Surface Functionality Influences pH- and Solvent-Dependent Colloidal Stability of Cellulose Nanocrystals

Abstract

Cellulose nanocrystals (CNCs) are promising sustainable nanomaterials for a wide range of applications. Yet, limitations remain in understanding their dispersion behavior in organic solvents and across pH conditions, especially after surface modification. In this work, CNCs were surface-functionalized with carboxy, amino, or vinyl groups and thoroughly characterized. Dispersion behavior was evaluated using dynamic light scattering, including measurements of hydrodynamic diameter and ζ-potential in water solutions at various pH levels and across different solvents. Colloidal stability strongly depends on surface functionality and solvent characteristics. A correlation with Kamlet-Abboud-Taft parameters was identified, linking dispersion behavior to solvent polarity and hydrogen-bonding ability, in relation to the pKa of the surface groups. These findings provide a semiquantitative framework for designing novel materials with CNC dispersions across diverse media.

1. Introduction

Cellulose nanocrystals (CNCs) have emerged as promising sustainable nanomaterials because of their high surface area, mechanical strength, and surface tunability. These nanomaterials are commonly obtained through the acid hydrolysis of biomass, resulting in materials with an abundance of hydroxy (-OH) groups and half-esters from the acid used. The functional groups added provide well-dispersed colloidal behavior in aqueous systems. Because of these colloidal properties, their use has expanded across biomedical, coating, packaging, and advanced material applications, where control over dispersion is critical for their performance. However, CNC dispersion in nonaqueous media remains difficult to predict because solvent polarity alone does not capture solvent–surface interactions.

Moreover, surface modification is needed to tailor key properties such as surface charge and hydrophobicity and to improve compatibility, particularly in nonpolar or hydrophobic matrices. The challenge here comes two-fold: to selectively modify the surface to achieve the desired interactions (preserving their original morphology, avoiding polymorphic transitions, maintaining crystal integrity) and improve their dispersibility. When functional groups are added, these will change the relationship between the material and solvent, requiring new predictive approaches that integrate both surface chemistry and solvent properties, which remain limited.

For aqueous solutions, the relationship between pH and the pKa of the introduced functional groups can be used as a design parameter to generate this predictivity. The pH of the water plays a critical role in modulating the ionization state of surface functional groups, directly affecting electrostatic interactions and colloidal stability. Conversely, for organic solvents, the Kamlet-Abboud-Taft parameters are a solvent classification system based on their dipolarity/polarizability (π*), which measures the ability of the solvent to stabilize a charge or a dipole through its dielectric effect, and on their basicity (β), which measures the solvent’s capacity to accept a proton in solute-to-solvent hydrogen bonding. These parameters provide a useful framework to describe solvent–solute interactions. For CNCs, these interactions are particularly relevant, as their colloidal behavior is governed by both surface functionality and the surrounding medium. This work aims to provide a comprehensive comparison of CNC dispersions with different functional groups in different media to understand how surface modifications influence CNC dispersion across different pH levels in aqueous media and in solvents of varying polarity.

2. Experimental Section

2.1. Materials and Reagents

CNCs were purchased from CelluForce Inc. (Montreal, Quebec, Canada), which produced them from hardwood via sulfuric acid hydrolysis. Glacial acetic acid (ACS grade) was obtained from VWR Chemicals (Radnor, PA, USA). Ethanol (99.5%) and acetone (99.5%) were purchased from Thermo Fisher Scientific (USA). Methanol (ACS reagent grade) was obtained from Fisher Chemical (Pittsburgh, PA, USA). Triethoxyvinylsilane (TEVS, 97%) was purchased from Thermo Fisher Scientific (manufactured in China). (3-Aminopropyl)­triethoxysilane (APTES, ≥98%) and 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO, 98%) were obtained from Sigma-Aldrich (St. Louis, MO, USA; both manufactured in China). Sodium hypochlorite solution (NaClO, 10–15%) was also purchased from Sigma-Aldrich. Standard solutions of sodium hydroxide (NaOH, 0.1 N volumetric solution) were obtained from J.T. Baker MACRON Fine Chemicals (Radnor, PA, USA). Hydrochloric acid (HCl, 0.1 N volumetric solution) was purchased from MACRON Fine Chemicals (Radnor, PA, USA). Polyvinyl sulfuric acid potassium salt (0.001 N, PVSK) and polydimethyl diallyl ammonium chloride (0.001 N, pDADMAC) were bought from BTG (Eclépens, Switzerland). Dimethyl sulfoxide (DMSO), dimethyl carbonate (DMC), trichloromethane (CHCl₃), N,N-diethyl ethanamine (TEA), N,N-dibutylbutan-1-amine (NBu₃), and n-heptane were purchased from Thermo Fisher Scientific (USA).

