Improved yield of carbon fibres from cellulose and kraft lignin

Chemical characterisation of raw materials

Prior to carbohydrate analysis, duplicates of lignin samples were hydrolysed according to SCAN-CM 71:09. Klason lignin was determined according to TAPPI T222 om-11, and the acid-soluble residue was measured according to TAPPI UM 250. The carbohydrates were quantified as monosaccharides from the acid hydrolysates. Ion chromatography combined with a pulsed amperometric detector (ICS-5000 IC System) equipped with an autosampler (AS 50) was available, and data evaluation was done using Chromeleon v 6.60, all from Thermo Fisher Scientific (Sunnyvale, CA, USA). The ash content was determined at 525°C according to ISO 1762. The OH and COOH groups in the lignins were determined using ³¹P-NMR spectroscopy (Bruker Avance 400 MHz, Bruker Corp., Billerica, MA, USA) based on Granata and Argyropoulos (1995), with 30 mg of lignin, 2-chloro-4,4,5,5-tetramethyl-1,3,2-dioxaphopolane as the phosphorylation reagent, N-hydroxy-norbornene-2,3-dicarboxylic acid imide as the internal standard and chromium(III) acetylacetonate as the relaxation agent. All reagents were of analytical grade and purchased from Sigma-Aldrich (Steinheim, Germany). Elemental analysis of the lignins was carried out by Analytische Laboratorien GmbH (Lindlar, Germany).

Thermal characterisation of raw materials

The glass transition temperature (T_g) was determined via differential scanning calorimetry (DSC) (Q1000, TA Instruments, New Castle, DE, USA). Details: N₂ flow 50 ml min⁻¹, and sealed aluminium pans containing 3.7±0.3 mg sample in each run. The moisture was removed in a pre-drying step at 105°C for 20 min. The temperature was quenched to 0°C and then the sample was heated to 220°C (10°C min⁻¹). Data evaluation was performed by the software, Universal Analysis 2000 (TA Instruments, New Castle, DE, USA). The reported T_g is the average of three replicates measured as the temperature at half the height of the transition. The decomposition temperature (T_d) and yield of the raw materials were determined by thermogravimetric analysis (TGA Q5000 IR, TA Instruments, New Castle, DE, USA). Duplicates of 15–17 mg samples were analysed. The samples were pre-dried at 105°C for 20 min and then heated to 1000°C (10°C min⁻¹) in N₂ (25 ml min⁻¹). T_d is the temperature, at which 5% of the material is destroyed. The TGA yield is the mass of the remaining material at 1000°C. The average value of duplicates are reported.

Preparation and evaluation of PFs

Lignin was dissolved together with cellulose in [EMIm][OAc] at 70°C for 1 h in a closed reactor with overhead stirring at 30 rpm to 18% solid concentration with a lignin/cellulose ratio of 70:30 by wt. Reference fibres of pure cellulose were spun from solutions with 12% solid content. Successful dissolution was confirmed by observation of solutions in light microscopy, Nikon Eclipse Ci-POL (Nikon Instruments, Tokyo, Japan), equipped with crossed polarisers. Deareation was done at 60°C below 10 kPa pressure for at least 5 h. The solution was analysed through oscillating rheometry with CS Rheometer (Bohlin Instruments, Cirencester, UK) to obtain the temperature-viscosity relation for each spin dope. The spinning temperature of the DP sample was 60°C and that of the KP sample was 80°C. The spinning temperature of the lignin/cellulose blends was set to either 45 or 60°C. The solution was spun in a bench-scale spinning instrument, customized by Swerea IVF, which consists of a spin pump, a spin bath and take-up rolls. Extrusion was performed through a multi-filament spinneret (33 holes, 120 μm capillary diameter, L/D 2) (Sossna, Marl, Germany) over an air-gap of 10 mm into a spin bath of water at 15°C. The fibres were then washed in water for 24 h and treated with fabric softener (Neutral^®, Unilever, Copenhagen, Denmark) for improved handling before they were dried at 80°C for 45 min.

