Hydroxyl groups on cellulose nanocrystal surfaces form nucleation points for silver nanoparticles of varying shapes and sizes.

In this study, we investigate the interactions between the cellulose surface and Ag nanoparticles (AgNPs) for the purpose of manufacturing hybrid nanomaterials using bacterial cellulose nanocrystals (BCNs) as a model substrate. We focus on the role of the BCN surface chemistry on the AgNP nucleation obtained by chemical reduction of Ag+ ions. Homogeneous hybrid suspensions of BCN/AgNP are produced, regardless of whether the BCNs are quasi-neutral, negatively (TBCNs) or positively charged (ABCNs). The characterization of BCN/AgNP hybrids identifies the -OH surface groups as nucleation points for AgNPs, of about 20 nm revealing that surface charges only improve the accessibility to OH groups. X-ray Absorption technics (XANES and EXAFS) revealed a high metallic Ag0 content ranging from 88% to 97%. Moreover, the grafting of hydrophobic molecules on a BCN surface (HBCNs) does not prevent AgNP nucleation, illustrating the versatility of our method and the possibility to obtain bifunctional NPs. A H2O2 redox post-treatment on the hybrid induces an increase in AgNPs size, up to 90 nm as well as a shape variation (i.e., triangular). In contrast, H2O2 induces no size/shape variation for aggregated hybrids, emphasizing that the accessibility to -OH groups ensures the nucleation of bigger Ag nano-objects.


Introduction
Cellulose is an almost inexhaustible biopolymer extracted from wood, cotton, algae, tunicates or bacteria [1][2][3], leading to variation in dimensions and structural organization [4]. Cellulose displays interesting properties such as high water retention capacity, high wet strength, low density, biocompatibility, non-toxicity and biodegradability, which make it very appropriate for several applications in biomedicine and pharmacology [5][6][7], cosmetics [8], paper and textiles [9]. Cellulose can also be hydrolyzed to increase its crystallinity, commonly by acid hydrolysis, leading to the socalled cellulose nanocrystals (CNCs) [10]. While the extraction of plant-sourced cellulose often requires the use of hazardous chemicals, bacterial cellulose (BC) is mainly secreted extracellularly by certain bacteria such as Gluconacetobacter xylinu [1]. Therefore, bacterial cellulose nanocrystal (BCN) contain no functional groups other than hydroxyl groups [11] and thanks to its high chemical purity, they are particularly interesting for food and medical applications.
To confer antimicrobial properties to CNCs, as to BCNs, it is quite common to graft or nucleate metallic nanoparticles (e.g., Cu, Ag, ZnO, Au) on their surface [12][13][14][15]. Among others, silver nanoparticles (AgNPs) have been widely studied. A silver precursor is generally incorporated by addition of a salt such as silver nitrate (AgNO 3 ), and several experimental methods (e.g., UV radiation [16,17], hydrazine reduction [18]) have been proposed to nucleate AgNPs on the cellulose nanocrystal surface, thus obtaining a hybrid material (i.e., CNC/AgNPs). Chemical reduction is one of the most common ways to synthesize metallic NPs, and sodium borohydride (NaBH 4 ) [19][20][21][22] is reported to be one of the most efficient reducing agents, inducing the rapid formation of 2-3-nm AgNPs [23]. NH 4 OH [24] and ascorbic acid [25] can also be used to reduce Ag + ions into AgNPs. To carry out this chemical AgNP synthesis, capping agents or stabilizers (e.g., trisodium citrate: TSC; polyvinylpirrolidine: PVP; Cetrimonium bromide: CTAB) are often introduced to better control the morphological properties of AgNPs, preventing their aggregation [26][27][28]. In other cases, nanocellulose is proposed as both a support and a reducer for the generation of metallic nanoparticles [29]. Xiong et al. [30] claim that cellulose promotes the generation of AgNPs and dendritic Ag nanostructures thanks to the extensive presence of surface hydroxyl groups. Furthermore, various surface modifications of CNCs can be performed (e.g., amidation, oxidation, esterification, etherification) [10], possibly affecting the reduction of AgNPs on the CNC surface. Most of the studies concerning CNC/AgNP hybrids [31][32][33] indicate how the introduction of negative surface charges (i.e., via TEMPO-mediated oxidation [32]) can provide the high-binding capability for the transition of metal species such as Ag + ions. Moreover, other surface modifications can be performed on the CNC surface to extend the possible application fields. For example, cationized cellulose can be produced by grafting different molecules on the cellulose surface (e.g., HPTMAC: hydroxypropyltrimethylammonium chloride [34], BriBBr: a-bromoisobutyryl bromide [35], HDTMA: hexadecyltrimethylammonium [36] or pyridinium [37]) so that it can be used for papermaking, cosmetics and drug-delivery [38]. The work of Shateri Khalil-Abad et al. [39] reveals that surface-modification of cellulose cotton fibers by 3-chloro-2-hydroxypropyl trimethyl ammonium chloride (CHPTAC) affects the absorption of AgNPs on the fiber surface. Indeed, such an adsorption is greater on cationized cellulose fibers than on untreated ones. To obtain these cationized fibers, Dong et al. [40] propose to graft ammonium ions onto the surface, thus creating two different pathways for the deposition of metal NPs (e.g., Au, Pt, Pd): the first one based on the electrostatic assembly of metal nanoparticles capped with negative citrate ions, and a second one where negative metal complex ions are adsorbed onto the cationic substrate and then reduced. Both approaches lead to a high surface coverage of the CNC surface by metallic nanoparticles.