2.2. Surface Modification of CNCs

2.2.1. Oxidation Mediated by 2,2,6,6-Tetramethylpiperidine-1-oxyl (TEMPO)

Carboxy-rich CNCs were obtained through TEMPO-mediated oxidation, using a sodium bromide (NaBr) and sodium hypochlorite (NaClO) system. , For this, 1 g was dispersed in 100 mL of water and sonicated for 5 min at 30% amplitude to achieve a uniform suspension. TEMPO (16 mg) and NaBr (0.1 g) were added, and the reaction was initiated by dropwise addition of NaClO under constant stirring and a pH of 10. After completion, excess oxidant was quenched with methanol, and the suspension was neutralized and dialyzed (MWCO 12–14 kDa) against DI water for ≥5 days.

2.2.2. Silanization with Triethoxyvinylsilane (TEVS)

CNCs were functionalized with triethoxyvinylsilane (TEVS). TEVS at 5 wt % was hydrolyzed in an ethanol:water (1.5:1 w/w) solution at pH 4 for 2 h at room temperature. CNCs (0.02 g/mL ethanol) were added and reacted at 48 °C for 2.5 h, followed by heating at 75 °C for 15 min and curing at 110 °C for 30 min. The modified CNCs were redispersed and centrifuged two times (5000 g × 30 min) with acetone to remove unreacted silane, then washed with ultrapure water for further analysis.

2.2.3. Silanization with 3-(Triethoxysilyl)­propan-1-amine (APTES)

CNC modification with APTES was performed in a 80/20 wt % water:ethanol solution adjusted to pH 4.0 with glacial acetic acid. To this solution, APTES at 0.5 wt % was added, and the mixture was stirred for 1 h at room temperature to promote silane hydrolysis. To minimize cross-linking and aggregation, CNCs were added at a ratio of 1.8 g of CNC per g of APTES. The mixture was ultrasonicated for 4 min at 30% amplitude, with a 5:2 s on–off cycle, and subsequently stirred for 2 h at room temperature to allow sufficient silane–CNC interactions.

2.3. Characterization

2.3.1. Atomic Force Microscopy (AFM)

AFM was performed to analyze the morphology, dispersion, and size of the modified CNCs (mCNCs). Substrates were P-type Test Grade silicon wafers, cleaned twice with ethanol and ultrapure water to eliminate dust particles. Subsequently, the sensors were subjected to UV-ozone treatment for 30 min using a Novascan PSD (Milwaukee, WI, USA). Sensor surfaces were activated by immersion in a 0.1 wt % polyethylenimine solution (PEI) for 20 min, followed by rinsing in ultrapure (UP) water for 5 min. Sensors were coated with 200 μL of a 0.01 wt % solution of each sample using a Laurell H6-23 spin coater at 4500 rpm (Lansdale, PA, USA).

The analyses were performed using the Asylum Research MFP-3D-SA Atomic Force Microscope (Goleta, California) along with an Asylum Research cantilever model AC160TS. Images were obtained using Igor Pro 6.37 in AC Air Topography mode. The size of the scanning areas was 2 × 2 μm, capturing the height parameter. Image treatment was enabled using Gwyddion 2.61.

2.3.2. Water Contact Angle (WCA)

The hydrophilicity of the mCNC and unmodified control CNC samples was evaluated using an FTA 1000 Analyzer System Contact Angle Goniometer (First Ten Ångstroms, USA) coupled with a digital camera (Watec Auto Iris) equipped with a Navitar 50 mm lens (f/1.4), at 25 °C. The WCA was measured using the sessile-drop method with 1 μL water drops onto the previously coated surface. The WCA was the average of 10 measurements at different spots on each surface.

2.3.3. Fourier Transform Infrared with Attenuated Total Reflection (FTIR-ATR)

To confirm functionalization, the different CNCs were characterized through FTIR-ATR using a Bruker ALPHA II FTIR system equipped with a diamond crystal (Billerica, MA, USA). Analyses were performed using 10–20 mg of the previously freeze-dried samples. The spectra were recorded per triplicate in the spectral range of 4000–400 cm^–1 with 4 cm^–1 resolution and 128 scans for each sample. FTIR data were baseline-corrected and normalized in software OPUS 9.0.6, and the spectra were plotted in GraphPad Prism (version 10.6.1; GraphPad Software, San Diego, CA, USA).

2.3.4. Thermogravimetric Analysis (TGA)

Variations in degradation profiles and thermal stability were used to confirm surface functionalization and as indirect indicators of the degree of substitution. The modified samples and the unmodified control were previously freeze-dried for 24 h at −80 °C. TGA analyses were performed in the TGA 550 thermogravimetric analyzer (TA Instruments, DE, USA). Approximately 5 mg of the sample was placed in ceramic pans, and measurements were conducted over the temperature range 35–699 °C at a heating rate of 10 °C/min. The experiments were carried out under a nitrogen atmosphere at a flow rate of 20 mL/min per triplicate. Data was analyzed in TA Instruments TRIOS 5.4 (USA). Thermal degradation temperatures Td’s were measured based on 5% mass loss after the initial 120 °C hold.