Tensile testing (Vibroskop/Vibrodyn, Lenzing Instruments, Austria) was performed on conditioned fibres at 20±2°C and 65±3% relative humidity (RH) with an extension rate of 20 mm min⁻¹ and a gauge length of 20 mm. Conversion to SI units (Pa) was done based on a fibre density of 1.3910 g cm⁻³, obtained by a He-pycnometer AccuPycII 1340 (Micromeritics Instrument, Norcross, USA) at room temperature.

Ultraviolet-visible (UV/Vis) absorbance (using SPECORDE^® 200 PLUS, Analytik Jena AG, Jena, Germany) at 280 nm was measured on spin baths for determining the concentration of leached lignin. The absorbance was converted to lignin concentration based on an extinction coefficient of 24.6 l g⁻¹ cm⁻¹ (Fengel et al. 1981). The possible absorption of carbohydrate conversion products at 280 nm, such as furfural, could be ignored as such products were not expected in the lignins under study. Correction was made for the contribution to the absorbance from [EMIm][OAc], which was about 1/10^th of that of lignin. This was determined by conductivity measurements of the spin baths at 23°C (inoLab Cond 720 Benchtop Conductivity Meter, Thomas Scientific, Swedesboro, NJ, USA) based on a linear calibration curve covering the range 0.3–0.7% [EMIm][OAc]. The soluble (dissolved) lignin did not affect the conductivity values in the course of the [EMIm][OAc] quantification.

Gel permeation chromatography (GPC) was performed on solutions with 25 mg ml⁻¹ lignin content. The mobile phase DMSO/10 mM LiBr was passed through a 0.2 μm filter before analysis (GHP Acrodisc^®, Merck, Darmstadt, Germany). Samples were made in duplicates, and each sample was injected twice. A PL-GPC 50 Plus integrated system connected with refractive index (RI) and ultraviolet (UV) detectors was used (Polymer Laboratories, Varian Inc., Iselin, NJ, USA). The separation was performed with a flow rate of 0.5 ml min⁻¹ on mixed columns (PolarGel-M, 300×7.5 mm) preceded by a similar guard column (50×7.5 mm) (Agilent, Santa Clara, CA, USA). Data analysis was performed via the Cirrus GPC software version 3.2, from the same manufacturer.

Preparation and evaluation of CFs

To prevent fibre shrinkage during thermal treatments, precursor fibres (PFs) were fixed on graphite bridges supplied by Gerken Nordiska Karma AB (Järfälla, Sweden). Stabilisation of the fixed fibres was performed in the presence of air (7 l min⁻¹) in a muffle furnace (KSL-1200X, MTI Corporation, Richmond, VA, USA) by heating at 0.2°C min⁻¹ to 200°C and then at 1°C min⁻¹ to 250°C (held isothermally for 1 h). Carbonisation was performed in a tube furnace (Model ETF 70/18, Entech, Ängelholm, Sweden) equipped with a ceramic (Al₂O₃) tube (70 mm diameter×1180 mm length). The fixed stabilised fibres were placed in the tube furnace, vacuum was applied for 3×5 min, and a N₂ flow of 200 ml min⁻¹ was applied before heating at 1°C min⁻¹ to 600°C and then at 3°C min⁻¹ to 1000°C. The gases were purchased from Strandmøllen (Ljungby, Sweden).

Tensile testing was performed in a fibre dimensional system (LDS0200) (Dia-Stron Ltd., Andover, UK) utilising laser diffraction for diameter determination (CERSA-MCI, Cabries, France) and a tensile system (LEX820; Dia-Stron Ltd.) including a UV 1000 control unit. Sample loading/unloading was done by means of ALS1500; Dia-Stron Ltd. automatic loading system, equipped with a pneumatically operated sample holder (PU 1100; Dia-Stron Ltd.). The extension rate during tensile testing was 0.5 mm min⁻¹ and the gauge length was 20 mm. Control and evaluation were done using UvWin 3.35.000 software (Dia-Stron Ltd.). Surfaces and cross-sections of the CFs were evaluated using scanning electron microscope (SEM) (Hitachi SU3500, Tokyo, Japan), 10 kV, secondary electron detector. The elemental composition (atomic %) of CF cross-sections was determined by energy dispersive X-ray analysis (EDXA) (XFlash detector, Bruker, Corp., Billerica, MA, USA), 10 kV, backscattered electron detector at a working distance of 10 mm. Data analysis was conducted with Esprit v. 1.9.3 software (Bruker, Corp., Billerica, MA, USA). The reported values are the average of three measurements of each cross-section. The yield of the CFs was determined gravimetrically by placing 270–290 mg precursor fibres in ceramic ships that were subjected to stabilisation and carbonisation of the CFs.

