Effect of thermal modification of wood particles for wood-PLA composites on properties of filaments, 3D-printed parts and injection moulded parts

Kiln-dried beech (Fagus sylvatica L.) wood boards were collected and cut into smaller lamellas (900 × 200 × 25) mm. They were dried to an absolute dry state for 24 h at 103 ± 2 °C and weighed to later calculate the mass loss during thermal modification. Thermal modification was carried out in a vacuum modification chamber (Kambič d.o.o., Semič, Slovenija) at 200 °C according to the Silvapro method (Pohleven and Rep 2001). The process of thermal modification is divided into the following three stages: temperature increase (heating), modification process in vacuum (constant temperature 200 °C for 3 h) and lowering of the temperature (cooling). It should also be noted that the degree of modification is determined by the degree of mass loss during thermal modification, and by this parameter, the severity of modification could be compared to other types of thermal modifications. After thermal modification, the wood boards were weighed again to determine mass loss during modification. Modified bords were subsequently cut into small cubes [(20 × 20 × 20) mm]. The cubes were also cut from non-thermally modified beech wood boards. All the cubes were ground in a laboratory cutting mill Retsch SM 2000 (Retsch, Haan, Germany), first with a 1 mm sieve and in a second step with a 0.25 mm sieve.

The transparent thermoplastic polymer used was PLA Ingeo™ 2003D (NatureWorks, Blair, Nebraska, USA) in granular form. The melt temperature of the material is 210 °C, tensile strength at break 53 MPa (ASTM D882), specific gravity 1.24 g cm⁻³ (ASTM D792), and melt flow index 6 g/10 min at 210 °C (ASTM D1238).

Wood particles were sieved through a 237 μm sieve. All materials were dried in a laboratory oven (SP-120 Easy, Kambič, Semič, Slovenia) prior to extrusion/compounding: wood for 12 h at 103 °C and PLA for 4 h at 70 °C.

The filament was produced on a co-rotating twin-screw extruder Haake PolyLab OS, Rheo Drive 7 (Thermo Fisher Scientific, Waltham, Massachusetts, USA), with a screw diameter of 30 mm and an L:D ratio of 20:1. The mixture of polymer and wood particles was fed into the main hopper of the extruder under vacuum using a gravimetric metering system. The filament was produced in a two-stage process to ensure better homogeneity. In the first step, the wood was mixed with a polymer matrix to form pellets, which were extruded into filaments in the second step. The screw speed was 20 min⁻¹ for compounding and 15 min⁻¹ for filament production. The temperature ranged from 160 to 180 °C (from hopper to die) for compounding and from 175 to 180 °C for filament production. Filaments were extruded through a round die with a diameter of 3.0 mm and withdrawn from a die with a filament spooler at a winding speed of 3.0 m min⁻¹. The winding/spooling speed had to be adjusted to the extrusion speed to achieve the desired filament diameter. The diameter of the filament was measured continuously in one direction during production using a laser measurement system. The fabricated filaments had a diameter of 1.50 ± 0.15 mm. A total of 5 filaments with different compositions were produced, which are listed in Table 1.

Dog-bone-shaped tensile specimens (length 75 mm) of the wood-PLA blend were created via injection moulding. Extruded filaments were directly pelletised to 1 mm length and injected into an ISO 527–2, 1BA mould, using a pneumatic piston injection moulding system Haake MiniJet Pro (Thermo Fisher Scientific, Waltham, Massachusetts, USA). The cylinder temperature was 180 °C, mould temperature 75 °C, pressure 90 MPa, and press time 10 s. Injection moulded specimens were used only for water contact measurements and for tensile strength measurements for comparison with 3D-printed specimens.

Prepared filaments were used to print dog-bone shaped test specimens (geometry according to a modified standard ISO 527–2: 1996 (Plastics—Determination of tensile properties)) on a Creality CR-10-V3 (Creality 3D Technology Co., Ltd, Shenzhen, China) 3D printer with a direct extruder. The 3D printer had a nozzle diameter of 1 mm (inner dimension). STL (Standard Tessellation Language) models were sliced and prepared for 3D printing in Cura software (Ultimaker, Utrecht, The Netherlands). 3D printing parameters were determined empirically from previous studies’ results. The nozzle with a 1 mm diameter was selected to minimise clogging during printing, and low printing speed was used to ensure proper melting of filament in the hot end. The infill density was set to 100%, the layer height was 0.3 mm with printing lines alternating at an angle of 45° (one layer + 45°, next layer – 45°), the print speed was 25 mm/s, and the nozzle temperature was 220 °C. The walls and top/bottom thickness were set to 0.6 mm (two printing lines), but since the infill was set to 100%, the whole volume of printed samples was filled with material. The printing bed temperature was 50 °C. Since filament diameters varied between wood content and type of wood particles, the material flow was individually adapted to each filament diameter before printing. This was necessary to compensate for the different filament diameters to avoid under-extrusion. There were evident differences in filament diameter between composites but less so in the same filament. The amount of filament (its length) used to print a specimen was, on average, not different.