2.3.5. Polyelectrolyte Titration

Total charge density was determined with the CHEMTRAC Laboratory Charge Analyzer LCA5 (Norcross, GA, USA). Samples were prepared at 0.04 wt % and adjusted to pH 6.5. According to the initial charge, 25 mL of either poly­(diallyldimethylammonium chloride) (p-DADMAC, 0.001 N) or potassium poly­(vinyl sulfate) (PVSK, 0.001 N) was added to 15 mL of the dispersion. The suspensions were mixed and centrifuged to eliminate large aggregates. Postcentrifugation, 10 mL of supernatant was titrated with the opposite charged polymer using the charge analyzer. All tests were conducted in triplicate. Charge density was calculated from the volume required to reach a streaming current value (SCV) of 0. Titration curve data were extracted from CHEMTRAC images using WebPlotDigitizer and calculated with the following equation:

$$Chargedensity=\frac{([pDADMAC]x{V}_{p‐DADMAC})-([PVSK])x{V}_{PVSK})}{W_{drysample}}$$

where

[pDADMAC] is the concentration of the cationic polymer,

V_p‑DADMAC is the used volume of p-DADMAC,

[PVSK] is the concentration of the anionic titrant,

V_PVSK is the consumed volume of the titrant, and

W_dry sample is the weight of the dry sample.

2.3.6. Conductometric Titrations

Conductometric titrations were conducted by measuring both pH and conductivity using a Mettler Toledo Seven Direct SD23 pH and conductivity meter (Columbus, Ohio, USA), which was previously calibrated with the respective buffer standard solutions. The conductivity values were adjusted to account for the dilution effect caused by the gradual addition of NaOH during titration. A volume correction was applied to normalize the conductivity measurements to the original suspension volume, ensuring that the observed changes solely reflect the neutralization of functional groups on the material surface. This correction was performed using the following equation:

$$Conductivit{y}_c=Conductivit{y}_m\times \frac{V_i+V_o}{V_o}$$

where

Conductivity_c is the corrected conductivity (μS cm^–1),

Conductivity_m is the measured conductivity for each data point (μS cm^–1),

V_i is the initial suspension volume (mL), and

V_o is the added volume of NaOH at each point (mL).

2.3.6.1. Sulfate Half-Ester Titration

Sulfate half-ester groups on the CNC surface, introduced during hydrolysis, were quantified through conductometric titration. CNC samples were dialyzed for at least 5 days (using Spectra/Por 4 membranes with a 12–14 kDa MWCO) until their conductivity stabilized, removing any residual sulfate ions. About 120 mg of the dialyzed sample was diluted to 100 mL with ultrapure water, then protonated by adding 1 mL of 0.05 M HCl to achieve a pH of around 3.5, along with 300 μL of 0.1 M NaCl to increase conductivity. Titration involved incremental additions of 0.05 M NaOH in 100 μL steps. The titration curves showed an initial decrease in conductivity, corresponding to proton neutralization from sulfate ester groups, followed by an increase once all charged groups were neutralized, and excess hydroxide ions appeared. The equivalence point, indicated by an inflection in the curve, was used to determine the volume of NaOH at a known concentration needed to achieve neutrality.

$$C_{R-OS{O}_{3}H}\left(\frac{\mathrm{m}\mathrm{m}\mathrm{o}\mathrm{l}}{\mathrm{g}}\right)=\frac{C_{NaOH}\times (V_2-V_1)}{m_{sample}}\times 1000$$

where

$C_{R‐OS{O}_{3}H}$ is the sulfate half-ester group content (mmol/g),

C_NaOH is the concentration of the NaOH titrant (mol/L),

V ₁ is the volume of NaOH required to neutralize excess HCl (L),

V ₂ is the total volume required to neutralize both HCl and sulfate half-ester groups (L), and

m_sample is the dry mass of the sample used for titration (g).

2.3.6.2. Carboxyl Titration

To ensure the successful functionalization of the CNCs, especially for the ones modified via TEMPO oxidation, the quantification of carboxyl groups (−COOH) was performed. 20 mg of each sample was redispersed in 10 mL of 0.01 M HCl. After stirring for 10 min, the suspensions were titrated with 0.01 M aqueous NaOH in 500 μL increments, gradually approaching pH 9–11. Conductivity was measured and corrected after each NaOH addition. The titration was carried out across a pH range of 2–12. The resulting conductometric curve indicated the presence of a strong acid, due to excess HCl, and a weak acid, attributable to the carboxyl groups.

$$C_{COOH}(\mathrm{m}\mathrm{m}\mathrm{o}\mathrm{l}\mathrm{g})=C_{NaOH}\times (V_2-V_1)m_{sample}\times 1000$$

$$C_{COOH}\left(\frac{\mathrm{m}\mathrm{m}\mathrm{o}\mathrm{l}}{\mathrm{g}}\right)=\frac{C_{NaOH}\times (V_2-V_1)}{m_{sample}}\times 1000$$

where

C_COOH is the carboxyl group content (mmol/g),

C_NaOH is the concentration of the NaOH titrant (mol/L),

V ₂ – V ₁ is the volume of NaOH required to neutralize the carboxyl groups, determined from the conductivity minimum (inflection point) (L), and

m_sample is the dry mass of the sample used for titration (g).