Characteristics of the lignin and cellulose samples

As a lignin source, KL and two membrane fractionated KL retentates, i.e. RL5 and RL15 are the focus, where 5 and 15 refer to the MM cut-off in kDa. As Figure 2 shows, the retentates are enriched with high-MM fragments. Either a commercially available dissolving pulp (DP) with an intrinsic viscosity of 460 ml g⁻¹ or a fully bleached paper-grade softwood kraft pulp (KP) with an intrinsic viscosity of 630 ml g⁻¹ served as cellulose sources for the mixed lignin/cellulose precursors. The analytical data confirm the expectations (Table 1); for example, the lower intrinsic viscosity and the lower amounts of xylose and mannose in the sample DP compared to the paper grade KP sample. The lignins contained about the same amounts of acid insoluble (94–98%) and acid soluble moieties and only traces of carbohydrates. The retentate lignins contain even less carbohydrates. All samples have an ash content below 1%. The retentate lignins contain similar concentrations of carboxylic groups as KL, whereas the RL5 and RL15 contain less OH_aliph and OH_phen groups (see Table 2). The total amount of OH groups and their composition are in accordance with the literature data (Brodin et al. 2009). The elemental analysis resulted in the expected carbon contents (64–65%) of SW-KL (Table 3) (Gellerstedt 2015). The theoretical carbon content of DP and KP was approximated by that of cellulose (44.4%).

Figure 2:

Size-exclusion elution profiles of kraft lignin (KL) and retentate lignins (RL5 and RL15), sample designations are explained in Figure 1.

The dotted vertical lines show the molecular mass peak (M_p) of pullulan standards for the corresponding elution time.

In the case of KL, the T_g (151°C) is in the same range as typical for SW-KLs (Brodin et al. 2009). In the retentate lignins, T_g could not be detected because of the absence of thermally mobile low-MM fragments (Yoshida et al. 1987; Norberg et al. 2012). Sorbed water, which is hard to remove due to extensive hydrogen bonds, strongly influences the DSC data of cellulose. By extrapolation, the T_g of dry cellulose has been estimated to be around 220°C (Szczesniak et al. 2008), but the T_g of cellulose could not be measured in the present work. Thermal stability, expressed here as mass loss in inert atmosphere up to 1000°C, differed considerably between the samples (Figure 3). Lignin samples have a slightly lower thermal degradation temperature (T_d, defined as 5% mass loss), reflecting an initial lower stability. The yield at 1000°C is 37.5% for KL and 40.2% for the two retentate lignins (average), compared to that of the two pulp samples (12.0% average). The higher yield of retentate lignins is due to their low carbohydrate contents, and because high-MM lignins are more stable due to crosslinking, which occur via radical reactions during kraft pulping (Gellerstedt 2015). The initially more stable pulp samples suffered a sharp mass loss above 300°C and ended up at a final residue of 12% at 1000°C. This has already been described for pyrolysis of wood-derived viscose (Liu et al. 2005). Once the adsorbed water is removed at 150°C, a three-step pyrolysis process begins (Tang and Bacon 1964). Above 200°C, dehydration of the equatorial hydroxyl groups takes place, presumably starting in the paracrystalline moiety and carbonyl groups are formed. The true pyrolytic degradation begins around 240°C leading to chain scission and formation of double bonds via free radical reactions and degradation of the cellulose units to four carbon units. In a competing reaction, levoglucosan is formed, ending up in tar products. This is the reason why the mass loss of the cellulose depends strongly on the heating rate (Brunner and Roberts 1980). Above 400°C, aromatisation is initiated through repolymerisation of the four carbon units.