CNCs can also be made hydrophobic to allow their dispersion in non-aqueous solvents. To do this, one of the most widely used approaches is amidation, where coupling agents are grafted onto TEMPOoxidized nanocrystals [10]. Araki et al. [41] propose the grafting of poly(ethylene glycol) with a terminal amino group onto the surface of TEMPO-oxidized CNCs. Zhang et al. [42] use a similar experimental approach to covalently functionalize CNCs via peptide coupling chemistry with different alkylamines (e.g., propylamine, n-butylamine, amylamine, hexylamine, heptylamine). Hu et al. [43] propose using tannic acid to covalently attach primary amine with a long alkyl chain (i.e., decylamine) to the CNC surface. In addition, Cunha et al. [44] produced inverse emulsions, tailoring the hydrophobicity of CNCs and nanofibrillated cellulose by chemical modification with lauroyl chloride (C 12 ). However, no studies have proposed bifunctional hybrids, using hydrophobically-modified CNCs as substrates for AgNP nucleation.
As for nanocellulose surface modification, AgNP characteristics (e.g., size, shape, crystalline structure) can strongly influence the final properties of the CNC/AgNP hybrid system. Some works report simple techniques for a shape-and size-selective synthesis of Ag nanostructures (e.g., from spherical AgNPs to triangular-shaped AgNPs or Ag nanoprisms, denoted AgNPrisms [45]) based on the use of hydrogen peroxide (H 2 O 2 ) redox post-treatment [20,21,46,47], as well as photo-induced [48] or solution-phase [49] methods. In particular, the size-shape transformation of AgNPs achieved with H 2 O 2 relies to redox capabilities linked to the autocatalytic decomposition on the Ag surface. These AgNPrisms are anisotropic (i.e., their lateral dimension is greater than their thickness) and are thus characterized by a strong localized surface plasmon resonance (LSPR). Their resulting size-and shape-dependent optical properties make them suitable for sensors [50,51] and biological imaging [52], catalysts [53], nanophotonic devices and circuits [54,55]. Parnklang et al. [46] show the shape transformation from spherical AgNPs to AgNPrisms when H 2 O 2 oxidizes the AgNPs to Ag + ions and thus reduces them into Ag atoms in nanometric silver particles. The H 2 O 2 redox post-treatment has also been applied to hybrid systems. Jiang et al. [20] propose the fabrication of films made of AgNPrisms and TEMPO-oxidized cellulose nanofibrils, which work as capping and shape-regulating agents. The presence of a predominant (111) peak in the XRD pattern allows them to affirm that AgNPrisms are bounded to cellulose nanofibers.
The result is that hybrids made of AgNPs fixed on the CNC surface have been produced, whereas the interactions and binding mechanisms have not been studied in detail [29,31,56]. In this paper, we propose to perform different surface modifications on BCNs in order to shed light on the effective role of cellulose nanocrystal surface chemistry on the nucleation of AgNPs. We also prepared bifunctional NPs that combine a hydrophobic surface and AgNPs. Moreover, we investigated the impact of the CNC surface modification on the H 2 O 2 redox post-treatment, linking the size and shape variation of AgNPs to their physicochemical characteristics (i.e., AgNP structure and oxidation state).

Synthesis of uncharged BCNs.
For the preparation of BCNs, the experimental procedure proposed by Kalashnikova et al. [57] was followed. Briefly, nata de coco cubes were dialyzed in ultra-pure water for 9 days, after which they were ground in a Waring blender, suspended in 0.5 N NaOH solution, and stirred for 2 h at 70°C. Finally, the sample was rinsed using ultra-pure water. A whitening treatment was performed twice, dispersing the slurry in a NaClO 2 solution (8.5 g/L) in a sodium acetate buffer (pH = 4.5) and stirring it for 2 h at 70°C. The bleached cellulose was rinsed with distilled water and then hydrolyzed with 2.5 M HCl solution at 70°C for 2 h. The suspension was centrifuged three times (10 min., 10000 g) and dispersed in ultra-pure water up to pH > 5. The final BCN suspension was again dialyzed against ultra-pure water for one week and finally homogenized by ultrasound (amplitude = 20; 5 min). The BCN content in the suspension (around 2 g/L) was determined by a drying test.