2.3.6.3. Amino Titration

The quantification of amino groups (−NH₂) was performed by resuspending 50 mg of the sample in 10 mL of HCl (0.05 M) and stirring for 10 min. The mixture was subsequently titrated with 0.05 M NaOH, adding 0.5 to 1 mL increments until reaching the equivalence point. The titration curve’s inflection point was identified, and the degree of functionalization was calculated from it.

2.3.7. Dynamic Light Scattering (DLS)

Their colloidal behavior (hydrodynamic diameter and ζ-potential) was evaluated using 0.01 wt % aqueous dispersions prepared with ultrapure water as the dispersion medium, which were previously adjusted to pH 6.5 by the dropwise addition of dilute NaOH or HCl solutions.

Likewise, samples analyzed in organic solvents were prepared at a final concentration of 0.01 wt % to keep comparisons consistent and to minimize multiple scattering and interparticle interactions. Before analysis, the never-dried CNCs were transferred from water to the corresponding organic solvent through four consecutive solvent-exchange cycles. During each cycle, the dispersions were centrifuged at 14,000 × g for 10 min, the supernatant was removed, and the pellet was resuspended in fresh solvent. After the final solvent exchange, the samples were sonicated for 2 min at 30% amplitude using a pulsed cycle of 5 s on and 10 s off. Measurements were performed in triplicate using a Malvern ZETASIZER PRO at 25 °C with the dip cell cuvette ZEN1002, letting the system stabilize for 60 s before each measurement. Parameters such as the ζ-potential model (Smoluchowski or Hückel), solvent-specific viscosity, and the dielectric constants based on each solvent were adjusted in the software ZS XPLORER 4.2.0.181 (Table S1).

The hydrodynamic diameters are reported as the intensity of the ζ-average. Since CNCs are rod-shaped but dynamic light scattering assumes spherical particles, these values serve as an apparent, sphere-equivalent size. Therefore, it was used mainly as a comparative parameter of colloidal stability rather than as an exact measure of the true crystal dimensions.

2.3.8. Statistical Analyses

The statistical analyses and graphing were performed using GraphPad Prism (version 10.6.1; GraphPad Software, San Diego, CA, USA). Significant differences among modifications were analyzed using one-way analysis of variance (ANOVA) with Tukey’s multiple-comparison test (p < 0.05). The relationships between solvent polarity, solvent basicity, or pH and the hydrodynamic diameter and ζ-potential were evaluated separately for each CNC modification using simple linear regression, assessed using the coefficient of determination (R ²), and regression slopes considered significantly different from zero at p < 0.05.

3. Results and Discussion

The first inferences were observed through morphology comparison (Figure ). Overall, surface functionalization did not significantly alter the morphology of the CNCs, as all samples retained the characteristic rod-like structure, were in the nanoscale range, and were monodisperse. Since DLS models particles as spheres, the hydrodynamic diameters reported below are sphere-equivalent values used as a relative, comparative metric of the dispersion state rather than as absolute particle sizes.

1.

Surface functionalization of CNCs and its effect on morphology and wettability. (A) Schematic representation of −OH (control), −COOH (TEMPO), −NH₂ (APTES), and CH₂ (TEVS) groups. (B) Atomic Force Microscopy (AFM) height images showing preserved rod-like morphology after modification. (C) Water Contact Angle (WCA) measurements, indicating increased hydrophobicity from −OH/–NH₂ to CH₂, while −COOH could not be quantified due to complete droplet spreading. Values are reported as mean ± SD; different letters indicate significant differences (p < 0.05).

Despite similar morphology, surface functionalization significantly affected hydrophilicity, as evidenced by WCA measurements. Unmodified CNC­(−OH) exhibited a low contact angle (12.52 ± 4.04°), confirming its inherent hydrophilic nature due to the abundance of surface hydroxy groups. Similarly, CNC­(−NH₂) samples showed no significant difference (13.68 ± 3.55°, p > 0.05), indicating that amino functionalities maintain strong interactions with water, likely through hydrogen bonding due to the polarity of this functional group. In contrast, oxidized CNC­(−COOH) exhibited complete spreading of the water droplet, precluding reliable quantification. This behavior is attributed to the high surface polarity and potential ionization of the surface carboxyl groups. Conversely, vinyl-functionalized CNC­(CH₂) exhibited a higher contact angle (66.33 ± 2.48°), indicating a reduction of the hydrophilicity of the surface, behavior attributed to the introduction of nonpolar hydrocarbon groups, reducing surface energy and hydrogen bonding. These changes in hydrophilicity are expected to directly influence dispersion behavior, as increased hydrophilicity enhances colloidal stability in aqueous media, while hydrophobic modifications may favor interfacial interactions and, therefore, dispersion with organic solvents.

