Valorization of Byproducts of Hemp Multipurpose Crop: Short Non-Aligned Bast Fibers as a Source of Nanocellulose

2.2. Production of Nanocellulose from Hemp Fibers

The disentangled hemp fibers (raw fibers) were chopped using a kitchen grinder, and then underwent a number of chemical and mechanical pretreatments before the final mechanical fibrillation step, as detailed in the following; they were never dried during the procedure. After each pretreatment step, samples of fibers were set aside for analysis.

The chopped hemp fibers were soaked in a 10 wt% aqueous sodium hydroxide (NaOH) solution at room temperature for two hours, with mild agitation, then in a 1M aqueous solution of hydrochloric acid (HCl) for one hour at 80 °C, and finally, in a 2 wt% aqueous NaOH solution for one hour at 80 °C. The ratio of liquid to fibers was equal to 1 L of each solution per 20 g of dry chopped hemp fibers. Between each step, the fibers were rinsed with water, with the aid of a 120-mesh sieve, until the pH was in the 6 to 8 range, and were drained of excess water, but not allowed to dry.

The chemically pretreated fibers were diluted to a 10 wt% concentration in demineralized water and processed with an immersion kitchen blender to obtain a finer pulp. The pulp was then further refined in a PFI mill-type laboratory beater (Metrotec SA, Kirchheim, Germany) at 2500 rpm, prior to the treatment with the FiberCare^® cellulase enzyme. The concentration of the enzyme was 300 ECU per gram of dry fibers. The refined fibers, placed in a reactor equipped with an overhead stirrer and a heated water bath, were further diluted to a 2 wt% concentration using a pH 5 acetic acid/sodium acetate trihydrate buffer solution. The temperature was raised to 50 °C, and then the enzymes were added: the fibers were left to stir for 2 h, before raising the temperature to 80 °C for 10 min to deactivate the enzymes. The pulp was then vacuum filtered using a precision woven screening polyamide 6.6 fabric with 1 µm mesh opening (NITEX, Sefar AG, Heiden, Switzerland), and re-dispersed in demineralized water. The filtration procedure was repeated three times in order to reach a neutral pH.

Finally, to obtain the nanocellulose, demineralized water was added to the enzymatically treated pulp, to obtain a solids content of approximately 1.4 g/L. The pulp was defibrillated in a Supermasscolloider ultra-fine friction grinder (Model MKZA6-2, disk model MKG-C 80, Masuko Sangyo Co., Ltd., Kawaguchi, Japan). The clearance gauge was initially set to the 0 position, and the suspension was passed 8 times through the grinder in order to homogenize it. Then, the suspension was passed through the grinder 4 times with the gauge set at −5, 8 times with the gauge set at −10, 1 time with the gauge at −12, and finally 2 times with the gauge at −15. A small sample of the nanocellulose aqueous suspension was dried overnight at 100 °C in order to verify the actual concentration, which was found to be 1.3 wt%.

2.3. Production of Nanopaper

Nanopaper handsheets were prepared using the 1.3 wt% nanocellulose aqueous suspension, as obtained after the ultrafine friction grinding, by filtering through a Rapid-Köthen standard sheet former (Frank-PTI, Birkenau, Germany) equipped with a precision woven screening polyamide 6.6 fabric with 1 µm mesh opening (NITEX, Sefar AG). The sheets were then dried in a dryer at 90 °C for 20 min, between two layers of the polyamide 6.6 fabric. The final thickness was 55 ± 1 µm.

3.1. Evolution of the Dimensions and Morphology of the Fibers during the Pretreatment

A photo of the hemp fibers, chopped and after each step of the chemical pretreatment, showing the color change of the fibers, is shown in Figure 2.

After the first alkali pretreatment, with a 10 wt% NaOH solution for 2 h at room temperature, the fibers had a lighter color than the chopped ones, due to the removal of the non-cellulosic compounds of the cell wall, i.e., hemicellulose and lignin. Accordingly, the alkali solution recovered after the treatment had a dark yellow color. After the acid and the second alkali pretreatment, with 1M HCl and with 2 wt% NaOH, respectively, both performed for 1 h at 80 °C, the fibers still retained a yellowish color. As the corresponding solutions recovered at the end of these treatments also had a pale yellow color, it is inferred that some hemicellulose and lignin that were protected inside the fibers were however further removed during these steps.

The evolution of the dimensions of the fibers during the chemical pretreatment is shown in Figure 3, and Figure 4 shows FESEM images of the surface of the fibers after each pretreatment step. Additional optical microscopy and FESEM images can be found in the open access Zenodo repository [41].

The raw hemp fiber sample was composed of primary and secondary bast fibers with a wide range of dimensions, with widths of few micrometers to several hundred micrometers and lengths up to 150 mm, and a small number of fragments of shives, with widths of up to 2 mm and lengths up to 10–20 mm.

