Effect of Silica Particles on Moisture Resistance and Mechanical Performance in Flax/Epoxy RTM Composites: Matrix Modification

Abstract

Natural fibre-reinforced composites (NFCs) have attracted attention as sustainable alternatives to synthetic fibre composites. However, their hydrophilic nature and susceptibility to moisture absorption, especially in combination with process-related defects, can compromise long-term performance. This study critically examines the effects of hydrophobic fumed silica, incorporated into an epoxy matrix, on the processing, moisture uptake, and mechanical properties of flax/epoxy laminates produced via resin transfer moulding (RTM). Epoxy systems containing 0–5 wt% silica were characterised in terms of particle dispersion, rheological properties, thermal behaviour, and water absorption. Corresponding laminates were analysed for void content, Fickian diffusion behaviour, and tensile performance in dry and saturated states. Despite its hydrophobic surface treatment, silica increased resin water uptake and, at 5 wt%, led to a substantial rise in viscosity, poor fibre impregnation, and increased porosity. The resulting laminates exhibited faster and higher moisture uptake and significantly reduced wet mechanical properties, especially for highly filled systems. While thermal stability improved slightly, the overall findings revealed that the chosen silica-based matrix modification led to clear trade-offs and processing limitations under RTM conditions. This study highlights the importance of assessing such limitations early in the design process and demonstrates that the selected silica type is not a viable strategy for improving moisture resistance in NFCs.

1. Introduction

The growing demand for lightweight and more sustainable materials has drawn increasing attention to natural fibre composites (NFCs) as potential alternatives to glass- and carbon fibre-reinforced polymers [1]. Compared to synthetic fibres, the production of natural fibres such as flax requires less energy, and flax cultivation contributes to a favourable carbon balance through CO₂ uptake during growth [2]. Flax also exhibits a markedly lower greenhouse gas footprint (0.9 kg CO₂eq/kg) than glass (1.4–2.9 kg CO₂eq/kg) and carbon fibres (29.4 kg CO₂eq/kg) [3]. Combined with its low density (~1.5 g·cm⁻³), relatively high tensile modulus, and moderate strength, these advantages make flax one of the most promising and widely studied natural reinforcements for polymer composites [4,5]. In terms of specific modulus, flax fibres are competitive with glass fibres, making flax/epoxy laminates suitable for lightweight structural applications [1]. Epoxy resins are typically selected as the matrix owing to their high mechanical performance and chemical resistance. More recently, bio-based epoxy resins have been developed to reduce the dependence on fossil resources. However, most commercial systems currently possess only partial bio-based content (19–46.8%) and are not yet fully renewable [6].

Despite their favourable mechanical performance and sustainability benefits, flax fibre composites face a major challenge in humid environments due to their sensitivity to moisture. The hydrophilic nature of flax fibres, significantly influenced by their hemicellulose content, makes them prone to water uptake that degrades mechanical performance [7]. Compared to glass/epoxy systems, flax/epoxy composites absorb significantly more water, and acoustic emission studies have confirmed the resulting pronounced weakening of the fibre–matrix interface [8]. Ageing studies have shown that moisture exposure reduces stiffness much more severely than strength, with matrix plasticisation and microfibril reorientation contributing to degradation [9]. More recent work has demonstrated that the swelling of flax fibres in humid air generates interfacial stresses that promote microcracking and fibre–matrix debonding [10], while long-term water immersion leads to substantial stiffness and strength losses beyond those observed in humid air exposure [11]. Despite these challenges, flax/epoxy composites still absorb less water than other NFCs and therefore exhibit comparatively better retention of their properties after exposure [12].

Different technologies can be considered to overcome this drawback of moisture uptake, with modification of the matrix system being one of them. To modify NFCs and achieve desirable properties, nanoparticles can be added to the system. The type and shape of the particles, as well as their size and incorporation method, can influence the properties of the composite [13]. A homogeneous presence of particles within the matrix leads to a more uniform stress distribution and can prevent or slow down crack initiation and propagation, thereby improving the mechanical properties of the composite [13]. However, while nanoparticles are characterised by a large surface area, they tend to agglomerate in higher concentrations, which has a negative effect and makes achieving homogeneous distributions challenging [14]. Consequently, they are typically added to the matrix at relatively low weight percentages, with investigations of the mechanical characteristics often revealing their best performance at these low contents [13].

Various particle systems have already been investigated in this respect, including metal oxide nanoparticles (e.g.,TiO₂ or MgO) [15,16], rubber [17], nanoclay [18,19], cellulose nanocrystals [20], as well as fly ash [21]. In addition, fumed silica is often used as a filler in plastics or as a thickening agent [22]. Due to the three-dimensional structure of the fractal particle aggregates obtained from the high-temperature pyrolysis synthesis, the particles tend to form a network when the synthesis product is dispersed in a liquid medium, which can lead to increased viscosity [23]. These properties protect against sedimentation and enable the formation of stable dispersions [23]. Furthermore, surface modification of nanoparticles can be used to create desired properties and to ensure optimal compatibility with the chosen matrix [24,25,26].

