For the synthesis of the bio-based adhesive, BSFL flour (kindly donated by Procens, from Balcarce, Argentina) was used, with a composition of 58% protein, 14% lipids, 8% moisture, 8% ash, 6% fibers, 5% Ca and 0.3% Mg (data provided by the supplier). The chemical reagents (pure for analysis—p.a. quality) used for the manufacture of the adhesive were sodium hydroxide (NaOH) and n-hexane, purchased from Biopack (Buenos Aires, Argentina). Eucalyptus grandis lumber was purchased from Aserradero Vagol (San Martín, Argentina), which was subsequently chipped and milled in the laboratory to produce the particles used for board manufacturing. Sugarcane bagasse particles were sourced from a core business unit located in Jesús Menéndez, Las Tunas, Cuba, and were used as received, without additional pretreatment, except for oven-drying to adjust the moisture content prior to board manufacture.

The BSFL flour was defatted by Soxhlet extraction with n-hexane at 70°C for 6 h [35]. Then, the sample was ground to reduce the particle size. The adhesive formulation followed the procedure described by García et al. [27], which consisted of preparing an alkaline solution of 6 g of NaOH in 150 g of H₂O and subsequently incorporating 37.5 g of degreased BSFL. The mixture was homogenized in an industrial stirrer at 300 rpm for 1 h at room temperature. At the end of the process, a homogeneous suspension with pseudoplastic behavior and suitable for spray application was obtained.

Rheological characterization of the BSFL-based adhesive was carried out using an Anton Paar MCR301 rheometer (Graz, Austria) equipped with a parallel plate geometry (diameter 50 mm, PP50). Viscosity (η) profiles were obtained in rotational mode, applying a shear rate (γ˙) sweep from 0.01 to 1000 s⁻¹. Viscoelastic behavior was assessed in oscillatory mode. Initially, an amplitude sweep test was performed to determine the linear viscoelastic region (LVR), using a constant frequency of 1 s⁻¹ and a strain range from 0.01% to 100%. Based on this, a strain of 0.1% (within the LVR) was selected for the subsequent frequency sweep test, which was conducted over a frequency range from 100 to 0.1 s⁻¹. This strain level ensures that the structure of the sample remains undisturbed during measurement. Thixotropic behavior was evaluated through a three-interval thixotropy test (3ITT), employing an oscillation–rotation–oscillation sequence. Interval 1 consisted of oscillatory measurement at 0.1% strain and 1 s⁻¹ frequency to establish the baseline. In interval 2, a high constant shear rate of 2000 s⁻¹ was applied in rotational mode to disrupt the structure. Interval 3 repeated the initial oscillatory conditions to monitor structural recovery post-shear. 3ITT results were analyzed by two different methods. First, the recovery ratio was calculated through Eq. (1): (1)Recoveryratio=G3′G1′⋅100where G1′ is the baseline of the storage modulus at interval 1, and G3′ is the final plateau of the storage modulus at interval 3. The second method is the time of recovery, which means the time at interval 3 when G′=G′′ and gel-like behavior is restored.

Sugarcane bagasse fibers were used; they were crushed and dried to produce the particle board. A hammer mill with a mesh diameter of 11 mm was used for this purpose. The sugarcane bagasse used in this study had a moisture content of 9% and the following chemical composition (dry basis), as provided by the supplier: 1.2% ash, 1.9% extractives (ethanol/benzene), 46.0% cellulose, 23.3% lignin, and 26.3% pentosans. The Eucalyptus grandis particles were obtained from 50 mm × 50 mm × 3000 mm wooden slats purchased from a local supplier and immersed in water for 24 h to increase plasticity and facilitate processing in a chipper. The resulting chips were then processed using a hammer mill to achieve the final particle size.

At this stage, the particle size distribution of both sugarcane bagasse fibers and Eucalyptus grandis wood particles was characterized by recording the proportion of material retained on a series of ASTM (American Society for Testing and Materials) sieves: 4760 μm (mesh 4), 2830 μm (mesh 7), 2000 μm (mesh 10), 1000 μm (mesh 18), 500 μm (mesh 35), and 250 μm (mesh 60). The proportion retained in each size category for both materials was analyzed using contingency table methodology. The individual dimensions of fibers and particles retained on each sieve were measured (length and thickness) using a digital caliper (n = 15, for each size category and material), and the slenderness ratio (length/thickness) was calculated. These results were analyzed through analysis of variance (ANOVA), followed by a post hoc comparison using Fisher’s Least Significant Difference (LSD) test.