Particle size and particle size distribution are key factors in composite materials. These parameters determine the mechanical properties of the composite/filament and provide information about its origin and history. Particle-size and particle-size distribution using a particle analyser with computer-controlled laser alignment showed a consequential difference between particles from beech and TM beech ground with a 250 μm sieve. The mean and median size of the particles varied between non-modified and modified beech. The median for beech was 193 µm and for TM 104 µm. The mean size was 242 ± 210 µm for beech and 163 ± 166 µm for TM beech (Fig. 1). The median and mean values of particles of TM beech are almost half lower than those of non-treated beech, which means after wood modification, smaller particles are obtained with the same wood preparation (grinding). This is mainly due to the more brittle nature of TM wood (Hill 2006). Moreover, variance and standard deviation are considerably lower for particles from TM beech wood, confirming our hypothesis that the modified particles have a more uniform size.

Although there are significant differences in the fracture behaviour of wood in different directions, when grinding, it is difficult to take advantage of this material property of wood because the wood particles are usually oriented randomly in the grinding zone.

The results of sieve analysis show that the percentage of particles smaller than 0.24 mm is 22.06 (wt%) for beech wood ground to 1.00 mm and 39.52 (wt%) for TM beech wood ground to 1.00 mm (Fig. 2). However, it should be noted that in Fig. 2, the results of the sieve analysis are given in mass fractions, while in Fig. 1, they are given in volume fractions. TM wood is ground into smaller particles because it is more fragile and brittle. In the temperature range of 150 °C to 200 °C, modification leads to damage to the cell wall (Tumuluru et al. 2011).

As the size of the sieve aperture size decreases, the fineness of particles increases in accordance with the specific surface area and the slenderness of the particles. In our research, we have attempted to justify and categorise the use of less ground larger wood particles (used sieve with 237 μm openings). Commercial filaments are made with wood flour with a mean size of particles of about 100 μm, with even more energy spent on grinding, which could be one of the reasons that commercial wood-PLA filaments are not cheaper than pure PLA filaments despite using low-cost wood residues from the woodworking industry.

The size and, in particular, the L:D ratio significantly affects the properties of the obtained composite. The size of the particles for 3D printing applications is limited with the printer nozzle, and larger particles lead to clogging and uneven extrusion. Smaller particles and short fibres can more easily flow through the printer nozzle but do not strengthen the composites as much as particles with a higher L:D ratio. Longer fibres can increase tensile strength and modulus of elasticity at the expense of toughness and ductility (Ning et al. 2015).

Image analysis is a very appropriate method for the characterisation of wood particles, as this method was used to provide information about their geometry. Image analysis was made on SEM images of wood particles sieved to 0.25 mm. For each photo, 10 measurements were made. The average aspect ratio was 4.5 ± 1.4 for beech and 3.9 ± 2.9 for TM beech. This shows that TM particles have a slightly smaller aspect ratio. However, the average beech wood particle is significantly longer than the TM beech wood particle. The average TM particles were about 40% shorter than the average length of beech particles. The aspect ratio between the two types of particles is quite close; moreover, the L:D ratio of TM wood particles is more variable, ranging from 10 to 2.

The aspect ratio depends to some extent on grinding methods, moisture content, and wood species, but the general trend is a decreasing aspect ratio with decreasing particle size (Karinkanta et al. 2018). Stark and Rowlands (2003) concluded that for wood-plastic composites, the aspect ratio rather than the particle size has the greatest influence on strength and stiffness. A high aspect ratio improves stress transfer from the matrix to the wood particles. In general, the aspect ratio in wood particles varies between 1 and 5.

Huang et al. (2011) found that as the aspect ratio increased, the tensile modulus of the plastic-wood composites increased by 28.4%, and at smaller sizes, the impact energy of the composites increased by 35.5%. A recent study by Golmakani et al. (2021) showed that reducing the aspect ratio increases the effect of filler reinforcement in composites and improves the stress distribution in the specimens.

According to the length and aspect ratio of the prepared particles, we would expect that wood-plastic composites with non-modified particles would exhibit better mechanical properties.

The density of prepared filaments varied due to their different compositions. The filament made of pure PLA showed the highest envelope density measured with a gas pycnometer (1.203 g/cm³). This value is still lower than the manufacturer's specifications for the density of PLA (1.24 g/cm³). The reason for this could be the manufacturing process of the filament and voids in the filament. With higher wood content with 10% and 20% beech wood content, the density decreased to 0.8270 g/cm³ and 0.6530 g/cm³ due to adding the wood with lower density than PLA, respectively. Specimens with 10% and 20% addition of TM wood had 1.1095 g/cm³ and 1.1616 g/cm³, respectively, which corresponds to a different trend (Table 2). The reason for this could be the filament production process, as PLA filaments with beech wood addition were very porous and irregularly shaped, which is visible to the naked eye.