Synthesis of anionic BCNs mediated via TEMPO oxidation (TBCNs). The experimental protocol
for the BCN surface carboxylation via TEMPO-mediated oxidation was adapted from the method proposed by Araki et al. [58]. A quantity of 100 mL of BCN suspension (2 g/L) was mixed with 0.25 g of NaBr and 0.05 g of TEMPO, after which 2.5 mL of NaClO were added to start the BCN carboxylation, leading to a pH increase. The resulting suspension was stirred for 24 h at room temperature while maintaining the pH at 10.3 through the controlled addition of 0.1 M NaOH solution using an automatic titrator (Metrohm 901 Titrando, France). The final suspension was centrifuged twice (10 min; 10000 g), dispersed in ultra-pure water while maintaining the BCN concentration at 2 g/L and, finally, homogenized by ultrasound (amplitude = 10; 5 min). The sample was dialyzed against ultra-pure water for one week to remove any residual contaminants (dialysis bath volume to sample volume = 10:1).

Synthesis of cationic BCNs (ABCNs).
For the BCN surface modification by AETMA, the coupling method proposed by Zhang et al. [42] was used with several modifications, working in a H 2 O medium instead of N,N-dimetilformammide (DMF). A quantity of 100 mL of TBCN suspension (2 g/L) was mixed for 5 min at room temperature with 230 mg of EDC. Then, 180 mg of NHS were added and the suspension was stirred for a further 30 min. Finally, a quantity of 210 mg of AETMA were introduced and the sample was stirred for 24 h at room temperature while maintaining the pH at 7 through the controlled addition of 0.1 M NaOH solution using an automatic titrator (Metrohm 901 Titrando, France). The final suspension was centrifuged to remove unreacted residual molecules and dispersed in ultra-pure water to be homogenized by ultra-sound (amplitude = 10; 5 min). The suspension was dialyzed for one day against a KNO 3 -saturated solution to promote ion exchange from chloride to nitrate, and then against ultra-pure water for one week (dialysis bath volume to sample volume = 10:1).

Synthesis of hydrophobic BCNs (HBCNs).
To obtain HBCNs, the TBCNs were functionalized with alkyl chains (C 8 ) using a coupling method similar to the one previously proposed for the preparation of    [58]. The data were recorded by Tiamo TM Titration software and the conductivity values were corrected from dilution effects and thus plotted against the added volume of sodium hydroxide solution. The inflection point was graphically determined from the intersection of the least squares regression lines fit and the data points in the distinct regions of the titration curves, evaluating the NaOH volume required for neutralization and, thus, the surface charge density [59,60], ρ (mol/g). Using ρ values, the degree of oxidation [61] (DO) was calculated as: where represents COOH moles (mol) deduced from the surface charge density, ; was the n COOH M AGU molecular weight of an anhydroglucose unit (162 g/mol); corresponded to BCN mass and m cellulose was the molecular weight of a carboxylic group.

M COOH
The zeta potential values were measured using a ZetaSizer Nano ZS (Malvern, UK). BCN suspensions were diluted at 0.1 g/L and filtered (pore diam. = 5 μm). Five measurements at T = 20°C were performed for each sample. For the case of HBCNs suspended in a H 2 O/EtOH mixture, the residual ethanol was removed by dialysis of several milliliters of suspension against water for 2 days. All the suspensions were sonicated just before the measurements.