The chemical functional groups were evaluated by FTIR spectra where the characteristic bands corresponding to each modification can be identified (Figure A). Distinctive cellulose bands were present in all samples. The most prominent band occurred at 3300–3400 cm^–1, corresponding to intramolecular O–H stretching vibrations. The C–H stretching vibration of the aliphatic CH₂ group appeared at 2890 cm^–1, and bands at 1170–1050 cm^–1 were associated with C–O–C bonds from glycosidic linkages. Oxidized CNCs exhibited a distinct band near 1745 cm^–1, associated with carboxyl (−COO^–) groups, confirming oxidation. CNCs with amino groups added via APTES showed slight broadening in the ∼3300 cm^–1 region, attributed to overlapping O–H and N–H stretching vibrations. Additionally, a weak band around ∼1545 cm^–1 was observed, which is associated with amino groups scissoring vibration. An increase in intensity in the 1000–1100 cm^–1 region indicates the formation of Si–O–C bonds after modification. , TEVS-silanized CNCs displayed bands at 1600 and 1400 cm^–1, attributed to the stretching and in-plane deformation vibrations of the vinyl groups (Si-CHCH₂). Additionally, bands around 780 cm^–1 are associated with silane carbon bonds (Si–C), confirming the addition of the functionality.

2.

(A) ATR-FTIR spectra comparison of mCNCs confirming the incorporation of different functional groups on CNC surfaces. (B) TGA and DTG curves of mCNCs, along with the extracted thermal parameters: T5% (temperature at 5% weight loss), Tmax (temperature at maximum degradation rate), and Ychar (char yield at 580 °C).

These chemical modifications also resulted in changes in thermal behavior (Figure B). Differences in the maximum degradation rate (Tmax) and the residue yield (Ychar) were observed depending on the introduced functional groups. Vinyl CNC­(CH₂) exhibited the highest thermal stability (Tmax 316.65 ± 8.7 °C) and the largest char yield (60.87 ± 2.3%). These could be a result of the incorporation of Si–O–Si linkages, which enhances thermal stability by forming a densely bonded polysiloxane network that promotes char formation. This effect is further supported by increased surface hydrophobicity, which effectively limits heat and mass transfer during degradation. ,

Oxidized CNC­(−COOH) showed an earlier onset of degradation (T5% ≈53 °C), attributed to the presence of carboxy groups, which are known to promote thermal decomposition because decarboxylation reactions accelerate aromatization and the formation of carbonyl groups. Previous studies show that the thermal stability of oxidized CNCs is also influenced by the surface charge content, which correlates with the values obtained in Table . In contrast, APTES-modified CNCs displayed the lowest Tmax (254.90 ± 1.7 °C), indicating reduced thermal stability compared to the unmodified CNC, despite the presence of Si at the surface. This phenomenon is attributed to an increase in surface charge due to the introduction of amino groups, which increase moisture content and catalyze thermal degradation.

1. Surface Charge Density and Functional Group Content of the Modified CNCs .

	–OH	–COOH	–NH2	CH2
Charge density (μeq/g)	972.87 ± 11.1b	5716.07 ± 10.0a	946.20 ± 17.9b	460. 47 ± 11.1c
Sulfate ester (mmol/g)	0.17 ± 0.0a	Nonquantifiable	0.02 ± 0.0c	0.04 ± 0.0b
Carboxy (mmol/g)	0.30 + 0.0d	1.95 ± 0.0a	1.09 ± 0.0b	0.37 ± 0.0c
Amino (mmol/g)	0.00 ± 0.0b	0.00 ± 0.0b	1.25 ± 0.0a	0.00 ± 0.0b

ⁱDifferent lowercase letters (a, b, and c) indicate statistically significant differences among the modified CNCs (p < 0.05).

Overall, the results indicate that thermal stability is strongly influenced by the nature of the surface functional groups, with polar and ionizable functionalization promoting earlier degradation, whereas hydrophobic modifications enhance thermal resistance and char formation.

Quantitative analysis of functionalization was then done by titrations (Table ), where the polyelectrolyte titration provided an overall value of charge density. Specific quantification of the sulfate ester, carboxy, and amino groups was measured by conductometric titrations.