Figure 3:

Thermal stability in inert atmosphere of lignin samples (KL, RL5, RL15) and cellulose samples (KP and DP).

Sample designations are explained in Figure 1.

The lower initial stability of the hemicellulose-containing KP-sample (T_d 290°C) compared to the DP_s-sample (T_d 303°C) and the slightly higher yield at 1000°C of the former are because of the lower thermal stability of xylan in the initial pyrolysis phase and the notably higher yield at 1000°C compared to cellulose (Yang et al. 2007).

Dope characteristics and spinning of PFs

Cellulose and lignin were mixed to form different spin dopes, which were dry-jet wet-spun to produce PFs. The rheological behaviour is an indication for the spinnability, thus the dopes were characterised by oscillatory rheology. The pulp types have a large impact on viscosity, but no significant differences were found between the lignin types. According to Michud et al. (2015), the zero-shear viscosity should be within 27 000–40 000 Pa s and the crossover point (COP) should be between 0.8 and 1.5 Hz. For solutions with these rheological properties, a sharp increase in the maximum draw ratio (DR) from 3 to 15 was observed. These viscosities seem to be acceptable for the PF production. During dry-jet spinning with IL as solvent, a medium DR of around 4–6 is often used to reach complete fibre orientation. In the present work, the highest DRs were attained, when COP was between 0.1 and 1 Hz and the complex viscosity at 0.02 Hz was still above 1000 Pa s. Thus, the solution temperatures were set to achieve these spinning conditions. Therefore, all trials with lignin and KP were spun at 60°C and the samples with lignin and DP at 45°C (Figure 4).

Figure 4:

Complex viscosity at 0.02 Hz versus frequency at the cross over point (COP) obtained from oscillatory rheology measurements at 30, 45, 60 and 80°C for all lignin-cellulose dopes.

Filled symbols represent the temperatures used for dry-jet wet spinning. KL, Kraft lignin; RL5 and RL15, retentate lignins; KP, fully bleached kraft paper grade pulp; DP, dissolving pulp. Sample designations are explained in Figure 1. In the legend, the KP-series is marked in black text and the DP-series is marked in grey text as is the corresponding temperatures in the graph.

Except for the neat DP, all PFs could be successfully spun with a DR of at least 4.0, whereas for the DP sample, a DR of 3.3 was selected to compare the mechanical properties between the PFs (Table 4). In the batch-spinning process, the target diameter of the precursor fibres was set to ≈25 μm to simplify handling during conversion and after carbonisation. With the chosen spinning method, the diameter could easily be decreased to match the PF for commercial CF, for example, by reducing the capillary diameter in the spinneret and/or by increasing the take-up speed.

PF properties

The solutions containing both cellulose and lignin were generally easier to handle during the dry-jet spinning compared to solutions with only cellulose. The results in Table 4 show that the PFs have a consistent diameter, between 22 and 27 μm. Expectedly, TS and TM decrease in the presence of lignin because of their lower MM and their amorphous character. This observation is in line with the literature data concerning similar cellulose-lignin mixed systems (Ma et al. 2015). It should be noted that there are no significant differences in mechanical properties of the precursor fibres containing different types of lignin or cellulose. Furthermore, the mechanical properties of the neat cellulose precursor fibres, KP-PF ~30 cN/tex and DP-PF ~25 cN/tex, are similar to those of commercial cellulose fibres (like viscose ~20–25 cN/tex, cotton ~24–36 cN/tex and Lyocell ~40–42 cN/tex) (Bredereck and Hermanutz 2005). Dry-jet spinning of neat lignin solutions was not possible, and addition of a fibre-forming polymer (such as cellulose) was necessary. The applied spinning method in the present paper can be considered as robust with a wide processing window with respect to raw materials and spinning parameters. The PFs of pure cellulose were less brittle compared to PFs produced from cellulose and lignin blends. The lignin-containing fibres were still easy to handle.