X-ray Absorption Near-Edge Structure (XANES) measurements were performed to investigate the AgNP oxidization state (i.e., metallic silver, Ag 0 ; ionic silver, Ag + ) while Extended X-ray Absorption Fine Structure (EXAFS) data made it possible to shed light on the AgNP bulk atomic structure (e.g., bond length, interatomic distance). Both XANES and EXAFS spectra were simultaneously recorded in transmission mode at the Ag K-edge (25250 to 27750 eV) on the SAMBA beamline at the SOLEIL synchrotron (Saint Aubin, France). The Si (220) monochromator was calibrated to 25515.6 eV at the first inflection point of the Ag foil XANES spectrum. To be analyzed, all the hybrids were freezedried and then compressed to obtain a circular pellet of 6 mm where the quantity of AgNP was high enough to reach an absorption edge jump close to 1. These pellets were placed on a sample rod, quenched in liquid nitrogen and then introduced into the He cryostat (T = 20 K). Silver foil (Agfoil) and AgNO 3 aqueous solution with 1 wt% glycerol (AgH 2 O) were used as standards. One continuous scan was recorded for each hybrid sample in the 25250 to 27750 eV energy range with a monochromator velocity of 5 eV/s and an integration time of 0.08 s/point. The obtained scans were then normalized and background-subtracted using the Athena software package [62]. The XANES spectra were analyzed by a linear combination fitting (LCF) procedure in the E 0 -20 eV, E 0 + 50 eV energy range, with E 0 set to 25514 eV and using Agfoil and AgH 2 O standards as components. All component weights were forced to be positive and the relative proportions of the components were forced to add up to 100%. Concerning the EXAFS oscillation, a background subtraction was performed before applying an autobk algorithm (Rbkb = 1, k-weight = 3). The Fourier transform of the k 3 -weighted EXAFS spectra was then calculated over a k range of 2.5-19.5 Å -1 , using a Hanning apodization window (width of the transition region window parameter = 1). The k 3 EXAFS fitting was performed in the 2.35-7.7 Å distance range with the Artemis interface [62] to IFEFFIT using leastsquares refinements. Paths used for fitting standards and samples were obtained from a metallic silver crystallographic model [63] using the FEFF6 algorithm included in the Artemis interface. Only paths with a rank higher than 7% were considered, and the E 0 value was set to 25520 eV. The amplitude reduction factor S 0 ² was determined to be equal to 0.978 by fitting the 1 st coordination sphere of the Agfoil spectrum over a range of 2.30-2.83 Å. This value was used in all the fitting procedures.
Degeneracy of the paths, energy shift ΔE 0 , radial distance shift ΔR, and thermal and static disorder σ² were fitted for each of the selected paths for a total of 57 independent points and 19 variables. All Rfactors were lower than 0.02.
The pellets prepared for XANES-EXAFS measurements were then analyzed by X-ray powder diffraction (XRD). The XRD diffractograms were recorded in 10 min on a Bruker D8 Discover diffractometer (France). Cu-Kα1 radiation (Cu Kα1, 1.5405 Å) produced in a sealed tube at 40 kV and 40 mA was selected and parallelized using a Gobël mirror parallel optics system and collimated to produce a 500-mm beam diameter. The data were collected in the 3°-70° 2 range. XRD θ measurements were also performed for initial BCN suspensions with different surface modifications.
The crystallinity index (CrI, %) was calculated as a function of the maximum intensity of the diffraction peak from the crystalline region (I 200 ) at 2 = 22.5° and the minimum intensity from the θ amorphous region (I am ) at 2 = 18°, according to the Segal equation [64]: θ The crystallite size (CS) was determined using Scherrer's equation [65]:

CS = Kλ βcosθ
where is the shape factor (0.9), is the X-ray wavelength (1.54 Å), is the full-width at half-K λ β maximum (FWHM) and is the angle of the diffraction peak of the crystalline phase (Bragg's angle). θ The FWHM was determined considering the characteristic peak at 2 = 22.5° for initial BCN θ suspensions and the peak at 2 = 38° for the AgNPs in hybrids. θ The light-visible absorbance of hybrid suspensions was measured in the 300-800 nm range using a Mettler-Toledo UV7 (USA) spectrophotometer equipped with a 10-mm quartz cell. All the samples were water-diluted (1:10) and ultra-pure water was used as a blank reference.
To determine the real content of AgNPs in hybrid suspensions, a volume of 1 mL of sample was digested by 40 mL water/aqua regia mixture (i.e., 30% v aqua regia, HCl/HNO 3 : 3/1). The resulting suspensions were then analyzed by atomic absorption spectroscopy, AAS (ICE 3300 AAS, Thermo Fisher Scientific, USA). A calibration curve was obtained by the measurement of a digested sample of silver standard solution (1000 μg/mL, Chem-Lab NV, Belgium) at different concentrations, from 0.5 to 10 ppm. Two independent measurements were repeated for each sample.