Particleboards were manufactured from sugarcane bagasse, with a density around 400, 600, 700, and 800 kg/m³. The samples were identified as D400, D600, D700, and D800, respectively (Fig. 1). The dry adhesive content was 12 wt.% in the final board and was applied by spraying.

For the manufacture, the impregnated particles were dried in a forced convection oven at 70 ± 1°C until their moisture content was reduced to 8%. Subsequently, pre-pressing was carried out at 10 MPa for 5 min at room temperature. Finally, curing was carried out in a press at 160°C using three successive compression intervals: 7.5 MPa for 3 min, 5 MPa for 6 min, and finally 2.5 MPa for 3 min. This stepped pressure schedule was selected based on our previous work on protein/UF adhesives [36], as it allows gradual removal of water, reduces vapor blistering and cracks compared with constant pressure pressing, and ensures more efficient curing and reproducibility of the panels. Once cured, panels were cut 415 mm long, 415 mm wide and 9 mm thick.

To investigate the influence of particle size on board properties, a panel with a target density of 700 kg/m³ was produced using Eucalyptus grandis particles (Fig. 1E), employing the same processing variables as those used for sugarcane bagasse panels. It is worth noting that Eucalyptus is one of the materials commonly used in large-scale commercial particleboard production [37]. The Eucalyptus grandis board was manufactured following the same methodology applied for the sugarcane bagasse-based panels.

The surface texture of the particleboard was evaluated with an electronic contact roughness tester according to the Deutsches Institut für Normung—DIN 4768 [38]. Three measurements were taken on each sample using a SURTRONIC 3+ roughness tester, with the following parameters: measurement modulus (R_z, cut off) of 2.5 mm and measurement length (L_m) of 12.5 mm. To qualitatively evaluate the surface porosity, a Lancet optical microscope with 20× optical magnification and a 12-megapixel resolution camera was used.

Regarding mechanical properties, two tests were carried out: the tensile perpendicular to the plane and the static bending test, both performed on a universal testing machine (INSTRON 5982, Norwood, MA, USA). The tensile perpendicular to the plane test was carried out following the guidelines of the European Standard EN 319 [39] standard on specimens of 50 mm × 50 mm × 9 mm. A crosshead speed of 7.2 mm/min was used. The internal bonding (IB) was calculated as the ratio between the maximum load and the cross-section of the test piece. A total of seven specimens were tested, and results are reported as mean ± standard deviation. The three-point bending test was performed following the guidelines of the EN 310 [40] standard with the aim of determining the MOR and the MOE. Nine specimens were tested for each condition, and the results are reported as mean ± standard deviation.

The swelling test was carried out following the guidelines of the EN 317 [41] standard with the aim of determining the water absorption (WA) and the thickness swelling (TS). For this test, specimens of 50 mm × 50 mm × 9 mm were immersed in water at room temperature. The thickness and weight of the swelled samples were measured at 2 and 24 h.

The measured MOR, MOE, IB, and TS values of the particleboards bonded with the BSFL-based adhesive were assessed according to the classification criteria defined in the European standard EN 312 for wood particleboards [42].

The MOR, MOE, IB, TS and WA values were compared by analysis of variance (ANOVA). Mean values were analyzed with the Tukey-Test for paired comparisons at a significance level P 0.05 using the Origin Pro 8 software (OriginLab Corporation, Northampton, MA, USA).

Fig. 2 presents the viscosity curves of BSFL-based adhesive. Given that particleboard adhesives are typically applied by spraying under high shear conditions (above 1000 1/s), the Carreau–Yasuda model (Eq. (2)) was employed to fit the experimental data and extrapolate viscosity values at shear rates beyond the measurement limits of the rheometer: (2)η(γ˙)=η0−η∞(1+(λ⋅γ˙)p1)1−pp1+η∞where p1 is the Yasuda exponent, λ [s] is the relaxation time, p is the power-law index, η0 is the viscosity zero, and η∞ is infinite shear viscosity. Based on this model, it was estimated that the viscosity of BSFL-based adhesive in a shear rate range from 6000 to 15,000 s⁻¹ is around 0.157 and 0.101 Pa·s.