Although silica is naturally hydrophilic, it can be surface-treated to become hydrophobic [27,28], giving the particles water-repellent characteristics while maintaining good compatibility with the epoxy. For NFCs reinforced with fumed silica manufactured by hand lay-up, studies have focused on mechanical and thermal properties: Parida et al. showed that fumed silica enhanced the tensile, flexural, and thermal properties of jute/epoxy laminates [29], while Arvinda et al. [30] studied jute–linen hybrid laminates manufactured from an epoxy matrix modified with 1–3 wt% fumed silica, reporting that 2 wt% gave the best mechanical performance and 3 wt% the highest thermal stability [30]. An often observed effect is the increase in thermal stability, with silica acting as a barrier on the surface and forming a protective layer over the polymer [31]. Silica can also improve the heat diffusion in the composite, therefore delaying ignition [32]. Fewer studies have examined moisture behaviour: in pure epoxy systems, silica nanoparticles have been shown to reduce moisture absorption [33], while in fibre-reinforced composites produced by compression moulding, results have varied from reductions [34] to no systematic change [35]. However, most silica-modified systems have been studied in the context of hand lay-up or compression moulding, where flow paths and impregnation conditions differ markedly from RTM. Consequently, the combined influence of silica-induced changes in the rheology of the resin and the flow behaviour of the RTM on the formation of the void, the uptake of water, and the mechanical response remains not adequately understood. In many cases, the correlation of chemical and microstructural findings with macroscopic tensile performance is also missing, leaving the link between nanoparticle modification, moisture transport, and structural properties underexplored, particularly regarding the potential limitations and trade-offs introduced by such matrix modifications in RTM-processed NFCs.

This interdisciplinary study addresses these gaps by examining how hydrophobic fumed silica affects both the epoxy matrix and the resulting flax/epoxy composites in an RTM process, combining nanoparticle and matrix characterisation with process- and performance-orientated composite testing to assess not only the potential benefits but also the technological limitations and degradation mechanisms introduced by silica modification. At the matrix level, the particle size distribution, dispersion state, rheology, thermal behaviour, and water uptake of the silica-modified epoxy are characterised. On the composite level, RTM-manufactured flax/epoxy laminates with varying silica contents are analysed in terms of void content, Fickian moisture uptake, and tensile performance in dry and moisture-saturated conditions. By systematically relating changes in resin properties to RTM impregnation quality, void formation, and the resulting moisture-induced degradation of stiffness and strength, this study links microscopic and chemical modifications to macroscopic composite behaviour. In doing so, it provides a mechanistic understanding of how silica-based matrix modification influences both the manufacturability and long-term performance of flax/epoxy composites, and it critically evaluates the trade-offs associated with silica-based matrix modification, highlighting that under the studied conditions, the selected modification may not provide a viable route for improving moisture resistance in RTM-processed flax/epoxy composites.

2. Materials and Manufacturing

2.1. Materials

The reinforcement material used in this study is a balanced 2/2 twill FlaxDry BL200 flax fabric (Eco-Technilin, Valliquerville, France) with an areal weight of 220 g·m⁻². The epoxy system consists of the bio-based resin SR GreenPoxy 56, containing 51% bio-based content, and the corresponding hardener SD 4771 (both supplied by Sicomin, Pluguffan, France). The silica nanoparticles used are hydrophobic fumed silica of the type Aerosil^® R104 (Evonik, Essen, Germany), hydrophobised with octamethylcyclotetrasiloxane (D4).

2.2. Matrix Modification Method

The silica nanoparticles were dispersed in the epoxy resin prior to hardener addition. Modified epoxy systems were prepared by incorporating silica powder into the resin to obtain nanoparticle contents of 1 wt%, 3 wt%, and 5 wt%, referring to the total weight of the matrix after the addition of the curing agent. The resin and nanoparticle mixture was stirred mechanically at 3000 rpm for 15 min, followed by further mixing at 2000 rpm for 45 min, and subsequently subjected to ultrasonication for 15 min to promote homogeneous dispersion. To remove trapped air, the dispersion was degassed under vacuum in a desiccator and placed on a vibratory plate before being used for composite manufacturing. Hydrophobically modified fumed silica (Aerosil® R104) was selected based on the commonly reported assumption that reducing silica surface polarity can limit water–filler interactions and thereby improve moisture resistance [36,37]. This study evaluated whether such an approach remained effective under RTM-relevant processing conditions, where dispersion and microstructural effects may dominate moisture transport behaviour.

2.3. Laminate Fabrication

The laminates were manufactured by RTM in a fixed cavity of 320 mm × 320 mm × 2 mm, defined by a rigid spacer (Figure 1). Resin infusion was performed under an injection pressure of 0.85 bar, assisted by a vacuum level of $-0.85$ bar applied at the outlet side. After infusion, the laminates were cured for 24 h at room temperature, followed by a post-cure for 24 h at 40 °C.