After chopping, the maximum fiber length was reduced to about 40 mm, with a large amount of shorter fibers, so that the median length was about 5 mm; dislocations (defects), appearing as transversal dark lines in the optical micrographs, could be observed. The mechanical action of the mixer blades not only cut the fibers, but also favored the splitting of the bundles of individual fibers. The fiber bundles had sections of irregular shapes, while the individual fibers had more round sections. Fiber widths were mostly in the 1–500 µm range, with a large part of the fibers with widths below 80 µm. Some fragments of shives with widths up to about 1 mm were still visible. The surface of the fiber bundles was rough as the middle lamella still surrounded the individual fibers; a few thin fibers with micron to submicron dimensions departed from the surface of the fiber bundles (Figure 4a).

After the first alkali pretreatment, the fibers still had lengths up to 40 mm, although, with respect to the chopped fibers, the number of shorter fibers increased, with the median value for the length decreasing to 2 mm. The widths of the fibers were similar to those of the chopped fibers. Due to the removal of the middle lamella, the individual fibers composing the bundles became clearly visible, but did not separate much further (Figure 4b), as even high concentrations of NaOH have been reported to not be able to completely remove all the lignin and pectin cementing the bundles [42]. The estimated widths of the individual fibers, 3 μm to 20 μm, were consistent with those reported in the literature [42,43].

The acid pretreatment overall reduced the length of the fibers, likely due to acid hydrolysis, which is known to weaken the fibers at dislocations [44]. The observed lengths were in the 0.1 mm to 16 mm range, with a median length of 0.9 mm. This range of lengths is in line with that reported for hemp individual fibers [43]. The treatment favored the separation of the individual fibers (Figure 4c), as also indicated in other works [23]. As a result, the amount of finer fibers increased, with the median value for fiber width decreasing to about 16 μm, although the upper limit for the observed fiber widths did not change, as not all the bundles were disrupted.

The second alkali pretreatment further loosened the individual fibers, removing non-cellulosic compounds; some thin ribbon-like filaments, probably parts of damaged individual fibers, could be seen protruding from their surfaces (Figure 4d). A slightly higher amount of shorter fibers could be observed, while the observed widths did not substantially change with respect to the previous step.

Refining greatly reduced the length and width of the fibers. The fiber length was in the 400 μm to 10 μm range, with a median value of about 70 μm. The majority of the fibers had widths ranging from submicron to about 20 μm, and the median value was close to 3 μm, and on the surface of the individual fibers, a fibrous structure of nanometric fibrils appeared (Figure 4e,f).

The enzymatic pretreatment did not further significantly change the fiber dimensions; nevertheless, the surface of the fibers was “cleaner”, and a structure of fibrils of nanometric dimensions became well visible (Figure 4g,h).

To summarize, the dimensions of the hemp fibers after each pretreatment step are compared in the box plots of Figure 5. One can conclude that a reduction in the length was obtained gradually throughout the pretreatments, with an abrupt decrease after refining. On the other hand, the width of the fibers did not decrease significantly with the chemical pretreatments, while a significant decrease was obtained with mechanical refining, suggesting that mechanical energy input is needed to separate the individual fibers. The pretreatment with enzymes did not significantly further reduce the fiber lengths and widths.

3.2. Morphology and Dimensions of the Nanocellulose

After the series of pretreatments, the fibers, still of micrometric dimensions, were subjected to mechanical defibrillation in an ultra-fine friction grinder, to obtain nanocellulose. The FESEM observation of the nanocellulose suspension highlighted that nanosized fibrils were present together with larger micron sized fibers (Figure 6). The nanosized fraction, separated by decanting as mentioned in Section 2.4 for the inspection by TEM (Figure 7), was composed of individual nanofibrils with lengths of 100–300 nm and widths of 5–12 nm, and stacks of fibrils with lengths of 380 nm to 3 μm and widths between 20 nm and 200 nm. The dimensions of the individual nanofibrils were comparable to those of the cellulose nanocrystals (CNC) obtained by Luzi et al. from Carmagnola hemp bast fibers by acid hydrolysis and to those obtained by Xu et al. from hemp and flax bast fibers by endoglucanase treatment (Xu et al., 2013; Luzi et al., 2014).

3.3. FTIR Characterization

The FTIR spectra of the raw hemp fibers and of the fibers after each treatment are shown in Figure 8. In the spectrum of the raw fibers, absorption bands that correspond to the stretching and bending vibrations of the main compounds present in lignocellulosic fibers, such as cellulose, hemicellulose and lignin, were identified. The peak assignments are reported in Table 1.