Amplitude sweep tests were conducted to identify the linear viscoelastic region (LVR). For the BSFL-based adhesive, the LVR was found to extend up to 0.1% strain (Fig. 3A), where both the storage modulus (G′) and loss modulus (G′′) remained constant with increasing deformation. Additionally, the flow point, defined as the crossover between G′ and G′′, was observed at a shear stress of approximately 10 Pa (Fig. 3B).

The frequency sweep test was carried out within the previously determined LVR to evaluate the viscoelastic behavior of the BSFL-based adhesive. As shown in Fig. 4, G′ remained higher than G′′ across the entire frequency range tested, indicating a predominantly elastic response. Both G′ and G′′ showed a slight increase with frequency, suggesting the presence of a weak gel-like network structure. The complex viscosity (η*) exhibited a decreasing trend with increasing frequency, consistent with shear-thinning behavior.

The 3ITT was performed to evaluate the structural recovery behavior of the BSFL-based adhesive under shear. As shown in Fig. 5, the crossover between G′ and G′′ occurred 12 s into Interval 3, indicating a rapid transition back to a solid-like structure after shear cessation. The recovery ratio was approximately 36.1%, which shows that the material quickly regained a gel-like character but only partially rebuilt its internal structure.

The morphological characterization of the bagasse particles is shown in Table 1. Analyzing the proportion of material retained in each mesh, a bimodal distribution was observed, with two peaks corresponding to mesh sizes 7 and 18. When comparing the morphometry of the retained fractions, differences were found in all three measured variables (length, thickness, and slenderness), although similar trends were observed for both length and thickness, with particle dimensions decreasing as the mesh number increased. However, the slenderness ratio remained statistically unchanged across mesh sizes for bagasse particles, ranging from 74.1 (mesh 10) to 87.8 (mesh 7), with a mean value of 78.2 for all sieves.

When comparing Eucalyptus grandis and bagasse particles, significant differences were observed in the distribution of particle sizes across mesh fractions, as confirmed by a highly significant Pearson Chi-square statistic (χ² = 54.15; p < 0.0001). Mainly, eucalyptus particles were shorter and thicker than bagasse particles. Both materials followed similar overall trends, with length and thickness decreasing as mesh number increased. However, the rate and magnitude of these changes differed by material: length reduction was more pronounced in bagasse, while thickness decreased more sharply in eucalyptus particles. For slenderness ratio, bagasse particles exhibited no significant variation across mesh sizes, while eucalyptus particles showed an increasing slenderness ratio as mesh size decreased. Moreover, the slenderness ratio was consistently and significantly higher in bagasse particles across all mesh fractions.

The measured density values of the manufactured boards, along with surface roughness, are presented in Table 2. Significant differences were found between the densities of the bagasse particleboards, except for D700 and D800. The density of the Eucalyptus grandis panel was 740 ± 73 kg/m³, comparable to the D700 and D800 panels. As shown in Table 2 and Fig. 6, the surfaces of the finished boards exhibited a good overall appearance: at higher densities (Fig. 6C,D), the surface was notably smoother, with fewer and smaller voids. The roughness of the D400 board could not be determined because it exceeded the maximum measuring range of the device. This parameter decreased as the density of the board increased. Surface roughness was not measured for the eucalyptus-based panel, as the particles were predominantly cubic and in the centimeter range, resulting in voids that prevented reliable measurement with the available equipment.

The results of MOR, MOE, and IB as a function of panel density are shown in Fig. 7. A significant increase in these properties was observed with increasing panel density. For instance, MOR for the D400 board was 5.34 MPa, whereas for the D800 board it reached 30.26 MPa, marking an increase of 466%. Similarly, MOE increased 333%, from 808 MPa for the D400 board to 3500 MPa for the D800 board. The IB values also showed a gradual increase with density: 0.09 MPa (D400), 0.18 MPa (D600), 0.43 MPa (D700), and 0.80 MPa (D800). Notably, the IB value of the D800 sample was 900% higher than that of the D400 sample. These results demonstrated a strong positive correlation between the mechanical properties and the density of the particleboard, indicating that higher compaction significantly enhanced mechanical strength. The Eucalyptus grandis-based panel exhibited lower MOR (11.0 ± 2.2 MPa) and MOE (1878 ± 370 MPa) values compared to the sugarcane bagasse particleboard. In contrast, IB was higher in the eucalyptus panel (1.25 ± 0.27 MPa).