For scanning transmission electron microscope (STEM) observations, the as-synthesized suspensions were water-diluted at 0.5 g/L in BCN content and 10 μL were then deposited onto glow-discharged carbon coated grids (200 meshes, Delta Microscopies, France) for two minutes, removing the excess by touching the edge of the drop with Whatman filter paper. The grids were dried overnight in air and then coated with a platinum layer (thickness = 0.5 nm) by an ion-sputter coater (LEICA EM ACE600, Germany). Brightfield images were recorded using a field emission gun scanning electron microscope (Quattro S, Thermo Fischer Scientific, USA) operating at 10 kV with a STEM detector. For each sample, the acquired images were analyzed by ImageJ software. The AgNP Feret diameter (i.e., the largest distance between two tangents to the contour of the measured particle) was determined considering the largest possible number of AgNPs (from 35 to 100, depending on the sample). FTIR spectroscopy assessed the effective surface-modifications. As shown in Fig. 2a, the unmodified BCNs showed well-defined peaks at 1000 cm -1 and 3000-3500 cm -1 , typical of-OH and -C-O-Cgroups, respectively [66]. The TBCN spectrum displayed a strong absorption band between 1630 cm -1 (carboxylate) and 1730 cm -1 (acid), which was characteristic of the carboxyl groups in their acidic form, validating a successful TEMPO oxidation of BCNs [61]. TBCNs were further surface-modified to obtain two types of surface chemistry. Firstly, a cholamine chloride derivative (AETMA) was used to synthesize aminated cationic BCNs (ABCNs). For this sample, the FTIR spectrum revealed the presence of two new bands at around 1580 cm -1 assigned to the C=O stretching of the amide I band and to the N-H vibration of the amide II band, respectively [67]. Secondly, an octylamine C 8 was covalently grafted onto the carboxyl group of TBCNs using NHS/EDC, leading to a hydrophobic BCN (HBCN). The good degree of coupling between TBCNs and C 8 was proved by the very low surface charge density measured by conductometric titration (i.e., 0.010 mmol/g). The detection of these same bands in the spectrum of HBCNs and the presence of asymmetrical and symmetrical CH 2 stretching from the C 8 alkyl chain at 2850 and 2930 cm -1 indicated the successful octylamine grafting [43].

Characteristics of surface-modified
Nevertheless, a residual peak at 1730 cm -1 was still visible probably due to remaining free COOH groups . FTIR results were corroborated by the variation of the zeta potential (), implying that a modification of the electrical charge environment of the BCN was induced by the different chemical surface modifications. While pure BCNs were quasi-neutral, TBCNs showed a negative  value of -25 mV, correlated to the variation of surface charge density (ρ). On the other hand, a clear increase in  to +16 mV was observed for the aminated ABCNs due to the substitution of COOgroups by the N + of the AETMA molecules. The slightly negative value of -9 mV was measured for the alkyl hydrophobized HBCNs because of the grafting of octylamine onto the surface carboxyl groups. This low value indicated that HBCNs were highly substituted and no longer electrostatically but, instead, sterically stabilized. These experimental evidences were consistent with a stable suspension of highlycharged colloids for TBCNs and ABCNs with respect to BCNs.
The surface modifications affected the nanocrystal dispersions, as shown from STEM images of dry samples in Fig. 1. Even though the global morphology of the nanocrystals remained unchanged after modification, it could be observed that TBCNs and ABCNs were better dispersed in comparison to quasi-neutral unmodified BCNs and HBCNs. Furthermore, TBCN and ABCN suspensions were less opaque than those of pure BCNs and HBCNs (Fig. 1). Analysis of STEM images indicated an average length of between 850 and 1750 nm and a width of between 20 and 40 nm for all surface-modification treatments, which was in agreement with several studies in the literature [57,68,69]. All the outcomes are summarized in Table 1, indicating that the BCN surface was successfully modified for all of the cases considered.  respectively, according to the triclinic indexation of Nishiyama et al. [70]. As reported in Table 1, the crystallinity was unchanged around 60% after correction of the amorphous for all the samples, and the chemical treatment did not affect the size of crystallites, which was 5.9 ± 0.2 nm for all the samples, in agreement with the value found by Vasconcelos et al. [71]. These results are consistent with the fact that TEMPO oxidation and other post treatments do not affect the crystalline part of the samples [72,73].
AgNP nucleation on BCNs with various surface modifications. The nucleation of AgNP was investigated on the four BCN types of interest (i.e., native BCNs, TBCNs, ABCNs and HBCNs).
Firstly, AgNO 3 aqueous solution was added to the BCN suspension and NaBH 4 aqueous solution was then introduced to reduce Ag + into AgNPs. As soon as the silver precursor was reduced, the initial translucent aqueous suspension turned light or dark yellow (inset, Fig. 3a), giving different shades of color, probably because of the initial dispersion state of the type of BCN used. However, such a color variation was not reflected in the UV-Vis spectra (Fig. 3a) since the effective AgNP content was very similar in all the hybrids (i.e., 7.9 wt% AgNP in BCN/AgNP; 6.8 wt% AgNP in TBCN/AgNP; 8.5 wt% AgNP in ABCN/AgNP; 8.6 wt% AgNP in HBCN/AgNP) with a same dominant in-plane absorption peak at ~ 400 nm. It confirmed the synthesis of well-dispersed AgNPs derived from λ max the coalescence of monomeric Ag particles obtained by a reduction to a zero-valence Ag atom [20].