Four laminate configurations were manufactured: flax-reinforced epoxy with unmodified resin and flax-reinforced epoxy modified with 1 wt%, 3 wt%, and 5 wt% silica. For tensile testing, laminates were fabricated with a stacking sequence of ${\left[(0/90)(90/0)\right]}_S$. Plates of 320 mm × 320 mm × 2 mm were cut into specimens of 250 mm × 25 mm × 2 mm, and glass fibre tabs were bonded to the specimen ends to prevent gripping damage. The fibre volume fraction was calculated using the geometric method based on the fixed cavity dimensions and the fibre mass per plate (80.0–82.5 g). Across all batches, this resulted in a fibre volume fraction of 27–28%. The rigid spacer ensured negligible variation in laminate thickness between plates.

3. Experimental Methods

This section describes the experimental procedures used to investigate the effects of silica modification on the epoxy matrix and the resulting flax/epoxy laminates. It covers the characterisation of the silica nanoparticles and the modified epoxy system, followed by an evaluation of laminate quality, water uptake behaviour, thermal properties, and tensile performance.

3.1. Nanoparticle Characterisation

3.1.1. Dynamic Light Scattering (DLS)

For particle size analysis, dynamic light scattering with a wavelength of 630 nm (Zetasizer Nano ZS, Malvern, UK) was used. Before measurement, a particle sample was added to ethanol and placed in an ultrasonic bath to achieve a homogeneous dispersion.

3.1.2. Scanning Transmission Electron Microscopy (STEM)

To characterise the distribution and morphology of the particles before and after dispersing them in the resin, transmission electron microscopy was performed using a focused ion beam scanning electron microscope of the type Helios 5 UX DualBeam (Thermo Scientific, Dreieich, Germany) equipped with a STEM detector. Sample preparation involved dispersing the particles in acetone and depositing a small drop of the mixture on a TEM grid. The resin samples were mixed with the hardener and a tiny amount of the mixture was placed on the grid and cured.

3.2. Resin Characterisation

3.2.1. Rheology

Viscosity measurements of the uncured matrix were performed using an MCR-302 modular compact rheometer (Anton Paar, Graz, Austria) with a cone–plate geometry (55 $\mathrm{m}$$\mathrm{m}$, 1°). The viscosity was measured over a range of increasing shear rates at a constant temperature of 25 °C. The rheological properties were considered to account for the potential impact of the presence of particles in the matrix on the manufacturing process.

3.2.2. Differential Scanning Calorimetry (DSC)

For the determination of the glass transition temperature, DSC measurements were performed using a DSC 3plus device (Mettler Toledo, Gießen, Germany). The device was calibrated using an indium crucible. About 10 $\mathrm{m}$$\mathrm{g}$ of the sample was placed in a 40 $\mathsf{\mu }$$\mathrm{L}$ aluminium crucible with a perforated lid. The measurement was carried out with a nitrogen flow of 50 $\mathrm{m}$$\mathrm{L}$·$\min$⁻¹ and a heat rate of 10 $\mathrm{K}$·$\min$⁻¹.

3.2.3. Thermogravimetric Analysis (TGA)

In order to investigate the influence of particles on the thermal decomposition of the composite, TGA was carried out using a TGA/DSC 3 STARe system (Mettler Toledo, Gießen, Germany). About 15 $\mathrm{m}$$\mathrm{g}$ of the sample was placed in a 70 $\mathsf{\mu }$$\mathrm{L}$ aluminium oxide crucible and heated under a nitrogen atmosphere with a heating rate of 10 $\mathrm{K}$·$\min$⁻¹.

3.3. Laminate Characterisation

3.3.1. Void Content

The void content of the laminates was determined from polished cross-sections prepared using a Tegramin-30 grinding and polishing machine (Struers, Ballerup, Copenhagen) and examined by optical microscopy using a VHX-7000 digital microscope (Keyence, Osaka, Japan). For each laminate configuration (0, 1, 3, and 5 wt% silica), three specimens were taken from different positions along the flow path: one near the resin inlet port, one from the mid-section of the mould cavity, and one near the resin outlet port, giving a total of 12 specimens. The void area fraction was quantified for each specimen individually in order to capture local variations in void distribution along the flow path. These results were subsequently compared between laminate configurations and related to resin viscosity, water uptake behaviour, and tensile performance.

3.3.2. Water Uptake

Water absorption behaviour was characterised in accordance with ISO 62. Prior to immersion, specimens were oven-dried at 50 °C for 24 h, then placed in a dry chamber for 30 min to reach room temperature while preventing moisture reabsorption. Specimens were weighed immediately thereafter to determine the conditioned dry mass.