The introduction of carboxy groups to the CNC resulted in a significant increase in surface charge density (5716.07 ± 10.0 μeq/g), which is significantly greater than CNC control (−OH) (972.87 ± 11.1 μeq/g). This increase is directly correlated with the high carboxy content (1.95 ± 0.0 mmol/g), confirming successful oxidation and the presence of negative charges in the system. CNC­(−NH₂) exhibited an amino group content of 1.60 ± 0.0 mmol/g; however, the overall charge density (946.20 ± 17.9 μeq/g) remained comparable to that of unmodified CNCs. This suggests that, under the measurement conditions (pH 6.5), the amino groups are not fully protonated, resulting in a lower net positive charge than the highly ionized carboxy groups in oxidized CNCs. Meanwhile, vinyl-functionalized CNCs (CH₂) showed the lowest charge density (460.47 ± 11.1 μeq/g) due to the predominance of nonpolar groups. However, the carboxy group content also increased, suggesting that during the modification, the conditions allowed slight oxidation of the −OH groups.

Lastly, the sulfate half-ester groups introduced during acid hydrolysis were present in all samples; the control retained 0.17 mmol/g, within the range previously reported by Jordan and collaborators for sulfuric acid CNCs (0.13–0.35 mmol/g). With their low pKa (∼1–2), these groups stay negatively charged across the whole pH range, adding a constant negative charge that largely explains the negative ζ-potential and their intrinsic colloidal stability. − These functional groups weakly contributed to the charged CNC­(−OH) and CNC­(CH₂), reinforcing the CNC­(−COOH) charge and partially offsetting the protonated CNC­(−NH₂). For the oxidized CNC­(−COOH), sulfate was nonquantifiable because its small conductometric signal is masked by the far larger carboxyl charge, compounded by partial desulfation during the alkaline TEMPO oxidation and dialysis. ,,

After confirming and quantifying the changes added with the modifications, the colloidal behavior was measured at different pH values. As expected, colloidal behavior was strongly dependent on the interplay between pH and the pKa of the functional group added. The pKa dictates the degree of ionization and, consequently, the surface charge and influences the dispersion state. In general, higher surface charge and ionic strength contribute to enhanced colloidal stability. As illustrated in Figure A, the pKa values provide a chemical framework for interpreting trends in the ζ-potential and hydrodynamic diameter. The shaded band in Figure C highlights the region of |ζ-potential|<30 mV, which is conventionally associated with poor colloidal stability, since electrostatic repulsion in this range is insufficient to prevent particle aggregation.

3.

Effect of surface functionalization at different pH levels. (A) Representation of the pKa values associated with each functional group added in the CNC modifications, (B) hydrodynamic diameter of the mCNC at different pH conditions, and (C) ζ-potential of mCNC at different pH levels. The shaded region denotes the |ζ| < ∼30 mV zone associated with limited colloidal stability.

Unmodified CNC­(−OH), characterized by a high pKa (9.5–13), showed limited ionization within the studied pH range. However, in acidic conditions (pH 2–5), the magnitude of the ζ-potential decreased, coinciding with an increase in the hydrodynamic diameter. This increase can be associated with the predominance of protonated species (weaker electrostatic repulsion), promoting particle–particle interactions and aggregation. At higher pH values, although ionization remains limited due to the high pKa, ζ-potential values contribute to improved dispersion, suggesting that stabilization is governed by weaker electrostatic interactions and hydrogen bonding.

Oxidized CNC­(−COOH) exhibited consistently low hydrodynamic diameters across the evaluated pH range, which correlates with their negative ζ-potential values. Given the pKa of carboxyl groups (∼4–5), deprotonation is expected from mildly acidic to neutral conditions, leading to a higher surface charge and improved electrostatic repulsion. However, at pH extremes, a decrease in ζ-potential indicates reduced colloidal stability, leading to slight increases in particle size due to limited charge at low pH and possible ionic effects at high pH. Likewise, CNC­(−NH₂), with a pKa around 10, showed a pH-dependent ionization behavior. At acidic conditions, protonation of amino groups reduced the ζ-potential, limiting electrostatic stabilization and leading to large hydrodynamic diameters. As pH increased, partial deprotonation occurred, and the ζ-potential became more negative, correlating with improved dispersion and reduced particle size. Nevertheless, the hydrodynamic diameter remained greater than that of the other systems.

CNC­(CH₂) shows variations in ζ-potential across the pH range, despite the nonionizable nature of vinyl groups. At low pH, an increase in hydrodynamic diameter is observed, suggesting aggregation. With increasing pH, changes in ζ-potential are mainly due to interactions with the surrounding medium rather than ionization. These CNCs are less affected by pH changes compared to CNCs modified with ionizable groups. This response to the pH can be an effect of the residual sulfate ester groups and the nonconsumed siloxane groups.

