [4][5], some breaking of the β1-4 glycosidic bonds may happen, producing more REs. The reduced MCC chains would tend to thermally degrade more easily than the original cellulose chain. This behavior was more pronounced for MCC-Si (1:5). A similar trend was reported by (Agustin et al., 2016) who studied the thermal characteristics of nanocrystals of cellulose, (Matsuoka et al., 2011) who studied MCC thermal glycosylation and degradation reactions, and (Leboucher et al., 2020) who studied hydrolyzation of cellulose nanocrystals. In all cases, the thermal stability was related to the chain ends rather than to crystallinity.
The results for grafted aminosilane amount onto MCC surface from TG curves are displayed in Fig. 2c. It is possible to observe that increasing the APTES content in the sample, increase the silane grafting amount. In other words, by increasing APTES in the solution, the quantity of aminosilanes able to bond with MCC slightly increases. These results corroborate with (Piscitelli et al., 2010) nds when they treated sodium montmorillonite clay with this silane.
The content of carbon (C), nitrogen (N) and hydrogen (H) elements in MCC-Si can be found in Table 1. The MCC treated with higher content of APTES showed a slightly higher nitrogen content, being the highest for MCC-Si (1:5). According to Hassanpour et al. (2017), higher N content can be related to a successful grafting of the aminosilane onto the cellulose surface. Saini, Belgacem, Salon, & Bras, (2016) also found that the initial silane concentration signi cantly affected the grafting e ciency. The content of APTES (M APTES ) in MCC calculated from the N amount, is also shown in Table 1. A Pearson correlation plot was done to identify possible correlations among the variables (Fig. 3). The red and blue ellipses refer to positive and negative correlations, respectively, and the smaller the ellipse, the stronger the correlation between the variables (also indicated by the p-value). The results indicate a better correlation between the (M APTES ) ( Table 1) and the N content than between the silane grafted amount and the residual TG (Fig. 2c).

Solid-state 13 C and 29 Si NMR results
In a previous publication , NMR spectra of solid-state 13 C from 130 to 30 ppm and NMR spectra of solid-state 29 Si from − 20 to -100 ppm were reported, from which cellulose carbons C1 to C6, characteristic of cellulose samples, and three Si structures of: T 1 (dimer), T 2 (linear) T 3 (three-dimensional structures), of the siloxane structure were reported. In the current paper, NMR measurements we performed at higher exposure times (total time of the analysis: 60h), extending the ppm axis. Figure 4 shows the 13 C NMR spectra of MCC and MCC-Si samples, and the chemical shift assignments of the 13 C are summarized in Table 2  α-C -10 10 10 11 Figure 5 shows the 29 Si NMR spectra of MCC and MCC-Si samples, from − 5 to -35 ppm and the chemical shift assignments of the 29 Si are summarized in  The type of structure formed is given by the self-condensation of the silanol groups, i.e. "T" structures.
However, in the current study, two other structures appeared, "D" and "Q", due to the longer exposure times in the 29 Si NMR analysis. The notations "D", "T", and "Q" represent the different types of silicate structures commonly encountered in 29 Si NMR studies, with two, three and four Si-O-bonds, respectively (see Fig. 5b). More information can be found in (Glaser & Wilkes, 1989, Brochier et al., 2005. Based on Fig. 5, "T" structures were found to be dominant, and more pronounced for higher contents of APTES (MCC-Si (1:5) and (1:10)). In all samples, "D" and "Q" structures also appeared. The chemical modi cation of cellulose is applied, in general, to add functional groups that will interact with some polymers, producing a stronger cellulose/polymer interface in a composite. Since the chemical condensation and grafting of APTES onto cellulose happens through Si-O-C bonds (Abdelmouleh et al., 2004), "T" and "Q" structures are interesting for having more oxygen to bond with cellulose. The FTIR study showed marked differences between the different samples ( Fig. 8 and Table 4). In general, in the -OH region, the bands assigned to the weak intramolecular O(2)H···O(6), to the intermolecular