Conditioned specimens were placed in closed containers filled with distilled water and stored in a PR-3J climate chamber (ESPEC, Osaka, Japan) at 23 °C. For each weighing, specimens were removed from the container, surface water was wiped off using a dry cloth, and the samples were weighed within 1 min after removal to minimise evaporation losses. The immersion test for the flax/epoxy laminates was conducted for 30 d. Specimens were weighed at selected time points using a 1702 analytical balance (Sartorius, Göttingen, Germany) with an accuracy of $\pm 0.1 \mathrm{m}\mathrm{g}$.

Water uptake ($M_t$) was expressed as the relative mass increase:

$$M_t={\displaystyle \frac{W_t-W_0}{W_0}}\times 100\%,$$

(1)

where $W_t$ is the specimen mass at time t and $W_0$ is the conditioned dry mass.

The water uptake curves were fitted to Fick’s second law for diffusion in flat plates:

$${\displaystyle \frac{M_t}{M_{\infty }}}=1-{\displaystyle \frac{8}{{\pi }^2}}\sum _{n=0}^{\infty }{\displaystyle \frac{1}{{(2n+1)}^2}}exp\left(-{\displaystyle \frac{{(2n+1)}^2{\pi }^{2}Dt}{h^2}}\right),$$

(2)

where $M_{\infty }$ is the equilibrium water uptake, D is the diffusion coefficient, and h is the specimen thickness. This model describes water uptake where the mass gain approaches equilibrium asymptotically with time.

3.3.3. Tensile Testing

Tensile tests were conducted in accordance with ISO 527-4 using a Z050 AllroundLine universal testing machine (ZwickRoell, Ulm, Germany) equipped with a 50 kN load cell. Strain was measured using strain gauges, and the tensile modulus was determined from the initial linear region of the stress–strain response. At least five specimens were tested per laminate configuration in the dry state, and three specimens per configuration were tested after water immersion.

4. Results and Discussion

In the following sections, the results of the investigated relationship between nanoparticle addition, resin rheology, laminate quality, water uptake behaviour, and tensile performance are shown. These effects are discussed in an integrated manner, establishing the causal chain from silica modification of the matrix to the final composite properties.

4.1. Structural and Rheological Properties of Silica-Modified Resins

Aerosil^® R104 silica particles exhibit the typical structure of fumed silica. The amorphous particles combine to form smaller aggregates. DLS measurement in ethanol reveals a broad particle size distribution with an average particle size of 295 nm, indicating pronounced agglomeration of the fumed silica (Figure 2). DLS measurements in ethanol are used for comparative purposes only and do not represent absolute agglomerate sizes in the epoxy matrix.

This observation is corroborated by the STEM micrographs of both the isolated particles and the particles embedded in the cured resin, which clearly show persistent aggregates rather than dispersed primary particles (Figure 2 and Figure 3). The contrast in the images of the modified matrix varies depending on the thickness of the sample. The structure of the particles in the resin does not differ from the powder form. A significant increase in particle density within the matrix can be observed, particularly in the resin sample with a 5 wt% silica content. However, due to the localised nature of STEM imaging, individual micrographs should be interpreted as representative examples rather than as a direct visual measure of particle concentration or network density. The presence of such aggregates demonstrates that mixing the particles into the epoxy system does not break them down to their primary size. Despite the combined use of high-speed mixing and ultrasonication, complete suppression of silica agglomeration is not expected for fumed silica due to strong particle–particle interactions and the absence of solvent or surface-reactive compatibilisers [23,38,39].

The presence of these stable aggregates, which can agglomerate directly, explains the rheological behaviour of the modified resin. As shown in Figure 4, the viscosity of the 1 wt% resin remains comparable to that of the neat epoxy, while the 3 wt% and especially the 5 wt% formulations exhibit a pronounced increase in viscosity. The 5 wt% resin additionally shows strong shear-thinning behaviour, which is characteristic of fumed silica, forming a three-dimensional percolated network that collapses under shear. This behaviour is also observed in other works [40,41]. The combined evidence for strong agglomeration (Figure 2 and Figure 3) and the increase in viscosity with particle content (Figure 4) demonstrate that the silica particles substantially modify the flow behaviour of the resin.

4.2. Influence on Laminate Quality

The addition of silica nanoparticles increases resin viscosity, which has critical implications for laminate manufacturing. Elevated viscosity reduces the ability of the resin to flow into fibre tows and fill narrow pores during RTM processing, thereby limiting impregnation quality and increasing the likelihood of void formation. Figure 5 shows representative polished cross-sections taken from the mid-section of the mould cavity. The 0 wt% specimen shows no detectable voids, while the 1 wt% and 3 wt% specimens contain a small number of isolated voids. In contrast, the 5 wt% specimen exhibits a noticeably higher void population, indicating impaired resin infiltration.