The results of thickness swelling (TS) and water absorption (WA) at 2 and 24 h are shown in Fig. 8. WA decreased with increasing board density, regardless of exposure time. According to TS data (Fig. 8A,C), TS after 24 h increased from 94.2 ± 8.6% in the D400 panel to 130.8 ± 3.2% in the D600 panel, followed by a reduction to 85.3 ± 3.6% and 58.4 ± 4.8% in the D700 and D800 panels, respectively. To contextualize these results, the Eucalyptus grandis-based panel, with a density similar to the D700 bagasse board (≈740 kg/m³), exhibited lower WA and TS values at both time points. Specifically, WA after 24 h was 133.7% for eucalyptus compared to 136.0% for bagasse, while TS after 24 h was 58.9% vs. 85.3%, respectively.

The experimental values of MOR, MOE, IB, and TS for the particleboards made from sugarcane bagasse and a BSFL-based adhesive were compared with the requirements of the European standard EN 312 for wood particleboards [42]. This comparison was conducted to assess potential applications of the boards (Table 3). According to this standard (Table 4), the D400 sample did not meet the minimum requirements for particleboards. The D600 sample reached the P1 rating for MOR and MOE but failed the IB requirement, thus not meeting the minimum classification either. The D700 sample showed good bending properties and an IB value of 0.43 ± 0.04 MPa, which would place it in P3; however, due to insufficient water resistance, its classification dropped to P2 (indoor dry environments). Finally, the D800 sample exhibited a TS of 58.4 ± 4.8% at 24 h, exceeding the 10% limit for P7 classification. Despite surpassing the mechanical requirements for P7, its classification was also P2 due to excessive swelling.

Rheological behavior is a critical parameter for non-Newtonian systems such as protein-based adhesives [43]. Viscosity curves are commonly employed to anticipate adhesive performance during industrial operations such as mixing, pumping, or spraying. Moreover, although the use of the Carreau–Yasuda model for protein-based wood adhesives is still limited, it has been widely adopted for describing the flow behavior of protein systems [44]. The pseudoplastic behavior observed for the BSFL-based adhesive in Fig. 2 is consistent with our previous findings [27]. Importantly, the extrapolated viscosity values (0.157–0.101 Pa·s at high shear rates) fall within the recommended range for particleboard production (0.1–0.15 Pa·s) [45], confirming that BSFL adhesives are suitable for spraying applications under typical processing conditions.

Oscillatory rheological tests are essential for characterizing protein-based adhesives because they provide insights into viscoelastic structure without inducing structural breakdown [46]. Determining the LVR ensures that subsequent frequency sweeps or time-dependent measurements are performed under conditions that reflect intrinsic material properties rather than structural damage. The flow point, observed in Fig. 3B at ≈10 Pa for the BSFL-based adhesive, is a particularly relevant parameter, as it defines the transition from solid-like to liquid-like behavior. This benchmark can be used to evaluate the mechanical stability of protein adhesives under external stress and to guide formulation adjustments for improving processing and performance during panel manufacture.

Frequency sweeps provide key insights into the molecular organization of protein-based adhesives, as they reflect the balance between elastic and viscous contributions [46]. The predominance of G′ over G′′ across the frequency range in Fig. 4 indicates that the BSFL-based adhesive behaves as a soft viscoelastic solid, like other physically structured protein systems. The observed shear-thinning trend of η* confirms the presence of a cohesive network capable of dissipating energy under stress while maintaining structural integrity. This behavior is advantageous for particleboard production, where an elastic component is essential to ensure bond stability during pressing and service life.

