Diafiltration

Diafiltration was mainly employed to remove impurities like lignin and small components from high molar mass GGMs. Its effect on the contents of the 30 kDa membrane concentrates from SWaH2,SWaH4and SWaH5solutions was investigated. Generally, the permeabilities behaved similarly during the DF of these concentrates.Fig. 5.1.7presents the permeability during the DF of the 30 kDa concentrate fromSWaH5. The dia-volume in the first DF step was two times its volume in the second and third steps, so that the permeability values during the first DF (more diluted) were higher than during the further DFs (Fig. 5.1.7). The viscosity of the diafiltered fraction increased by 34% from 5.6 to 7.5 cP at 65 °C after the three DF stages. This could be explained by the fact that the average molar mass of the concentrated GGMs increased by the removal of low molar mass compounds in DF. This purification enhanced the steric interactions between the high molar mass macromolecules, resulting in high viscosity (Kök et al. 1999). As DF proceeded, an increase in the concentration of high molar mass GGMs might have led to formation of a gel-like structure on the membrane surface. Such gelation behaviour is strongly associated with the increase of viscosity (Richardson and Norton, 1998). These phenomena probably caused lower permeabilities in the second and third diafiltration stages, although the amount of dissolved solids was even 50% lower after the third diafltration stage.

Compared with the former UF step, the initial permeability of the 30 kDa concentrate fromSWaH5

during DF was much lower, about seven times, than during the UF ofSWaH5.The higher content of GGMs in the DF feed (SWaH₅ concentrate, ~ 60 g/L) than in the originalSWaH₅(11 g/L) could be the reason for the lower permeability during DF. The viscosity of the DF feed was also 10 times higher than the water viscosity at 65 °C, which decreased the initial permeability

significantly. Moreover, the fouling of the membrane during DF (53%) was higher than during UF (~30%). However, pure water permeability was recovered by ~ 90% with alkaline cleaning.

Figure 5.1.7 Permeability of the GGM concentrate fraction (30 kDa membrane) in three diafiltration steps ofSWaH5 (Pressure 1-3 bar, temperature 65 °C, flat sheet membrane module, cross-flow velocity = 2 m/s).

In general, a narrower molar mass distribution (lower polydispersity) could be observed in the concentrated fractions after DF. The increase in the molar mass with DF was due to the removal of low molar mass GGMs. This removal caused a decrease in the TDS including the content of the total GGMs. For example when the resin-treated 30 kDa concentrate (fromSWaH4) was subjected to DF(2 dia-volumes), the average molar mass increased from 17 to 22 kDa, and the polydispersity decreased from 1.5 to 1.3. During this DF, the reduction in TDS was about 40%. Moreover, the average molar mass of the DF permeate (~8 kDa) was higher than that of the UF permeate (~3.5 kDa). This difference in the average molar mass of the permeates could confirm the higher possibility to form a gel layer by big macromolecules, resulting in lower permeability during DF than UF.

The effect of DF on the content of the 30 kDa concentrated fraction fromSWaH₂is presented in Table 5.1.5and Paper I.The results showed DF-facilitated separation of xylans from high molar mass GGMs, where their content in the concentrate fraction declined by 80 %. Moreover, about 30 % of lignin, 55 % of uronic acids, 95 % of monosaccharides, and 11 % of turbidity were removed from this fraction during DF. It could be assumed that the uronic acids were removed with DF, including the free uronic acids as well as the ones are originally attached to either the hemicellulose and/or pectin chains. In general, the direct UV absorbance in the concentrate fractions decreased by 20-40 %. The overlapping of the GGMs and lignin molar mass distributions, and the fact that most of the free low molar mass lignin had already been removed in the concentration filtration step, could be the reasons for the limited removal of lignin by DF. Based on previous observations, DF enabled partial separation of various impurities from GGMs. It thus improved the purity of the hemicelluloses having the highest molar mass, even though the overall hemicellulose purity decreased by ~2%. Andersson et al. (2007) report that DF increased hemicellulose purity from 57 to 77 % with 1 kDa PVDF membranes. Higher purity values of GGMs (> 90%) have been reported when using ethanol precipitation and size-exclusion chromatography (Willför et al. 2003 a, b; Andersson et al. 2007), but difficulties related to high ethanol consumption or high expenses make DF more suitable and cost-efficient for the purification of hemicelluloses in a large scale. Several studies have shown that UF combined with DF allows concentration and purification of hemicelluloses, and as a result, concentrates containing large molecules with narrow molar mass distribution suitable for the production of packaging films and hydrogels are produced (Hartman et al. 2006 a, b; Persson, 2009; Edlund et al. 2010; Albertsson et al. 2010). As a conclusion in this study, the removal of low molar mass components from high molar mass GGMs was achieved with DF, but DF could only partially purify GGMs from lignin fragments. As shown inPaper I, the loss of GGM yield was higher with the membranes having higher cut-offs. High GGM concentration (also viscosity) in the DF feed, high fouling and increase in the viscosity (associated with the gel layer) during DF explain the lower permeability during DF than UF.

Table 5.1.5 Variation in the content of 30 kDa concentrated fractions (produced fromSWaH₂) before and after DF (2 dia-volumes).

Analysed compounds Content before DF Content after DF

GGMs, mg/L 9000 4900

Xylans, mg/L 1600 360

Lignin*, mg/L 2200 1500

Uronic acids, mg/L 430 200

Monosaccharides, mg/L 600 30

Turbidity, NTU 5200 4650

Direct UV absorbance 70 40

*: MTBE-treated sample