Figure 6 shows the relationship between resin viscosity and void content with increasing silica loading. Void content is quantified as a global measure across the full area of the sampled cross-sections, as the optical microscopy method used does not permit a reliable distinction between intra-tow and interlaminar void locations. While the void content remains below 1% for 0–3 wt% silica, it increases sharply to 11.1 ± 5.9% at 5 wt% silica, with large scatter along the flow path. This sharp increase coincides with the elevated resin viscosity at 5 wt% silica, which increases by a factor of ∼2.4 relative to the 0 wt% silica system. The viscosity is measured at a process-relevant shear rate of ∼2.5 s⁻¹, corresponding to the nearest available experimental data point to the estimated RTM shear rate. The shear rate is calculated from an RTM flow velocity of 1 × 10⁻³ $\mathrm{m}$·$\mathrm{s}$⁻¹ and a cavity thickness of 2 $\mathrm{m}$$\mathrm{m}$ using $\dot{\gamma }\approx 4v/h$. Accordingly, processing-induced effects on laminate quality become significant only in 5 wt% silica, where elevated viscosity leads to increased void formation, with direct consequences for subsequent water uptake and mechanical performance.

4.3. Water Uptake Behaviour

The water uptake $M_t$ is calculated according to Equation (1). Figure 7a shows the water uptake behaviour of flax/epoxy laminates at varying silica contents. All silica-containing laminates exhibit higher equilibrium water uptake than the reference material, with uptake increasing from 6.19% at 0 wt% silica to 9.27% at 5 wt% silica. The particularly high uptake observed for the 5 wt% silica laminates corresponds to their increased void content, which results from elevated resin viscosity during RTM injection. The 1 wt% and 3 wt% silica laminates also show increased uptake, although their void contents remain close to that of the 0 wt% material. This indicates that the increase at intermediate particle contents cannot be explained by voids alone.

A matrix-related mechanism was therefore examined separately. Fibre-free resin plates containing 0 wt% and 5 wt% silica were manufactured by RTM to ensure identical processing conditions and enable direct comparison between resin and laminate systems. The absence of the fibre architecture eliminated dominant mechanisms such as local flow hindrance and bundle-related air entrapment. This fibre-free resin system therefore isolated the effect of silica on the water uptake behaviour of the epoxy matrix (Figure 7b). As expected, all laminates absorbed more water than their corresponding resin specimens due to the hydrophilic and porous structure of flax fibres.

The water uptake curves were fitted to a single-stage Fickian diffusion model (Equation (2)), which was applied consistently across all silica contents to enable direct comparison of fitted parameters. The model described the experimental data well ($R^2>0.95$). In this context, $M_{\infty }$ represents the long-term equilibrium water content, while the diffusion coefficient D characterises the initial transport rate. Increases in $M_{\infty }$ indicate a higher sorption capacity, whereas changes in D reflect alterations in transport behaviour.

The fitted parameters in

Table 1. Equilibrium water uptake $M_{\infty }$, diffusion coefficient D, and their coefficients of variation (CVs) obtained from Fickian fits for flax/epoxy laminates (0–5 wt% silica; 30 d immersion) and resin specimens (0 wt% and 5 wt% silica; 92 d immersion).

System	Silica Content (wt%)	${\mathit{M}}_{\mathit{\infty }}$ (%)	CV (${\mathit{M}}_{\mathit{\infty }}$)	D (${10}^{-13}$ m2·s−1)	CV (D)
Flax/epoxy laminates (30 d)
Flax/epoxy	0	$6.19$	0.0043	$2.599$	0.061
Flax/epoxy	1	$7.49$	0.0058	$2.927$	0.027
Flax/epoxy	3	$7.24$	0.0120	$3.305$	0.091
Flax/epoxy	5	$9.27$	0.1328	$3.341$	0.088
Resin specimens (92 d)
Resin	0	$3.13$	0.0021	$0.371$	0.019
Resin	5	$4.49$	0.0083	$0.348$	0.031

show that silica primarily affects $M_{\infty }$ rather than D. In the laminates, $M_{\infty }$ generally increases with silica content, although this trend is not strictly monotonic. D changes only slightly across the series, indicating that silica has little influence on the transport kinetics. This pattern is consistent with a sorption-controlled mechanism. The coefficient of variation (CV) in $M_{\infty }$ provides further insight. In the laminates, the scatter remains low up to 3 wt% silica but rises sharply at 5 wt%, consistent with the increased void content and more heterogeneous structure at this concentration.

The resin specimens containing 0 wt% and 5 wt% silica show a similar increase in $M_{\infty }$ with increasing silica content while maintaining low scatter, and D remains nearly constant. This confirms that the enhanced water uptake at 5 wt% silica originates at the matrix level. While resin specimens are not tested at intermediate loadings, the low void content in the 1 wt% and 3 wt% silica laminates suggests that a similar matrix-level mechanism operates at these loadings. The higher variability observed in the laminates can be attributed to the amplification of this matrix-driven effect by fibre architecture and processing-induced heterogeneities.