The 3ITT test is a valuable tool for assessing the structural recovery behavior of viscoelastic materials, particularly relevant for bio-based adhesives subjected to high shear during processing or application. In the context of protein-based adhesives, where the formation and recovery of a physical network governs performance, thixotropic analysis provides insight into both formulation robustness and industrial applicability [43]. The rapid but partial structural recovery observed in Fig. 5 for the BSFL adhesive is advantageous, as it ensures that mechanical strength is regained shortly after application. This behavior is especially beneficial in spraying or spreading processes, where fast recovery minimizes sagging, enhances substrate wetting, and promotes uniform adhesive distribution, ultimately contributing to strong and reliable bonding in particleboard or plywood production.

To better illustrate how the rheological properties of the BSFL adhesive relate to its bonding performance, Fig. 9 presents a simplified scheme of the proposed adhesion mechanism between protein chains and bagasse particles during hot-pressing. In this conceptual model, protein unfolding and partial denaturation under alkaline conditions expose functional groups capable of establishing hydrogen bonds and hydrophobic interactions with the wood surface, while the liquid medium ensures adequate wettability and penetration into pores. Short chains can penetrate into the surface porosity, favoring adsorption and mechanical interlocking, whereas longer chains remain at the interface, contributing to cohesion across particles. These interactions, combined with the consolidation effect of pressing and drying, explain the ability of the BSFL adhesive to form continuous networks and generate adequate internal bond strength in the manufactured panels.

The dimensions of the bagasse particles used in this study (Table 1) were slightly higher than those reported by Brito et al. [47], who observed lengths between 9.2 and 18.78 mm, thicknesses between 0.22 and 0.27 mm, and slenderness ratios ranging from 42 to 69. Also, the morphological distinctions between Eucalyptus grandis and sugarcane bagasse particles are critical, as they are expected to influence the mechanical behavior of the resulting panels. Bagasse particles, with their higher slenderness ratio, provide greater contact area and potential for mechanical interlocking, while eucalyptus particles, being shorter and thicker, tend to adopt a more cuboidal shape, leading to increased void content and reduced surface area available for adhesive bonding. These differences in particle morphology are therefore expected to influence the mechanical properties of the panels.

Surface roughness is a quantifiable measure of surface irregularities, which can significantly influence the interaction between panels and melamine-impregnated paper commonly used as the external face in veneers [48,49]. This coating not only improves the finishing of the panels but also enhances the water resistance of the boards [50]. The decrease in surface roughness with increasing density observed in Table 2 is consistent with previous reports for particleboards [51]. Furthermore, surface roughness has been shown to depend on multiple factors, including particle type and morphology, panel density, and pressing conditions [52,53]. In agreement with these findings, our results suggest that particle morphology plays a key role in surface quality, with particles of higher slenderness ratio leading to more homogeneous and smoother surfaces.

The contrasting mechanical behavior observed in Fig. 7 between bagasse and eucalyptus panels can be attributed to particle morphology. According to the literature, both MOE and MOR are positively influenced by particle length and negatively by particle thickness [54], with slenderness being the most critical factor, exerting a direct positive effect on these properties [55]. In this context, the higher slenderness ratio of bagasse particles (Table 1) is consistent with the superior MOR and MOE values obtained in this study (Fig. 7A,B). Regarding IB, previous reports indicate that this property is influenced by particle geometry at fixed density, with higher values occurring when particles are shorter and thicker, i.e., with a lower slenderness ratio [56]. This trend was also observed in our results, where the eucalyptus panel, composed of shorter and thicker particles, achieved higher internal bonding compared to bagasse panels. These findings reinforce the key role of particle morphology in determining the mechanical performance of particleboards manufactured from alternative lignocellulosic raw materials.

The decrease in WA with increasing density observed in Fig. 8 is consistent with previous findings that denser panels exhibit reduced water diffusion [57]. This trend may also be influenced by the adhesive content, since at fixed resin loading (12 wt.%), denser boards concentrate a higher adhesive amount per unit volume, contributing to reduced WA [58]. The marked increase in TS from D400 to D600 can be attributed to the greater amount of lignocellulosic material present in the D600 panel, which increased water uptake capacity, combined with a porous internal structure that facilitated moisture penetration. In contrast, the reduction in TS from D600 to D800 correlates with enhanced compaction of the board, leading to reduced void content and limited water intrusion. This structural effect was supported by optical microscopy (Fig. 6), which showed fewer interparticle gaps in higher-density panels [59,60]. The comparison with Eucalyptus grandis highlights the dual role of particle morphology and board compaction in determining dimensional stability. Despite a coarser and more porous panel structure, the eucalyptus particles were thicker and less slender than bagasse particles, which may have limited water absorption at the particle level. This explains why eucalyptus boards showed lower WA and TS values than bagasse panels of similar density.