O(6)H···O′(3) and to the intramolecular O(3)H···O(5) showed a displacement towards higher wavenumber
values (Table 4) which is assigned to a constrained mobility of these -OH groups. The "weak" band is often assigned to the most accessible cellulose portions and it is strongly shifted after APTES treatment, which supports a grafting at the periphery of the crystalline portions. The main -CH band near 2900 cm − 1 is also affected by the treatment and it shows a slight wavenumber increase of 4 cm − 1 at the lowest relative concentration (MCC-Si 1:3), which progressively disappears as the concentration is increased. It is often thought that this band has little sensitivity to treatments, but this is in fact not true and it can be affected both by chemical and thermal treatments (  The X-ray diffraction spectra were analyzed using a bimodal size distribution model, shown. The results are shown Fig. 8 and Table 4. The crystallite thickness in the direction orthogonal to the (200) plane is ~ 2.R x ± 7% and an increase of this value can be understood as a co-crystallization occurring on the more hydrophobic planes, as detailed in previous works (Duchemin, 2017;Leboucher et al., 2020). In contrast, the crystallite thickness in the direction orthogonal to the (010) plane is equal to 2.R y and an increase of this value can be understood as a co-crystallization occurring on the more hydrophilic planes. Several things can therefore be assessed from the X-ray tting work. Regarding the MCC sample, the relative allomorphic content, the crystallite shape and the dimensions of the two allomorphs are very much in line with our previous work, despite differences in the MCC origin and measurement setup (Duchemin, 2017). The exposure to the APTES in the acidic medium had several effects on the initial MCC. Regarding the dimensions of the larger cellulose I β crystallites, their diameter 2.R x is decreased by ~ 25% along the hydrophobic sides for all the treatment conditions except for the treatment with the highest APTES concentration (MCC-Si 1:10) for which 2.R x remains the same (Table 4). More strikingly, the diameter 2.R y is roughly halved by all of the treatment conditions, but this is also slightly less marked for the MCC-Si 1:10 sample. This means that, statistically, the silanization splits the larger I β crystallites in half and along the more reactive hydrophilic  These measurements can be interpreted in the light of the previous 29 Si NMR and ATR-FTIR results. During the APTES treatment, reactive silanols are created and they condense with the hydroxyl groups of cellulose. Since these hydroxyl groups are found in priority along the hydrophilic surfaces, the condensation reaction should occur on these sites rst. This is what is observed here since the 2.R y values are halved after the APTES treatment, meaning that the crystallites were split apart along the planes. Interestingly, the silane treatment is therefore able to act as a spacer between the hydrophilic sides of the cellulose crystallites.
This transformation is correlated with the cross-sectional shape of the I β phase: a swap from a lateral diamond shape (MCC) to a square shape occurs at all treatment conditions ( Fig. 9 and Table 5). It is also correlated with the content reduction of slender IV I crystallites with respect to larger cellulose I β , possibly due to dynamic reorganizations and perhaps recrystallization of the former phase.
When the highly reactive silanol structures are formed, they can also react to some extent with the less reactive "hydrophobic" cellulose sides, which explains the crystallite size change along a (R x values). However, when the APTES concentration is increased in the acidic solution, the acidic hydrolysis of the APTES moieties evolves and certain forms become more frequent. For instance, the prevalence of the more reactive T 3 structure seems to be at an optimum for the 1:5 concentration (Fig. 6a), resulting in the lowest R x values observed (Table 5). However, when the APTES concentration is too high, self-condensation might occur and bulkier, less reactive silanols are created that are unable to react along with the accessible "hydrophobic" groups. This explains the similarities between the initial and 1:10 condition in terms of crystallite size (R x remains the same), lattice parameters (less lattice distortions due to lower reactivity), and allomorphic content (the unchanged prevalence of the IV I form).

Conclusions
An in-depth study regarding MCC surface-functionalized using different ratios of APTES was performed. In all samples, APTES was grafted onto the MCC surface as proved by two different methods, CHN and TG residual mass. For the latter, higher APTES content yielded higher grafting amount, whereas for the former, the highest grafting amount was found for the MCC-Si (1:5) sample. A higher Pearson correlation was found for the grafting amounts calculated from the %N content.
Moreover, the solid-state 29 Si NMR of the MCC-Si samples presented dominant "T" structures (three Si-Obridges) compared to "D" and "Q" structures (two and four Si-O-bridges, respectively). The allomorphic content obtained by XRD analysis presented dominance of IV I crystallites for MCC and MCC-Si (1:10), and prevalence of I β crystallites for MCC-Si (1:3) to (1:5)). During the APTES treatment, the condensation reaction occurred mostly along the hydrophilic surfaces, acting as a spacer between the hydrophilic sides of the cellulose crystallites. For the MCC-Si (1:10), due to the high concentration, the greater similarity with the initial MCC is justi ed in terms of self-condensation and due to the less reactive silanols which could not react any further with the accessible hydrophobic groups.
Declarations Figure 1 Illustration of the whole silanization process and the characterization carried out.
Page 17/23   Solid-state 13C for MCC and MCC-Si samples.  Si amount (mg/L) before and after APTES condensation onto MCC for the MCC-Si samples.

Figure 8
Absorbance FTIR spectra of the samples before and after APTES condensation on MCC. The spectra were area normalized for each region of interest in order to improve readability.