The filaments and 3D-printed specimens made from neat PLA were smooth and had minimal voids, but with increasing wood content, they became rougher, more porous and had visible clusters of wood particles. Porosity was expected to be lower for filaments with added TM beech wood than for filaments with added non-treated beech wood (Fig. 3).

The porosity of the material also reflects on the surface and its roughness (Fig. 3). The surface roughness of filaments was evaluated via CLSM, whereas a detailed microstructure of the surfaces and cross sections was analysed with an SEM. The surface roughness parameter (Sa) is commonly used to evaluate surface roughness and is defined as the arithmetic mean of the absolute ordinate values within a defined range.

The surface roughness of filaments increased with the wood ratio. As expected, filaments with TM wood exhibit lower roughness (14% smaller roughness for a 10% ratio and 28% for a 20% particle ratio in composite) compared to filaments with non-treated wood particles at the same ratio of particles in the composite. This is partly because of smaller TM particles and smaller porosity.

Thermally modified wood is more brittle (Kubojima et al. 2004); consequently, brittle fractures occur under stress. Phuong et al. (2007) studied the effect of thermal modification on the brittleness of Styrax tonkinensis wood and found that the degradation of amorphous polysaccharides during thermal modification was the main reason for the increase in brittleness. In addition to the degradation of hemicelluloses, brittleness is also affected by a change in the degree of crystallinity (Awoyemi and Jones 2011). This higher brittleness also reflects on particle shapes and sizes obtained by grinding and cutting.

Wood particles of beech wood (non-modified and thermally modified) ground with sieve sizes 1 mm and 0.25 mm were observed. TM beech wood particles have more versatile dimensions and shapes, while the beech wood particles are more uniform in size and shape (Fig. 4).

Specimens of four different filaments were cut in the transverse direction and observed.

The cross-section of a PLA filament with 10% beech wood is very porous and irregularly shaped (the cross-section is ellipsoid), whereas the cross-section of a PLA filament with 10% TM beech wood is much less porous and whose cross-section is round (Fig. 5).

A cross-section of a PLA filament with 20% beech wood, as expected, is even more porous than a filament with 10% added beech wood. A cross-section of a PLA filament with 20% TM beech wood is much more homogeneous and ‘compact’ than a PLA filament with natural beech wood. Part of the porosity arises during filament production due to the lack of mixing during the production of the wood-based filament and due to the lack of melting and blending pressure during the FFF extrusion process (Guessasma et al. 2019; Le Duigou et al. 2016).

Wood cells and the penetration of the PLA polymer into the cell lumina is much more pronounced in TM beech wood (Fig. 6). When wood is thermally modified, the hemicelluloses, which contain most of the hydroxyl groups, are degraded, and the wood becomes less polar and, consequently, more compatible with the PLA polymer.

The strong mechanical interlocking of the polymer in the lumen of the wood cells is evident in the SEM micrographs above. We assume that the mechanical properties of the wood-plastic filaments and the 3D-printed specimens printed from these filaments are also better than those of filaments with the addition of non-treated beech wood. From the SEM figures, the TM beech wood particles have very different dimensions and shapes, while the beech wood particles are more uniform in size and shape. This is due to the greater fragility of the TM wood, which is favoured by the degradation of hemicelluloses and the altered degree of crystallinity after the modification process.

Contact angle (CA) was measured on injection-moulded specimens. The water CA was the smallest for the injection moulded PLA specimens (62.3°), followed by specimens TM10 (77.1°) and TM20 (75.6°). As expected, it is smaller on specimens with the addition of TM beech wood compared to specimens with the addition of non-treated wood. This is mainly due to the fact that TM wood is less polar and, therefore, interlocks more with the PLA matrix. The difference between B10 (78.8°) and B20 (78.6°) was insignificant.

The mechanical properties of composites depend on many parameters, such as the properties of the polymer matrix, filler properties, interfacial adhesion, and morphology (Fu et al. 2008). If parts are produced by the FFF 3D printing process, the mechanical properties are also strongly dependent on the 3D printing parameters (Spoerk et al. 2020). The strength of a material is defined as the maximum stress that the material can withstand under tensile loading. In micro- and nanoparticulate composites, this depends on the effectiveness of stress transfer between the matrix and fillers. Factors such as particle size, interfacial strength between particle and matrix, and particle loading significantly affect the strength of the composite material (Fu et al. 2008).