As is evident from these results, the weakest point of the studied particleboards is water resistance, since all samples presented TS values at 24 h considerably higher than those required for use in humid environments (Table 3). This limitation is consistent with other studies on protein-based adhesives, which typically show lower water resistance compared with synthetic resins [61]. Nevertheless, strategies such as incorporating cross-linking agents have been reported to improve the hydrophobicity of protein-based adhesives [62–64]. In addition, surface roughness analysis (Table 2) showed that higher-density boards presented smoother and more homogeneous surfaces, which could favor the adhesion of melamine-impregnated papers with excellent water resistance, often surpassing UF resins [65]. Therefore, by combining protein-based adhesives with hydrophobic modifications or overlay techniques, the D700 and D800 panels could potentially achieve the more demanding EN 312 classifications for use in humid environments. These findings reinforce the potential of bagasse-based boards bonded with natural resins to reach high-performance applications under appropriate modifications.

The mechanical and physical performance of the particleboards produced in this work was compared with several previous studies that employed sugarcane bagasse as the main lignocellulosic raw material (Table 5). Although some formulations reported higher absolute values in flexural or bonding strength, such as the castor oil polyurethane (PU) system by Sugahara et al. [66] or the bone glue-based panel from Islam et al. [25] those panels have a higher density than those studied in this work; also, the results must be contextualized by the adhesive type, board composition, and sustainability profile.

The binderless boards reported by Widyorini et al. [67] achieved IB values close to 0.42 MPa, comparable to our bagasse-based formulation (0.43 MPa), but those boards yet required high pressure and long pressing times. In contrast, our adhesive system based on BSFL protein achieved these results with milder processing conditions and without the need for synthetic additives or chemical pretreatments. Nadhari et al. [69] also researched binderless systems with lower MOR and MOE values.

When benchmarked against UF and phenol urea formaldehyde (PUF) systems, our formulation displayed superior MOR, MOE and IB values. For instance, Magzoub et al. [68] reported 16.9, 2101 and 0.34 MPa for MOR, MOE and IB, respectively, with PUF adhesives, while our bagasse board reached 20.6, 2229 and 0.43 MPa. Similarly, Mendes et al. [70] achieved an internal bond of 0.46 MPa using UF at 8 wt%, although the panel exhibited lower density (672 kg/m³) and no data were provided for flexural strength. Compared to these conventional formulations, the BSFL protein adhesive offers competitive bonding strength while being derived from renewable and circular sources.

The castor oil-based polyurethane PU adhesives presented by Fiorelli et al. [71] and Sugahara et al. [66] yielded similar and strong mechanical properties, with Sugahara’s panel achieving 31 MPa in MOR and an exceptionally high IB of 2.52 MPa. However, the density of this panel was 882 kg/m³, significantly higher than our system (740–745 kg/m³), and the formulation included 60% wood particles. Additionally, these resins still require petrochemical-derived components (e.g., isocyanates), raising concerns regarding volatile organic compounds (VOC) emissions and toxicity during production and use.

The system by Islam et al. [25], based on bone glue, reported very high MOE (4302 MPa) and good MOR (26.22 MPa), though its internal bond was not reported. Although technically effective, animal-based adhesives such as bone glue may face limitations regarding consumer acceptance and traceability under circular bioeconomy frameworks.

In this context, the BSFL protein adhesive used in our study offers a balanced combination of mechanical performance and environmental benefit. It is derived from upcycled organic waste via insect bioconversion, aligned with circular bioeconomy principles and offering an innovative alternative to conventional protein-based adhesives (soy, casein, blood). While water resistance (TS 24 h = 85% for bagasse boards) remains a challenge, our approach presents a viable bio-based route to particleboard production without synthetic resins.