Even if the surface modification did not seem to affect the intensity of the spectra (i.e., similar AgNP content for all the hybrids), the full width of the main peak and the background intensity increased for Complete size distributions are reported in Fig. S1. The nucleation growing method resulted in wellgrafted AgNPs on the cellulose surface for the neutral and modified CNCs. In contrast, AgNPs synthesized in the same conditions without BCN rapidly aggregated. This proved that BCNs form a perfect substrate for the nucleation point to obtain well-dispersed quite-monodispersed AgNPs without the need for any other capping agents or stabilizers [74].
Nevertheless, these results suggested that AgNP nucleation on the cellulose nanocrystal surface nonspecifically occurred on hydroxyl groups and/or on the additional negative surface charges. The aim was to determine if these surface charges represent an additional nucleation point that interacts with Ag + ions or if they just acted as promoters of nanocrystal dispersion, facilitating the accessibility to the hydroxyl surface groups. The TEMPO carboxylated TBCNs were then compared to aminated ABCNs.
In this case, the functionalization by the positive surface charges preserved the electrostatic repulsion and, consequently, the good dispersion. At the same time, these positive charges could not interact with Ag + ions and thus did not represent a possible nucleation point for AgNPs. The STEM image of the ABCN/AgNP hybrid (Fig. 3b) showed that AgNPs were well-nucleated on both surfaces, indicating that the -OH surface groups represented the effective AgNP nucleation point. The OH groups and Ag + ions are complexed throught ion-dipole interactions [22] and the extensive number of hydroxyl groups on the BCN surface can promote the complexation of Ag + ions also acting as passivation contacts for AgNP stabilization [75].
To our knowledge, such a result represents the first experimental proof that hydroxyl surface groups on cellulosic surfaces are the real nucleation points for metallic nanoparticles and that additional negative surface charges just improve the dispersion state, thereby increasing the accessibility to the nucleation sites. This result was corroborated by the fact that for the HBCN/AgNP hybrid, most of the AgNPs were well-grafted onto the HBCN surface as well. In this case, the only real possible nucleation point was represented by -OH surface groups since most of the carboxylate groups were removed by grafting of C 8 molecules. The presence of several AgNPs not anchored at the HBCN surface was visible in STEM images and could be linked to the suboptimal AgNO 3 dissolution due to the presence of EtOH in the suspending medium.
Concerning the structural characterization, all the XRD patterns of BCN/AgNPs (Fig. 3c) [20]. For all the samples, the same average AgNP diameter of 3 nm was estimated from XRD patterns. The nucleation mechanism of AgNPs could affect their growth on the nanocrystal surface and, thus, their final characteristics. Therefore, the oxidation state of AgNPs grafted onto BCN with different surface modifications (i.e., Ag 0 and Ag + contents) was determined by a linear combination fitting (LCF) of XANES spectra of the hybrids (Fig. S2a and Fig. S2b). For each sample, the R-factor and the Chi-square values of the fits are reported in Table S1. A quantity of 92 ± 9% of Ag 0 was found for all the hybrids, showing the efficiency of the AgNP reduction and nucleation, independently of the surface treatment of the initial BCNs. All Fourier transform spectra of BCN/AgNP ( Fig. S2c and Fig. S2d) were fitted with the crystallographic structure of metallic silver with an R-factor systematically lower than 0.015. The shifts in R values (i.e., interatomic distance) obtained from the fits were systematically negligible (< 0.06 Å -1 ; all of the values are presented in Table S2), showing that the interatomic distances in BCN/AgNP case did not significantly change in comparison to the metallic silver distances and that the space group of BCN/AgNP sample still corresponded to the fcc silver structure, as suggested by XRD. It follows that the initial and final crystal structural organization of AgNPs were not affected by the various surface-modification treatments. These results showed the NaBH 4 reduction with -OH surface groups as nucleation points represented an efficient nucleation mechanism where the speciation and the crystalline structure of AgNPs were not affected by the surface modifications performed on cellulose nanocrystals. isolated or even attached to surface-carboxylated cellulose nanofibrils [20]. In this case, we proposed a more detailed study on the effect of the H 2 O 2 redox post-treatment on properties of AgNPs (i.e., This was already visible at  equal to 0.17, since the absorption peak became broader and shifted to 495 nm, showing the presence of two shoulders at 345 nm and 385 nm related to an early modification of primary AgNPs. At  = 0.25, the in-plane plasmon peak was further shifted to 638 nm and finally moved out of the measurement window for