Morphology and rheology of cellulose nanofibrils derived from mixtures of pulp fibres and papermaking fines

Samples used in the present study were never-dried bleached softwood sulphite pulp (F0) and never dried bleached sulphite fines (F100) kindly provided by a project consortium. Solid contents as received were 12 and 20 %, respectively. Both samples were ash-free.

To obtain pulp samples consisting purely of long fibres without fines, the original pulp sample was treated as follows: Its solid content was first brought to 1 % by dilution with tap water and the pulp was stirred for 10 min. It was then fractionated using a lab pressure screen implemented at Graz University of Technology, at a rotor speed of 17 m s⁻¹, a flow rate of 12 l min⁻¹ and an overpressure of about 1 bar. The screen had a hole diameter of 100 µm. The system was operated at 22 °C. The inlet flow was divided into two outlet flows: the reject-flow, containing the washed pulp (free of fines) having a solid content of approximately 2 %, and the accept-flow, containing only fines with a solid content of around 0.02 %.

Whereas the solid content of the reject flow suits the needs of the present study, the solid content of the accept-flow is very low. The production of a sufficient amount of pure fines to perform the planned sample preparation (fibrillation) and measurements using this pressure-screen method is extremely time consuming. For these reasons, it was decided to use a thicker fines suspension available at another sampling point in the paper machine circulation system to perform the experiments described in this paper.

Samples as received were brought to a solid content of 0.5 % and stored at 5 °C for a few days to ensure complete swelling of the fibres. These samples were then used to produce mixtures with different fines/long fibre contents (compare Table 1). In the next step the different mixtures were ground using a Masuko MKCA6-2 J CE supermasscolloider (Masuko^®, Japan) at 1500 rpm and a gap clearance of −50 µm. This step was repeated 5 times. Samples were subsequently homogenised with an APV1000 (SPX^®, USA) homogeniser. To avoid getting the fibres stuck in the homogeniser, the pressure was gradually increased: In the first cycle, fibres were passed through the homogeniser without applying any pressure. In the second cycle a pressure of 100 bar and in the third one a pressure of 400 bar was applied. Homogenizing pressure in the following passes was held constant at 800 bar. The first pass at 800 bar was considered as pass number one in the graphs presented in the “Results and Discussion” section. Samples (around 50 ml) were taken from the suspension after different numbers of homogenisation passes and stored in a refrigerator until use.

A few millilitres of each sample taken during the homogenisation were poured in an aluminium cup, weighed and placed into an oven at 103 °C. After complete evaporation of the water, the dry samples were taken out of the oven and placed in a desiccator until they reached room temperature. Their mass was then recorded again, and the solid content was calculated as follows:where s is the solid content in  %, m _wet and m _dry are the mass of the wet and of the oven-dried sample in [g], respectively.

Previous studies have shown that solid content has a significant influence on viscosity (Silveira et al. 2002; Derakhshandeh et al. 2011). In the present study, a systematic increase in solid content after each homogenisation cycle due to evaporation of water was noticed (compare S1 in supplementary data) and had to be compensated.

The target solid content was set (0.5 % for the first, 1 % for the second measurement series). Samples with a solid content higher than the target were diluted with the appropriate amount of water. Samples with a lower solid content were centrifuged for 10 min at 10,000 rpm (Z36HK, Hermle Labortechnik GmbH). The excess amount of water was removed with a syringe and the fibres were resuspended using a laboratory vortex mixer.

All measurements were performed with a cone-and-plate rheometer (CS50, Bohlin Instruments) equipped with a thermostat set to 20 °C. The cone diameter was 4 cm, the cone angle 4°, and the truncation gap 150 µm. 1.5 ml of the fibre suspension was poured on the plate with a syringe. Measurements at constant shear rate (20 and 200 s⁻¹) were performed for each sample. Before starting the measurement, equilibration at the same shear rate was performed for 10 s. A cycle of averaging for 10 s followed by a measurement pause of 3 s, the cone still rotating at the given shear rate, was then performed 10 times. The viscosity value used for further calculations was determined by discarding the first 2 measurement cycles and averaging the individual values obtained in the 8 other cycles. As each individual sample could be measured within a short time (140 s), evaporation was considered negligible and no solvent trap was used.

A L&W Fiber Tester Plus (Lorentzen & Wettre) having a resolution of 3.3 µm/pixel was used to measure the fibre length distributions of the different samples. Furthermore the fines content [according to the standard implemented in the device, ISO 16065-2 (ISO 2014)] was evaluated. For each measurement, approximately 0.1 g of dry fibres diluted in 150 ml of deionised water was used. The measurement was stopped after 72,000 particles were detected. Two replicates were measured for each sample, followed by a washing step with distilled water after each measurement.

Samples were diluted to approximately 0.001 %. Two drops were given on a mica plate and dried at 103 °C for 15 min. This step allowed obtaining single fibres sticking on the mica surface. The samples were imaged with an AFM (Dimension Icon, Bruker) using an OTESPA cantilever. Gwyddion 2.42 was used to process the output files from the AFM.