To determine whether the observed changes were linearly associated with pH, simple linear regression analyses were conducted for each surface modification, using pH as the independent variable and hydrodynamic diameter or ζ-potential as the dependent variables (Table ). These results showed that pH was significantly associated with the ζ-potential of CNCs­(−OH) (R ² = 0.3475, p = 0.0207), CNCs­(−NH₂) (R ² = 0.4437, p = 0.0067), and CNCs­(CH₂) (R ² = 0.4165, p = 0.0094). All three regressions had negative slopes, indicating a shift toward more negative ζ-potential values as pH increased. In contrast, no linear relationship was observed for CNCs­(−COOH) (R ² < 0.001, p = 0.9914). Hydrodynamic diameter decreased significantly with increasing pH for CNCs­(−COOH) (R ² = 0.4776, p = 0.0043), CNCs­(−OH) (R ² = 0.4662, p = 0.0050), and CNCs­(CH₂) (R ² = 0.5605, p = 0.0013). No significant relationship between pH and hydrodynamic diameter was detected for CNCs­(−NH₂) (R ² = 0.1042, p = 0.2406)

2. Linear Regression Analysis of ζ-Potential and Hydrodynamic Diameter as a Function of pH Solutions for Surface-Functionalized CNCs.

Parameter	mCNC	Linear regression in function of the pH	R 2	p	Significance
ζ-potential	–COOH	y = 0.02008x – 23.36	9.187e-006	0.9914	Not significant
	–OH	y = −3.143x – 9.390	0.3475	0.0207	Significant
	–NH2	y = −1.950x – 19.58	0.4437	0.0067	Significant
	CH2	y = −5.498x + 14.90	0.4165	0.0094	Significant
Hydrodynamic diameter	–COOH	y = −4.056x + 141.4	0.4776	0.0043	Significant
	–OH	y = −7.605x + 181.3	0.4662	0.0050	Significant
	–NH2	y = 5.036x + 274.8	0.1042	0.2406	Not significant
	CH2	y = −1307x + 11667	0.5605	0.0013	Significant

The dispersion behavior of mCNCs in organic solvents was influenced by solvent dipolarity/polarizability (π*), as shown in Figure . As a general qualitative trend, solvents with higher π* (>0.70) tended to give more stable dispersions, whereas at low π* (e.g., n-heptane), ζ-potential values near neutrality were measured, indicating poor stabilization and aggregation consistent with previous reports. In Table S2, pictures showing the dispersion appearance are shown. This trend, however, was not uniform across modifications: as shown by the regressions below, π* significantly modulated the ζ-potential of only CNC­(−NH₂) and CNC­(CH₂), and it was a weak predictor of aggregate size, so polarity alone did not control the colloidal behavior.

4.

(A) Kamlet–Abboud–Taft solvent diagram of selected solvents. (B) Hydrodynamic diameter of mCNCs in the corresponding solvent series plotted as a function of solvent dipolarity/polarizability (π*). (C) ζ-potential of mCNCs as a function of solvent dipolarity/polarizability (π*). (D) Hydrodynamic diameter of mCNCs as a function of solvent basicity (β). (E) ζ-potential of mCNCs as a function of solvent basicity (β). The shaded region denotes the |ζ| < ∼30 mV zone associated with limited colloidal stability.

Several solvents with intermediate π* values, including ethanol, chloroform, and acetone, still produced relatively large particle sizes, indicating that dipolarity/polarizability alone is insufficient to explain the observed colloidal behavior. In this context, solvent basicity (β) provided additional insight. As shown in Figure E, DMSO, which combines relatively high β (>0.76) with strong charge-solvation capability, produced more negative ζ-potential values, consistent with enhanced solvation of charged surface groups and improved electrostatic stabilization. By contrast, H₂O also yielded negative ζ-potential values, although this behavior is attributed to its high polarity and dielectric constant rather than to basicity.

Importantly, in this system, negative ζ-potential values did not always correspond to small hydrodynamic diameters. Solvents such as acetone and ethanol could still show relatively large aggregate sizes despite moderately negative potentials, suggesting that electrostatic repulsion alone was not sufficient to completely disperse the nanoparticles. This behavior indicates that additional solvent properties, including dielectric constant, basicity, and solvent–particle affinity, also contribute to colloidal stability. In contrast, H₂O and acetone, despite their low basicity, also yield negative ζ-potentials due to their high polarity and high dielectric constant, which promote effective charge dissociation. Conversely, solvents with intermediate β values exhibit ζ-potentials closer to neutrality, which correlates with the formation of large aggregates observed in DLS measurements.

Surface functionalization further modulated these trends. Highly charged CNC­(−COOH) showed the strongest dependence on solvent properties, particularly those related to charge solvation. In contrast, amine-functionalized CNCs­(−NH₂) exhibited more variable behavior, likely due to their solvent-dependent ionization state, which can reduce or destabilize electrostatic contributions in nonaqueous media. Hydrophobically modified CNC­(CH₂) showed a weaker dependence on both π* and β, suggesting that their dispersion is governed more strongly by solvent–surface affinity than by electrostatic interactions. The combined analysis of hydrodynamic diameter and ζ-potential reveals a nonlinear dispersion behavior governed by the interplay between solvent parameters and surface chemistry. Rather than being controlled by polarity alone, mCNC colloidal stability depends on the balance among dipolarity/polarizability, basicity, dielectric effects, and the nature of the surface functional groups. These observations are consistent with previous studies showing that CNC dispersibility in organic media improves in solvents that better stabilize interparticle charge and promote favorable solvent–particle interactions.