Together, these findings point to a two-level mechanism. Matrix modification alone increases equilibrium water uptake, as demonstrated by the resin systems. In the laminates, this matrix-driven effect is either amplified or moderated by processing-induced features such as voids. At 1–3 wt% silica, the response is predominantly matrix-driven, whereas at 5 wt% silica the higher uptake results from the combined contribution of increased matrix sorption capacity and additional transport pathways introduced by microstructural defects.

The morphology of the silica particles provides additional insight into this matrix-level uptake mechanism. The Aerosil^® R104 fumed silica used in this study forms branched, interconnected structures, as observed in Figure 2. These aggregates become progressively denser with increasing silica content, as shown in Figure 3. This network formation creates extended particle–matrix interfacial zones that may facilitate moisture uptake.

Whether these interfacial pathways act as preferential moisture routes likely depends on the bonding quality at the particle–matrix interface. In the present system, the hydrophobic surface treatment lacks functional groups capable of forming covalent bonds with the epoxy network. When dispersed in the matrix, the particle–matrix interface may rely more on physical interactions, which could be more susceptible to moisture infiltration. This interpretation is supported by the observations of weak interactions between hydrophobic fumed silica and epoxy resin reported by Ekin et al., which led to moisture-induced degradation at the particle–matrix interface [35].

The interfacial pathway mechanism may help to explain why certain approaches reported in the literature lead to reduced water uptake. Surface treatment with coupling agents directly addresses the bonding issue by forming covalent particle–matrix links, which reduce water uptake [33]. Alternatively, particle morphology can limit the extent of interfacial pathways: discrete spherical particles form isolated particle–matrix regions rather than the interconnected networks characteristic of branched aggregates, resulting in substantially lower water absorption [34]. The present results suggest that without epoxy-reactive functionalisation, hydrophobic fumed silica increases water uptake due to the combined effects of weak interfacial bonding and branched aggregate geometry.

Beyond water uptake behaviour, the glass transition temperature ($T_g$) can be affected by both silica addition and water absorption. Silica nanoparticles influence $T_g$ through free volume effects [33], while water reduces $T_g$ through plasticisation [42]. As silica addition also influences the extent of water uptake, the magnitude of the water-induced change in $T_g$ is expected to depend on silica content. The influence of silica content on $T_g$ under dry and wet conditions is therefore examined in the following subsection.

4.4. Effects of Silica and Moisture on $T_g$

DSC was performed to evaluate the glass transition temperature (T_g) of the cured resin samples with and without particles before and after water absorption (see Figure 8).

The cured epoxy has a T_g of 52.25 °C. It shifts to a lower value at 50.16 °C with a 5 wt% particle content. Water absorption also shows a decreasing effect on T_g. For the 0 wt% sample, T_g decreases to 43.34 °C and for the 5 wt% sample, T_g occurs at 38.03 °C after water absorption. The measurements show a gradual reduction in the dry glass transition temperature with increasing particle content, which can be attributed to disruptions in the polymer network by agglomerates and local free volume effects. After water absorption, T_g decreases further for all configurations, with the strongest reduction observed in the 5 wt% laminate. This can be explained by the higher moisture content, which increases chain mobility and plasticises the polymer network. The combined presence of silica and moisture therefore amplifies the softening of the matrix. This is in line with existing studies that report a decrease in T_g for epoxy composites highly filled with bulk moulding compounds (BMCs), which can be explained by the increase in distance between polymer chains due to the presence of BMC powder [43]. In other cases, the addition of modified silica nanoparticles to epoxy composites has been reported to lead to an increase in T_g. Babaranova et al. achieved higher T_g in modified epoxy composites by using a greater amount of silica particles (10–30 wt%) obtained from a 30% colloidal dispersion with surface-treated silica nanoparticles (10–15 nm) as well as varying hardener concentrations. The increase in T_g was explained by the increase in junctions in the cross-linked polymer (showing a maximum at 20 wt% silica). A decrease in T_g at higher wt% was explained by a hindrance of the curing process and microphase separation caused by the higher viscosity of the system. Also, a higher concentration of the hardener as well as a longer curing time were reported to be beneficial to obtaining higher T_g in this case [44]. Ou and Shiu compared the influence of different types of fumed silica on the T_g of modified epoxy composites by adding varying wt% (0.3–7) of Aerosil^® (A90, A200, A300), with A300 showing the strongest increase in T_g. An increase in T_g with a decrease in particle size was observed and explained by good adhesion between the polymer and particles, which could have restricted segmental motion. Finer particles provided more contact area between the epoxy matrix and the silica, supporting interfacial forces [45]. Zeng et al. reported a shift to higher T_g values for epoxy composites reinforced by mixing noncovalently functionalised graphene sheets, explaining this with the nanosheets restricting movement and decreasing free volume, as well as causing a higher cross-linking density by influencing the curing reaction [46]. Similar findings were also reported, with a higher T_g observed after the addition of graphitic and ceramic-based nanofillers, which was attributed to the restriction of movement [47]. As shown in Figure 8, water absorption also decreases the T_g. This can be explained by water acting as a plasticiser. As they are present between the polymer chains, the water molecules effectively increase the free volume. With the restraining forces weakened and more space available, the polymer chains require less thermal energy for motion. This results in the material transitioning from a rigid to a softer state at a lower temperature, as often observed for epoxy composites after water absorption [48,49]. The lower values for the 5 wt% sample, showing an even greater decrease after water absorption than the 0 wt% sample, could be explained by the particles causing cracks and cavities in the composite, leading to a greater absorption capacity, as shown in the water absorption tests.