The influence of the number of homogenisation passes on viscosity is shown in Fig. 1. From the figure it is apparent that no results could be obtained for the first homogenisation passes if the initial suspension was containing a low amount of fines. In these cases, clearly visible millimetre-sized fibre flocs were expelled from the side of the cone-and-plate system, sometimes causing the rotation of the cone to slow down or even completely stop. When this happened, the measurement was aborted and it was not taken into account. For fines contents of 100, 75, or 50 %, the viscosity increases with the number of homogenisation passes, until a plateau value is reached after approximately 20 cycles. This value is in accordance with the number of passes used by our group to obtain nanocellulose from pulp (Veigel et al. 2012; Gindl-Altmutter et al. 2015). For fines contents of 25 and 0 % (F25 and F0), an increase in viscosity could not be observed because of fibre flocculation. Once these samples were sufficiently homogenised, their viscosity could be measured. The values obtained are oscillating between 200 and 260 mPa s. No reason could be found for this behaviour. As for the other samples, a plateauing was also assumed in this case. However, due to the low amount of samples measured, this cannot be clearly stated here.

The above mentioned behaviour (increase in viscosity and subsequent stagnation after 20 passes) was reproducible for all shear rates and solid contents used in the present study (raw data from the measurements is available in supplementary information S3). Figure 2 shows the influence of the long fibre content, shear rate and solid content on the value of the viscosity plateau. To determine this value, viscosity values measured for the last five data points of Fig. 1 were averaged for each sample type. As it was expected, a higher solid content causes a higher viscosity, whereas a higher shear rate causes a lower viscosity. The latter is due to the shear-thinning behaviour of fibre suspensions, which is well known and has already been studied for several types of fibres (Mallick 2008; Derakhshandeh et al. 2011; Iotti et al. 2011; Rezayati Charani et al. 2013; Naderi et al. 2014). For all combinations of shear rate to solid content, the average plateau value first increases slightly and then steadily decreases when the fines content exceeds 25 %. This result is the exact opposite to what has been found with similar measurements on hard sphere suspensions, where a viscosity minimum was observed at a content of 25 % of small spheres (Parkinson et al. 1970). This indicates that the long fibres, because of higher fibril length, are able to form network structures having a high resistance to flow. The addition of fibres with shorter fibrils will break this network structure and decrease the interactions in the flow, resulting in a lower viscosity.

Numerical models have already been established to calculate the rheology of rigid, rod-like particle suspensions (Powell 1991). However, in the case of natural fibres, flexibility has to be taken into account and these models are no longer valid. Moreover, the surface chemistry of natural fibres is very heterogeneous (contrarily to e.g. carbon or glass rods). For fibres subjected to homogenisation, even more complex interactions due to fibrillation of the fibres on the nanometre scale have to be considered. All these parameters make it very difficult to define and quantify the interactions occurring in the suspension subjected to flow. The high complexity of this topic could explain why models of the flow behaviour of highly fibrillated fibres are often valid only under given conditions, e.g. in a defined shear rate range (Puisto et al. 2012).

The weight-weighted and length-weighted length distributions of unprocessed and homogenised samples F0, F50 and F100 as measured with the L&W Fiber Tester Plus can be seen in Fig. 3. As expected, fibre length distribution changes gradually during the homogenisation process. The images taken with the camera of the Fibertester and shown in Fig. 4 support this observation. For the full set of samples including all mixtures of fibres and fines as well as several grinding and homogenisation steps, the reader is referred to supplementary data S2.

After 10 passes in the homogeniser, a differentiation between the different samples from the images is impossible due to the insufficient resolution of the device. It should be mentioned that the grinding pre-treatment influences the mean fibre length as well as the fines content, as shown in Table 2. This table also shows that fines could not be totally removed from the sample F0 and are accounting for 3.4 % of the total fibre length.

The length-weighted fines content is shown in Fig. 5. The fines content of all samples is converging towards 100 %. In the first homogenisation passes, the fines content of F100 decreases slightly, which means that the relative total length of fibres longer than 200 µm increases. At a first glance, this result may seem contradictory. It can however be quite easily understood if one considers the fibrillation mechanism in the homogenizer, where fibre shortening and fibrillation occur simultaneously (Turbak et al. 1983; Gharehkhani et al. 2015). A long fibre can be shortened several times before it is considered as fine material. In the first homogenisation passes, long fibres will be separated into several shorter fibres which still will not be classified as fines. Depending on the amount and the length of fibres longer than 200 µm initially present in the suspension, this may cause the relative total length of long fibres to increase.

The mean fibre length (length-weighted) is shown in Fig. 6. It is converging towards a limit value of around 50 µm for all samples. For the same reason as mentioned above, the mean length is constant or even slightly increasing in the first homogenisation passes for sample F100. Moreover, all samples are close to the limit value after 20 homogenisation passes, independently of the mean fibre length before the homogenisation. This confirms the conclusions drawn from the viscosity measurement.

The mean fibre length and the fines content followed similar trends when the experimental data was mass-weighted, surface-weighted, volume-weighted, or arithmetically weighted (data not shown).

As AFM is a technique allowing investigations at a very local scale, samples containing only one type of fibres (F0: only long fibres, F100: only short fibres) were chosen for imaging. Typical images obtained by scanning these samples are shown in Fig. 7. After 15 homogenisation passes, long fibres are already highly fibrillated (Fig. 7c), whereas the sample derived from fines does not show a strong fibrillation (Fig. 7a). The difference in the degree of fibrillation is less pronounced after 35 homogenisation passes, where a similar fibril diameter was measured for both samples (Fig. 7 b, d). Samples derived from fines (F100) showed a much lower viscosity than those obtained from pulp (F0) (Figs. 1, 2). One can conclude that the final fibril length and the degree of fibrillation are much higher for F0 compared to F100. This supports the conclusions drawn from the viscosity measurements and the size characterizations performed with the L&W Fiber Tester Plus.