The same regression approach was applied to examine the relationships between solvent dipolarity/polarizability (π*) and the colloidal properties of each surface modification (Table ). Solvent dipolarity/polarizability was significantly associated with the ζ-potential of CNCs­(−NH₂) (R ² = 0.3944, p = 0.0005) and CNCs­(CH₂) (R ² = 0.5632, p < 0.001). The negative regression slopes indicate that the ζ-potential of these modifications became more negative as π* increased. No significant relationships were detected for CNCs­(−COOH) or CNCs­(−OH), showing that the electrokinetic response to solvent polarity was dependent on the introduced surface functionality.

3. Linear Regression Analysis of ζ-Potential and Hydrodynamic Diameter as a Function of Solvent Dipolarity/Polarizability for Surface-Functionalized CNCs.

Parameter	mCNC	Linear regression in function of the dipolarity/polarizability	R 2	p	Significance
ζ-potential	-COOH	y = −31.73x – 8.353	0.1274	0.0676	Not significant
	–OH	y = −22.54x – 31.09	0.0607	0.2155	Not significant
	–NH2	y = −65.36x + 14.93	0.3944	0.0005	Significant
	CH2	y = −55.37x + 10.45	0.5632	<0.0001	Significant
Hydrodynamic diameter	–COOH	y = 275.7x + 1109	0.0038	0.7751	Not significant
	–OH	y = −1510x + 2195	0.1031	0.1558	Not significant
	–NH2	y = −3123x + 4838	0.1039	0.1542	Not significant
	CH2	y = −4722x + 6577	0.1889	0.0235	Significant

In contrast to ζ-potential, hydrodynamic diameter was not strongly predicted by solvent dipolarity/polarizability. A significant relationship was observed only for CNCs­(CH₂), but its relatively low coefficient of determination (R ² = 0.1889, p = 0.0235) indicates that π* accounted for less than 20% of the variation in particle size. Thus, statistical significance in this case does not imply that solvent polarity was the principal factor controlling aggregation. The absence of consistent relationships for the other modifications suggests that CNC solvent interactions were also influenced by specific interactions, including hydrogen-bond acceptance, ion solvation, and the accessibility of the functional groups on the CNC surface. The corresponding regression analysis against solvent basicity (β) is provided in the Supporting Information (Table S3).

It is important to consider that in this system, all analyses were performed from aqueous CNC suspensions, which may preserve a residual hydration layer on the particles. This interfacial water could play a significant role in redispersibility and stability, particularly in organic media, where solvent–surface interactions may differ under completely anhydrous conditions. Using never-dried mCNCs helps maintain their structure, avoiding issues related to irreversible agglomeration and hornification, which limit redispersion and alter their intrinsic colloidal behavior, as is commonly observed in freeze-dried, organic solvent-based methods. Another perspective of this work would be the study of significant modifications with multiple degrees of substitution to expand the predictability of the parameters. Likewise, making a thorough study of different CNCs from different sources and sizes could be highly beneficial. Therefore, the findings reported here are specific to solvent-exchanged systems rather than to strictly dry or fully anhydrous environments.

4. Conclusion

The colloidal behavior of the functionalized CNCs depended on both the surface group and the medium. In water, stability followed the relationship between pH and the pKa of each group. The ionizable −COOH and −NH₂ modifications were the ones that responded to pH in their ζ-potential and hydrodynamic diameter, whereas −OH and CH₂ changed little across the range. In organic solvents, polarity by itself did not explain the results. Solvents with higher dipolarity/polarizability (π*) tended to give smaller aggregates, but how strong this effect was depended on the surface chemistry, and solvent basicity (β) added a further contribution through hydrogen bonding. Because of this, the Kamlet–Abboud–Taft parameters track the observed trends only in a semiquantitative way; the regressions were significant but had low R ² values. The residual sulfate half-esters, which stay negatively charged over the pH range tested, also add to the baseline charge of every sample.

Taken together, the results show that CNC dispersibility can be adjusted by matching the surface functionality to the ionization or solvation capacity of the solvent. This is useful when choosing surface chemistries and solvents to incorporate CNCs into advanced materials and nonaqueous formulations.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c06143.

• Table S1: solvent parameter values of the Smoluchowski/Hückel model per solvent, Table S2: pictures of the different mCNCs in the tested solvents, and Table S3: linear regressions using basicity as an independent variable (PDF)

The authors declare no competing financial interest.