higher  values. In this case, a lower intensity peak was detected at = 335 nm that is usually associated with an out-of-plane quadrupole resonance peak, λ max thus representing a good indicator for general prismatic architectures since it strongly depends on aspect ratio [48]. Such a peak became sharper for  values of 0.33 and 0.52 (i.e., 160 μL and 250 μL of H 2 O 2 , respectively). The presence of a low-intensity peak at 335 nm, at the same time as the shift of the in-plane resonance peak out of the measurement window, and the absence of the primary AgNP peak is proof that an AgNP size-shape variation is induced, as already observed in literature [20]. To better describe the impact of the H 2 O 2 redox post-treatment on the AgNP morphology STEM acquisition were performed (Fig. 4b). The analysis of STEM images indicated that AgNPs initially had an average diameter of 17.5 ± 12 nm that regularly increased to reach 94.1 ± 69 nm for  equal to 0.52, where several AgNPs_H 2 O 2 with quite irregular edges and a sometimes vaguely triangular shape could be observed. These results agreed with the experimental evidence reported in a recent study of our group [76], in which we investigated the effect of the H 2 O 2 redox treatment in hybrid suspensions where primary 10-nm AgNPs are nucleated on wood cellulose NCs, proving the efficient size-shape transition from 10-nm spherical AgNPs into 300 nm AgNPrisms. In this work, we proposed a H 2 O 2 redox mechanism where H 2 O 2 induced the oxidative dissolution of primary AgNPs, generating Ag + ions. When  equal to or greater than 0.20, the H 2 O 2 oxidation interest most of the AgNPs, except the contact location where AgNPs are effectively grafted onto the NC surface which could actually work as nucleation sites for the formation of newly formed AgNPs with a triangular shape. Finally, all the images clearly showed that AgNPs_H 2 O 2 , even the bigger ones, were still attached to the TBCN surface, exactly as for the reference sample before the H 2 O 2 post-treatment, showing the importance of cellulose as a support for AgNPs_H 2 O 2 nucleation, growth and stabilization. The formation of nonperfectly prismatic silver objects and the increase in the average diameter associated with an increase in the polydispersity in size (Fig. S3a and Fig. S3b, Table S3) could be linked to the absence of any additional stabilizer, which implied that the structural change was heavily affected by the aggregation of newly-formed AgNPs_H 2 O 2 [46] .
To establish how the H 2 O 2 post-treatment affected the oxidation state of AgNPs, XANES spectra were recorded and then fitted using the LCF procedure ( Fig. S4a and Table S4). Data showed that the amount of Ag 0 in AgNPs_H 2 O 2 slightly decreased with respect to the reference case, from 88% before H 2 O 2 post-treatment down to 69% at  = 0.08, as long as the H 2 O 2 addition did not induce a size-shape variation (corresponding to  from 0.08 to 0.25), while remaining equal to 91% and 87% for  values of 0.33 and 0.52, respectively (Table S4). The formation of AgNPs_H 2 O 2 mostly composed of Ag 0 silver was also corroborated by the presence in the XRD patterns of characteristic (111), (200) and (220) Ag 0 peaks, confirming the presence of a lattice plane of a face-centered cubic structure (Fig.   S4b). It could be observed that the H 2 O 2 post-treatment did not induce a structural modification of AgNPs_H 2 O 2 with respect to the primary AgNPs before the H 2 O 2 treatment and that a H 2 O 2 /AgNP mass ratio equal to 0.33 and 0.52 determined an increase in the intensity of the characteristic fcc Ag (111) peak, which corresponded to the highest recorded values of Ag 0 . Such a result suggested that a  parameter at least equal to 0.33 represented the optimum value necessary to reach an effective sizeshape variation that maintains the conversion of the initial Ag + to Ag 0 ratio. Finally, the EXAFS Fourier transform spectra (Fig. S4c and Fig. S4d) of the AgNP_H 2 O 2 in hybrids were fitted with the crystallographic structure of metallic silver, (R-factor systematically lower than 0.020), presenting systematically negligible shifts in R space (< 0.06 Å -1 ; all values are in Table S5). This result confirmed that the interatomic distances in AgNP_H 2 O 2 did not significantly change in comparison to the metallic silver distances and that the space group of the samples still corresponded to the fcc silver structure, as observed by XRD. The final crystal structural organization was not affected by the H 2 O 2 redox post-reaction nor by the particle size variation, as observed by Ma et al. [77] Conversely, the AgNP diameter measured by XRD was about 10 nm, which was consistently smaller than the 40-80 nm measured by STEM. Such a difference could be explained by the fact that XRD makes it possible to evaluate the size of the coherent diffraction domains in AgNPs, i.e., perfect repetition in singlecrystal particles, twinned or imperfect particles that have more than a single diffraction domain [77].
The characteristics of all the investigated samples are reported in Table 2.