The first alkali pretreatment caused the disappearance of the two sharp peaks at 2917 cm⁻¹ and 2849 cm⁻¹, indicating the removal of fatty acids. Furthermore, the peaks at 1734 cm⁻¹ and 1246 cm⁻¹, as well as at 1261 cm⁻¹ and 1230 cm⁻¹ decreased, confirming that the content of hemicellulose and lignin, respectively, diminished. The peak at 1636 cm⁻¹ assigned to bound water strongly decreased. Peaks at ca. 1104 cm⁻¹, 1053 cm⁻¹ and 1029 cm⁻¹, which can be assigned to cellulose alcohols, became sharper. The broad band between 3700 cm⁻¹ and 3000 cm⁻¹ showed some changes, particularly in the intensities at 3330 cm⁻¹ and 3280 cm⁻¹, which may indicate changes in the type of hydrogen bonds formed within and between the cellulose fibrils.

The following acid and second alkali pretreatments caused a further slight change in the spectra following the same trends, indicating the removal of some residual non-cellulosic compounds, in agreement with the visual observation reported in Section 3.1. After the enzymatic pretreatment and after the defibrillation in the grinder, only some changes in the relative heights of the peaks in the region from 1104 cm⁻¹ to 985 cm⁻¹ were detected, possibly indicating changes in the conformation of the hydroxyl groups of cellulose. The enzyme-pretreated fibers showed a lower intensity of the peak at 3280 cm⁻¹ in comparison to that of the peak at 3330 cm⁻¹, suggesting the disruption of intramolecular hydrogen bonds, as well as a lower intensity of the bands at 1002 cm⁻¹ and 985 cm⁻¹, compared with the neighboring peaks at 1056 cm⁻¹ and 1030 cm⁻¹. The spectrum of the dry nanocellulose on the contrary showed an increased intensity of the peaks at 3280 cm⁻¹ and at 1002 cm⁻¹ and 985 cm⁻¹; the increase in the peak at 3280 cm⁻¹ suggests that more intermolecular hydrogen bonds were formed upon drying.

3.4. X-ray Diffraction

Native cellulose typically consists of two crystalline forms, Iα and Iβ, which can coexist in a single microfibril in various proportions depending on the cellulose source. Cellulose Iα, which is predominant in, e.g., algae and bacterial cellulose, has a triclinic unit cell, containing one cellulose chain. Cellulose Iβ, which is the main component in fibers from higher plants such as hemp, cotton and the like, has a monoclinic unit cell, containing two parallel cellulose chains [55,56,57]. All fiber samples in Figure 9 clearly showed diffraction peaks at 2θ values of 14.8°, 16.5°, 22.5° and 34.5°, corresponding to the (11⎯⎯0), (110), (200) and (004) crystallographic planes of semi-crystalline cellulose Iβ. Only the raw hemp showed an additional peak at 2θ of 25.9°, which could not be attributed to any of the constituents of natural fibers. As the peak disappeared after rinsing the fibers in cold distilled water, it was attributed to the presence of impurities. Cellulose II exhibits characteristic peaks for the (11⎯⎯0), (110) and (020) reflections located at 12.2°, 20° and 22.1°, respectively. While the shoulder around 20° that appeared in the diffractograms of all fiber samples after the first alkali pretreatment may correspond to the superposition of a weak (110) reflection of cellulose II and of (012) and (102) reflections of cellulose Iβ appearing in randomly oriented samples [55,58], a very weak peak at 12.2° was only visible in the nanocellulose pattern. This suggests that only a negligible presence of cellulose II was present in the samples after the alkali treatments. Strong alkali treatments are known to induce a partial transformation of cellulose I to the more stable cellulose II structure, the so-called mercerization. This happens particularly in combination with bleaching treatments [23,25,59,60]. On the other hand, the presence of residual hemicellulose and lignin, which in our case is suggested by FTIR (Section 3.3) and thermogravimetric analysis results (Section 3.5), is thought to inhibit the transformation, possibly by hindering the penetration of alkali in the cellulose fibrils [25].

To estimate the size of the crystallites, we assumed that the lateral crystallite size is the primary contributor to broadening of X-ray peaks. The crystallite size was determined from the broadening of the (200) reflection because it was the most clearly resolved of the three X-ray reflections. The dimension which is measured is in a direction normal to the surface of a crystallite. Table 2 summarizes the positions of the fitted peak and the width of the peak at half height (β), the degree of crystallinity (CI) and the size of the crystallites.