Mass loss in PF preparation

It can be assumed that cellulose is not lost during the spinning. On the other hand, hemicelluloses, especially in the case of paper-grade pulp (KP), may partly be released into the spin bath. However, only a marginal decrease in xylose and mannose contents were observed compared to theoretical values. To maximise CF yield, the lignin loss during wet spinning should be minimised. The lignin loss in the dry-jet wet spinning step was estimated by characterisation of the coagulation bath with respect to the UV-absorbing chromophores of lignin and conductivity due to leached [EMIm][OAc] as well as the MM distribution of lignin in the PF. Based on the assumption that the same amount of coagulated fibre was passing through the spin bath, the lignin loss in the spin bath was slightly higher for the KL sample 2.2±0.3% compared to 1.6±0.2% for RL5 and RL15, but this difference is not statistically significant.

The difference between KL and retentate lignins was confirmed by GPC analysis (Figure 5). The MM distribution (MMD) of the lignin isolated from the cellulose-containing PFs (KL:DP-PF and KL:KP-PF) reflects the mass loss of low-MM fragments into the spin bath. A small difference in MM can be observed concerning the low-MM fragments between the samples from different cellulose sources. Apparently, pure cellulose (DP) has a somewhat greater ability to retain the low-MM lignin fragments from the KL sample compared to the paper-grade cellulose (KP) containing hemicelluloses. However, this qualitative difference between KL:DP and KL:KP could not be quantitatively verified. It cannot be ruled out that the qualitative difference in lignin composition influences the thermal conversion of the fibres and may also have a negative impact on the recyclability of the IL solvent in a large-scale process.

Figure 5:

Size-exclusion elution profiles of raw material lignins (KL and RL15) and lignin retained in the corresponding lignin-cellulose precursor fibres (KL:KP-PF, KL:DP-PF and RL15:KP-PF, RL15:DP-PF).

Sample designations are explained in Figure 1. The dotted vertical lines show the molecular mass peak (M_p) of Pullulan standards for the corresponding elution time.

Thermal conversion and CF properties

The PFs were stabilised and carbonised under the same conversion conditions. Shorter stabilisation times have been reported for laboratory-made cellulose-based CF production (Dumanli and Windle 2012), but the temperature programme described here is typical for lignin-based PF (Norberg et al. 2012), and was also successful for stabilisation of dry-jet wet spun lignin-cellulose PFs (Olsson et al. 2017). Neat cellulose PFs should preferably be stabilised, fixed to keep the molecular order attained during spinning and to counteract shrinking of the fibre (Karacan and Gül 2014). Tensioning during carbonisation has a positive influence on the mechanical properties of cellulose-based CFs (Bahl et al. 1998) and has been applied on a pilot-scale CF line (Byrne et al. 2016). In the present study, the PFs were mounted on graphite bridges prior to the thermal treatment and converted in laboratory-scale batch equipment. Keeping the fibres in a fixed state during the carbonisation has a positive impact on the modulus of the final CF (Olsson et al. 2017).

By definition, the carbon content of CF should be more than 90%. The elemental analysis was performed by EDXA, which does not detect the hydrogen content. The CFs made from neat cellulose pulp (DP-CF and KP-CF) contained 95 atomic% carbon, and a minor fraction of oxygen (4.6 atomic%). For CFs made from the lignin-cellulose blends, the carbon content was in the 92–97 atomic% range. Traces of sulphur were detected in all lignin-based CFs, but expectedly, sulphur is absent in neat KP-based CFs. The large lignin fractions (70%) have no detrimental impact on the appearance of the CF. Instead, the outer surfaces are smooth, and no flaws or irregularities in the CF cross-sections are visible on the SEM images of the neat KP-CF and the KL:KP-CF samples (Figure 6).

Figure 6:

Scanning electron microscopic images of the surface, (a) and (c) (10 kV and 2000X, secondary electron detector spot 60), and cross-sections, (b) and (d) (10 kV and 4000X, secondary electron detector spot 60), of carbon fibre made from neat cellulose (KP-CF; a and b) and lignin:cellulose blends (KL:KP-CF; c and d).

Sample designations are explained in Figure 1.