All these results, coupled with the experimental proof that -OH groups on the CNC surface are the AgNP nucleation points, allowed us to conclude that our approach to tune the morphology and the oxidation state of the AgNP grafted onto cellulosic support may be extended to all the hybrid systems where metallic NPs are anchored on a polysaccharide substrate (e.g., cellulose fibers, films or chitin NCs).   (Fig. 5a). This suspension showed an in-plane absorption peak shifted to higher (i.e., out of the measurement window), and a sharp out-of-plane quadrupole resonance peak was λ max visible at 333 nm (Fig. 5b). This result agreed with the transformation of the primary AgNPs of 16.1 ± 13 nm into bigger AgNPs_H 2 O 2. of 95.4 ± 87 nm, as confirmed by STEM (Fig. 5c). On the other hand, the addition of H 2 O 2 to the BCN/AgNP turned the suspension a darker gray color, and the particle size slightly changed with respect to the reference case (i.e., 32.5 ± 19 nm). Moreover, the in the UV-λ max Vis spectra remained at 402 nm and a poorly-defined low-intensity broad peak appeared at 340 nm.
These results indicated an uncomplete transformation of the AgNPs, suggesting that the BCN surface modification affected the H 2 O 2 transformation of primary AgNPs since the BCN surface modification influenced the BCN dispersion rather than its chemical structure.
When the surface charges provided a good repulsion between nanocrystals (the cases of TBCN and ABCN), the addition of H 2 O 2 made it possible to obtain the formation of AgNPs_H 2 O 2 of about 100 nm, whereas the size of AgNPs_H 2 O 2 for the almost neutral surface charge in the BCN case remained constant (i.e., about 20 nm). It appeared that well-dispersed nanocrystals facilitate the accessibility to the surface groups, thereby promoting the H 2 O 2 redox action.
Concerning the hydrophobically-modified surface (i.e., HBCN/AgNP case), the introduction of H 2 O 2 did not induce a size-shape modification of primary AgNPs, as shown in the STEM image (Fig. 5c).
Furthermore, the UV-Vis spectrum of the HBCN/AgNP_H 2 O 2 suspension did not change with respect to the reference HBCN/AgNP. In this case, not only the absence of repulsion prevented good nanocrystal dispersion, but the suspending medium containing EtOH could promote aggregation and affect the efficiency of the H 2 O 2 redox reaction as well. To check the influence of EtOH on the H 2 O 2 treatment, we compared aqueous TBCN/AgNPs suspension (highly charged nanocrystals to ensure dispersion) to the same TBCN/AgNPs diluted in a H 2 O/EtOH (60/40 v/v) mixture, like in the case of the HBCN/AgNPs. In pure water, the addition of H 2 O 2 induced a color variation from yellow to blue, whereas the hybrid prepared in the H 2 O/EtOH mixture just turned a dark yellow (Fig. S5). This result confirmed that the presence of ethanol limited the formation of AgNPs_H 2 O 2 that could be linked to a weak oxidation of AgNPs to Ag + and the subsequent reduction in a non-aqueous medium. Size distributions of AgNPs_H 2 O 2 in BCN hybrids are reported in Fig. S6 and the average diameters are summarized in Table S6.
Interestingly, all the reformed AgNPs_H 2 O 2 appeared to be nucleated on nanocrystal surfaces, irrespective of the applied surface modification. This means that the BCN still serves as a substrate for AgNP_H 2 O 2 formation, ensuring adequate AgNP dispersion. We hypothesized that the H 2 O 2 did not completely oxidize the AgNPs, leaving hooks available as a large amount of re-nucleation points for AgNP_H 2 O 2 nanoparticles in addition to the -OH surface groups, as reported in another work of our group [76]. Moreover, the analysis of XANES spectra (Fig. S7a, Table S7) showed that the addition of H 2 O 2 did not affect the Ag 0 content (i.e., 95 ± 10% compared to 92 ± 9% of the reference cases), indicating that the oxidation state did not change. Finally, the XRD patterns of hybrid suspensions without ( = 0) and with the addition of H 2 O 2 ( = 0.33), Fig. 5d., indicating the presence of the Ag (111), Ag (200) and Ag (220) peaks, characteristic of the fcc silver model. This confirmed the formation of metallic AgNPs_H 2 O 2 well-grafted onto the BCN substrate, irrespective of the surface modification. Moreover, the fitting of EXAFS Fourier transform spectra (R-factor < 0.016) did not reveal any significant modification of the crystallographic structure in comparison to fcc metallic silver ( Fig. S8a and Fig.  S8b, data in Table S8). These studies confirmed that all native and modified BCN/AgNP samples were composed quasi-exclusively of Ag 0 with similar crystallographic organization.

Conclusion
In this study, we investigated the interactions involving well-dispersed AgNPs grafted on a bio-based