4.5. Thermogravimetric Analysis

While T_g is sensitive to changes in polymer chain mobility and interfacial free volume, thermal stability assessed by TGA is governed by degradation kinetics, which can be positively influenced by the barrier effect of silica particles [50,51]. In order to determine the influence of the particles on the thermal properties of the composite material, the decomposition is analysed using TGA. The course of relative mass versus temperature and the corresponding first derivative of the fibre and matrix material, as well as the matrix-modified composites, are shown in Figure 9.

The thermal degradation of flax fibre occurs in three stages (see Figure 9a). The first two can be attributed to water evaporation and the decomposition of pectins and hemicellulose. The decomposition of cellulose causes the main weight loss [52,53]. This behaviour is also observed for other natural fibre materials [54]. The thermal degradation of the matrix material is affected by the addition of particles (see Figure 9b). The 5 wt% matrix sample shows weight loss in a broader temperature range, while the degradation temperature is not significantly influenced. For the epoxy with particle content, a higher amount of carbon can be observed. The 0 wt% sample also shows carbon residue as the leftover weight is not 0 %, which can be explained by the carbonisation of the epoxy matrix as well as the mass of the remaining particles [55]. For the composite laminates (Figure 9c), the point of greatest weight loss occurs at about 300 °C to 310 °C and shifts to slightly higher temperatures in the presence of particles in the composite. The curve for the pure composite (0 wt%) also shows a second stage of decomposition around 370 °C, which decreases as the particle content increases and is not visible for the 5 wt% sample. The effect of the particles on the degradation temperature in this case is small. There is a significant difference between composites without and with incorporated silica particles, but there is no significant difference between the different higher particle contents. The presence of particles leads to slightly higher thermal stability, as the weight loss is shifted to higher temperatures in the presence of particles. This is in agreement with other findings [56]. Kim et al. reported an increase in thermal stability resulting from the addition of fumed silica to epoxy resin compared to pure epoxy resin. They furthermore observed an increase in char residue with the increase in silica content [55].

4.6. Mechanical Characterisation

Figure 10 shows the mechanical properties of flax/epoxy laminates before and after 30 d of water immersion. Before immersion, the incorporation of silica did not significantly affect the mechanical properties. The reference laminate exhibited a tensile strength of 87.1 MPa and a tensile modulus of 8.86 GPa. Laminates containing 1, 3, and 5 wt% silica showed only minor variations in strength (+4%, +1%, and $-1$%, respectively) and modulus (+1%, +5%, and 0%), all of which fell within the experimental scatter and showed no systematic dependence on silica content.

After 30 d of water immersion, a degradation in mechanical performance was observed for all laminates, with the most severe losses occurring in 5 wt% silica. The retention of tensile strength decreased from 85% for the reference laminate to 74%, 75%, and 63% for the 1, 3, and 5 wt% silica laminates, respectively. A similar trend was observed for the tensile modulus, with the retention of the tensile modulus decreasing from 36% for the reference laminate to 24%, 26%, and 17% for the 1, 3, and 5 wt% silica laminates, respectively. This progressive loss in mechanical performance was consistent with the increased equilibrium water uptake measured for the silica-containing laminates.

The stress–strain responses further supported these trends (Figure 11). In the dry state, all laminates exhibited an approximately linear initial stress–strain response, followed by a gradual deviation from linearity and eventual failure near 1.5% strain. After water conditioning, the stress–strain curves showed reduced initial slopes, lower peak stresses, and increased scatter. The laminate containing 5 wt% silica exhibited the most pronounced degradation, displaying a markedly more compliant response than both the reference laminate and the lower silica contents.

The mechanical degradation observed after water immersion resulted primarily from matrix plasticisation and interfacial weakening. The substantial reductions in tensile modulus can be attributed to matrix plasticisation, which lowers the $T_g$ of the epoxy matrix. Previous studies on flax/epoxy composites have similarly shown that water acts as a plasticiser, leading to significant reductions in stiffness [9]. Water absorption reduced $T_g$ by approximately 9 °C for the reference epoxy and 12 °C for the 5 wt% silica-modified system, such that the conditioned laminates operated closer to $T_g$ during tensile testing at 23 °C. This increased proximity to $T_g$ enhanced polymer chain mobility and led to a marked reduction in stiffness. The larger $T_g$ depression in the silica-filled system was consistent with its higher equilibrium water uptake and explained the more severe modulus loss.