The degree of crystallinity increased abruptly after the first alkali treatment, most probably due to the removal of amorphous materials such as lignin, hemicellulose and waxes. At the same time, the crystallite size increased as well. Then crystallinity slightly increased up to 80% for the final nanocellulose. The progressive decrease in the width at half height values (β) following the chemical pretreatments is reflected in the increase in crystallite sizes. A correlation between crystallite sizes and CI was found, in agreement with other researchers who observed an inverse correlation between X-ray broadening (β) and X-ray fractional crystallinity [61,62,63]. This is consistent with the recent theory that amorphous cellulose concentrates at the surface of crystallites [17]. Smaller crystallites have a larger surface area to volume ratio, and therefore a larger fraction of amorphous cellulose surrounding the crystalline part, leading to lower CI, compared with larger crystallites, which instead have smaller surface area to volume ratios, and therefore lower amounts of amorphous cellulose surrounding the crystalline fraction, resulting in higher CI.

3.5. Thermal Analysis

The results of the thermogravimetric analysis performed on the raw and pretreated hemp fibers and on the nanocellulose are reported in Figure 10. Generally, all the thermogravimetric curves presented the following features: (i) an initial weight loss due to evaporation of residual water, around 100 °C; (ii) between 250 and 300 °C, the thermal decomposition of hemicellulose; (iii) in the 350–400 °C region, the thermal decomposition of cellulose; (iv) from 250 to 600 °C, the thermal decomposition of lignin, which usually happens in a wide range of temperatures, due to the complexity of the lignin structure; (v) a residue at 800 °C of 11–17 wt%.

Comparing in detail the first derivative of the weight curve of the raw and chopped hemp fibers with that of fibers treated with 10 wt% NaOH, one notices in the latter only a very slight decrease in the two shoulders, corresponding to thermal decomposition of hemicellulose and lignin, which appeared at the two sides of the peak attributed to thermal decomposition of cellulose; in addition, the temperature of maximum decomposition rate (T_max) for cellulose slightly increased after the alkali pretreatment. These small changes can be attributed to the removal, after the 10 wt% NaOH pretreatment, of the non-cellulosic compounds, hemicellulose and lignin, at the surface of the fibers. After the acid pretreatment, with 1M HCl, the low temperature weight loss, associated with the residual water present in the fibers, decreased to less than 2%, consistently with the decrease in the corresponding peak in FTIR analysis. The T_max for cellulose increased, while the residual weight at 800 °C decreased to about 11%. These effects are attributed mostly to acid hydrolysis of hemicellulose and cellulose, leading to removal of hemicellulose and to molecular weight decrease for cellulose [64]. The second alkali pretreatment, consistently with what was seen by FTIR analysis, caused only minor changes, in the same direction as the acid pretreatment. After the enzymatic pretreatment, the weight loss at low temperature attributed to residual water increased, and T_max decreased: finer fibers are likely to form quite densely packed structures, hindering water removal during the predrying in the oven and slowing down thermal decomposition. The onset of cellulose thermal decomposition was at a lower temperature, probably due to the reduction in the molecular weight caused by the endoglucanase enzyme action.

The main differences in the thermal decomposition behavior of the dry nanocellulose with respect to that of the enzymatically treated sample are an initial slow and almost linear weight loss (2–3%) until about 190 °C, which can be attributed to residual non-freezing water, and the increase both in the onset temperature for cellulose thermal decomposition and in the residual weight at 800 °C, due to the formation of a denser structure upon predrying in the oven.

The DSC thermogram of the dry nanocellulose (Figure S1) during the first heating run showed a broad endotherm peak between 0 °C and 200 °C, centered around 90–95 °C, which was not present in the second heating run. The presence of this peak, attributed to the evaporation of non-freezing water, is consistent with what seen in thermogravimetric analysis. No other transitions appeared in the observed temperature range.

3.6. Compatibility of Nanocellulose with Solvents

As nanocellulose is an efficient reinforcing agent in polymeric composites, where the matrix usually has a low polarity, the compatibility with different solvents was tested. Samples of the suspension collected after the ultrafine friction grinding were diluted by adding either distilled water, ethanol or acetone, to obtain a solids content of 0.25 wt%. The obtained suspensions were left to stand at room temperature for 12 days, as shown in Figure 11. The suspension in water was stable, showing no settling. In water/ethanol and water/acetone, there was some settling of the cellulose fibrils; while after 1 h only the water/acetone solution showed some settling, after 12 days a large settling was visible both in water/ethanol and in water/acetone. The content of ethanol and acetone in the solutions was close to 82 wt%, and the density of the solutions was estimated to be close to 840 kg/m³ [65,66]. The settling in the organic solvent mixtures may be attributed to the lower density with respect to water, which allows a higher settling velocity; destabilization of the fibrils in the presence of the organic solvents may also be ascribed to their different polar/dispersive characters.

The authors thank the NanoBio-ICMG Platform (FR 2607, Grenoble) for granting access to the Electron Microscopy facility.