A clear trend exists between injection moulded and 3D-printed specimens (Fig. 7). 3D-printed specimens made of PLA with the addition of TM beech wood have higher tensile strength than specimens with the addition of non-treated beech wood. As already mentioned, TM filaments have a rounder cross-section and fewer voids. Voids and porosity contribute greatly to the reduction of mechanical stiffness and strength (Oztan et al. 2019), as cracks can form in them. 3D-printed biocomposites could have a microstructure with relatively high porosity (even around 20%), which leads not only to damage mechanisms but also to high and rapid water absorption and swelling (Le Duigou et al. 2016). One reason for the improper mixing of the wood particles with the filament matrix is the clogging of the nozzle, which leads to resistance of the heat flow, resulting in the formation of voids and, thus, poor strength properties (Chawla et al. 2020). Filaments with higher wood content did not flow evenly through the nozzle and often clogged it, resulting in poorly printed parts that were not evenly filled with material. As particles are deposited on the surface, the nozzle cross-section decreases, and the pressure drop increases. This leads to a continuous reduction of the volume flow. Dendrites can form, which change the local flow field and break off due to particle collisions (Beran et al. 2018). There were also notable differences in filament diameter between different composite compositions that could alter the amount of material extruded into the printed parts and cause uneven flow and deposition of material. Therefore, the filament diameter was measured before each print and the flow rate was individually adjusted before printing to the actual diameter of the filament, which varied depending on the wood content and type of wood particles. This flow correction should minimise the effect of varying diameters on 3D-printed parts properties. The contact area between the printed lines and layers, and consequently the strength of the printed part, is lower for 3D-printed parts. However, it should also be noted that fewer 3D-printed specimens were tested than injection-moulded specimens due to problems with nozzle clogging during 3D printing.

The wood particles used had a relatively small aspect ratio and could not contribute to the mechanical properties to the extent that longer fibres/particles would. The nozzle diameter limits the size of wood particles used, and if larger particles were used, clogging would occur more frequently. As concluded in a previous study by Dominkovics et al. (2007), larger particles act as stressors and create sites for the onset of cracking, causing the composite to break more easily. Moreover, fine particles have a larger specific surface area due to their greater abundance. As a result, they are more evenly distributed, and the compatibility between the particles and the additional/underlying material is greater.

Corresponding results were found in a study by Ecker et al. (2019), who concluded that 3D-printed specimens had significantly poorer mechanical properties than injection-moulded specimens.

It is considered that agglomerates of wood particles in the composites can fail catastrophically under loading and thus represent zones of potential stress concentration. Previous research has shown that wood content of 20% (w/w) or more can lead to increased interaction or agglomeration of wood particles, which increases the effects of stress concentration zones in the final composite material (Petinakis et al. 2009).

The process-induced voids were evident in all computed tomography (CT) scans (Fig. 8). As evident and expected to some degree, the voids were largest in the filaments with beech wood addition.

Specimen B20 has the biggest pores and also the largest porosity, followed by specimen B10. The pores in specimens with added TM wood are much smaller in all directions, and the overall porosity of specimens TM20 and TM10 is much lower compared to specimens B20 and B10. Thermal modification of wood additive leads to a decrease in the pore size and the total porosity of the specimen compared to non-treated beech wood additive. These results are in line with the porosity determined using pycnometry (Fig. 3). Voids are elongated in the direction of extrusion in all specimens and are interconnected to some extent in specimens B20 and B10. It is not surprising, but interesting, that the voids tend to form at the interface between the wood and the polymer matrix, which is also consistent with the adhesion between wood and polymer mentioned earlier. The surface morphology of the filaments is also strongly influenced by the type and amount of the additive. In general, it can be said that the amount of the additive has an influence on the roughness of the filament as filaments with 10% added wood were smoother compared to filaments with 20% added wood, but the influence of thermal modification was most profound in this respect. The observed effect of TM on the wood particles is described in detail in Sect. 3.6 (SEM). The TM10 specimen is the smoothest and has the roundest cross-section among tested materials. This is consistent with the smaller particle size of the TM filler in general and the roughness measurements described above (Fig. 3). The distribution of the wood particles over the specimen cross-section is most homogeneous with TM specimens, specifically with the TM10 specimen. When non-treated beech wood is used as an additive, the particles tend to aggregate less. This is due to better mixing and lower interfacial energy of TM wood and polymer.

We believe that pores observed in the B20 and B10 specimens are problematic with regard to printing, as they can lead to porosity in the printed parts but can also cause problems in the printing process itself (filament ripping, poor adhesion between layers, etc.). The µCT method used in this study is suitable for observing wood-plastic hybrid materials, but the contrast between the two phases is not very pronounced; therefore, detailed numerical analysis of porosity or particle distribution is not possible. For similar work in the future, we recommend the use of contrast agents that allow easier segmentation and better analysis.