In addition to matrix plasticisation, reductions in strength and stiffness may also arise from interfacial effects associated with the incorporation of fumed silica. In the absence of coupling agents, the resulting particle–matrix interfacial regions may be mechanically less efficient under tensile loading. After water exposure, these regions are particularly susceptible to local weakening, which can impair stress transfer and promote stress concentrations. Due to the branched, interconnected aggregate structure of fumed silica, such weakened interfacial zones may become spatially connected, thereby amplifying their influence on the macroscopic tensile response. Beyond particle–matrix effects, fibre–matrix interface degradation may also contribute to reductions in mechanical performance. Flax fibres are hydrophilic and absorb moisture, which can lead to fibre swelling and weakening of the fibre–matrix bond.

At the highest silica loading, processing-induced voids introduced an additional degradation mechanism. Void analysis revealed that the 0–3 wt% silica laminates contained an average void content below 1%, whereas the 5 wt% silica laminate exhibited an average void content of approximately 11% due to impaired resin infiltration during RTM processing. These voids acted as internal defects that could reduce the effective load-bearing cross-section and promote stress concentrations, while also providing preferential sites for moisture accumulation and damage initiation. The combined effects of severe matrix plasticisation, interfacial weakening, and processing-induced voids accounted for the significant losses in stiffness and strength observed in the 5 wt% silica system.

5. Conclusions

This work investigated how the incorporation of hydrophobic silica nanoparticles influenced both the epoxy composites and RTM-produced flax/epoxy laminates, connecting nanoparticle-driven changes in resin behaviour to laminate quality, moisture absorption, and tensile properties. The key findings were:

• Fumed silica formed aggregates in the epoxy, as evidenced by DLS and STEM measurements, and did not break down to primary particle size during processing. These aggregates caused a strong, non-linear increase in viscosity with rising particle content, with 5 wt% silica leading to pronounced shear-thinning behaviour.
• The higher viscosity at 5 wt% silica content impaired RTM impregnation and led to a substantial increase in void content, whereas laminates with 0–3 wt% silica showed negligible porosity. Thus, matrix modification directly influenced laminate microstructure through its effect on flow behaviour. Future work should explore process parameter optimisation (e.g., injection pressure, mould temperature, and resin preheating) to reduce void formation at silica loadings above 3 wt%.
• Both the epoxy composites and the flax/epoxy laminates showed higher equilibrium water uptake with increasing silica content. In the epoxy composites, this was attributed to silica-related sorption sites and interfacial free volume. In the laminates, elevated void contents, particularly at 5 wt%, further accelerated moisture ingress via capillary transport, causing the highest and fastest uptake in the highest tested concentration.
• DSC revealed a decrease in glass transition temperature with increasing particle content and after water absorption, with the largest decrease seen for the 5 wt% epoxy composite. Together with TGA, which showed only a modest gain in thermal stability, this indicated that silica-induced softening and moisture plasticisation dominated over any small improvements in high-temperature behaviour.
• Under dry conditions, tensile stiffness and strength were only weakly affected by silica addition. After water saturation, however, stiffness and strength were significantly reduced for all laminates, with the most severe degradation in the 5 wt% configuration. This behaviour was consistent with the combined effects of higher moisture uptake and void-related stress concentrations in the highly filled material. As mechanical testing was conducted at 23 °C, the observed degradation reflected both moisture-induced effects and the proximity of the reduced $T_g$ to the test temperature. Multi-temperature testing is recommended for future studies to further distinguish these mechanisms.

Overall, the results demonstrate that silica-based matrix modification in flax/epoxy NFCs involves a trade-off between the potential benefits in thermal behaviour and the detrimental effects of increased viscosity during processing, void formation, and moisture sensitivity. For RTM processing, silica loadings up to 3 wt% appear more suitable, as they limit porosity while the mechanical performance of the unfilled system is largely preserved, while higher loadings compromise manufacturability and long-term tensile properties under wet conditions, as is observed for the 5 wt% silica content. These findings are specific to vacuum-assisted RTM processing with Aerosil^® R104. Applicability to other manufacturing processes (e.g., VARTM and HP-RTM), silica types, or surface treatments would require further experimental verification. Notably, despite the promising effects reported for hand lay-up or compression moulding, modifying the matrix did not improve the moisture resistance of RTM-manufactured composites. These findings suggest that the addition of silica to the resin is not the most effective way to enhance moisture resistance in NFC manufactured by RTM. Targeted strategies, such as placing silica primarily at the fibre–matrix interface (e.g., through direct fibre coating), may mitigate the drawbacks associated with increased matrix viscosity while enabling interfacial property tailoring. Further investigation of such approaches is warranted in future studies.