Tensile testing gave similar results for all samples (Table 5). It has been suggested that the ash content of raw materials should be below 0.1% to avoid impairing the mechanical properties of CF (Kadla et al. 2002). The ash content of the raw materials in the present study was in the range 0.1–1%, and obviously it did not dramatically affect the tensile properties of the CF. The coagulation and washing steps in the PF processing probably purify the precursors with respect to inorganic components.

Both TS and TM were in the same range as previously reported for SW-KL:DP (70:30) carbonised in a fixed state (Olsson et al. 2017). Thus, the TM of CF made from lignin and bleached paper-grade pulp can be improved by fibre fixation during carbonisation. It appears that KL as well as KP can be used with equal success as the more refined (and more expensive) retentate lignins and dissolving grade pulp (DP). The level of the average mechanical properties is below the target set by the US automotive industry (1.72 GPa TS and 172 GPa TM), but meets the target demand on elongation (>1%). The mechanical properties of the neat cellulose-based CF are in the same range as generally reported for cellulose-derived CF (Dumanli and Windle 2012), but inferior to CF made from high-strength textile-grade viscose PF that, after optimisation of the thermal conversion steps, rendered exceptionally high TS (2.0 GPa) and TM (84 GPa) (Spörl et al. 2017). The TS (720–1000 MPa) and TM (64–74 GPa) reported in the present study are better than those reported for neat SW-KL-based CF (300–465 MPa TS and 30–32 GPa TM) (Nordström et al. 2013; Salmén et al. 2015). Thus, from the perspective of mechanical properties, lignin-cellulose combinations are a promising alternative for producing lignin-based CFs. In spite of the difference in diameter and tensile properties between the two cellulose PFs, there are no differences in this regard between the corresponding CFs. It is remarkable that the less refined KP performs as good as the DP.

Yield of CF

The gravimetric yield of CF, measured as mass retained from the PF after oxidative stabilisation and carbonisation at 1000°C, reveals a significant impact of lignin (Figure 7). The theoretical yield was estimated from the PF composition considering the carbon content of cellulose (44.4%) and those determined on different lignins (Table 2), taking into account the initial relative contents of the two components (lignin:cellulose/70:30).

Figure 7:

Theoretical yield (wt.%) (white) calculated from the carbon content of the raw materials and the fraction of respective component in the prefibre, and the gravimetric yield of the corresponding CFs: DP-series (striped) and KP-series (grey).

KL, Kraft lignin; RL5, retentate kraft lignin from using 5 kDa cut-off membrane; RL15, retentate lignin from using 15 kDa cut-off membrane; DP, dissolving grade pulp; KP, fully bleached kraft paper-grade pulp. The standard deviation for the yield of the KL:KP-CF sample was ±0.27 wt.% (n=5).

The gravimetrical yield was improved for all oxidatively stabilised samples compared to thermal treatment of the lignin powder and cellulose samples in inert atmosphere (Figure 3), illustrating the protective effect of oxidation. Also, the gravimetrical CF yield was increased from 21 to 22% of the neat cellulose-based CF to about 40%, when the CF was made from lignin-cellulose blends, irrespective of the KL source.

In general, CF yields are seldom reported in the literature. A comparison was made based on CFs made from neat HW-KL (46% yield) and CF from HW-organosolv lignin (42% yield) (Qin and Kadla 2012). The mechanical properties of the CFs were around 400 MPa and 40 GPa for TS and TM, respectively, i.e. lower than those obtained in the present study. The best cellulose-based CF yield reported so far is 38%, obtained by adding ammonium dihydrogen phosphate to prevent tar formation, which is, in general, responsible for the comprehensive mass loss during thermal treatment (Spörl et al. 2017). There is still a lot of possibility for further optimisation of the conversion of the lignin-cellulose PFs to CF.

In contrast to the thermal treatment in an inert atmosphere, the yield relative to the theoretical at 1000°C of the neat cellulose DP-CF sample produced via oxidative stabilisation was higher (51%) than the corresponding KP-CF (48%). This indicates that cellulose from DP is more susceptible to oxidative treatment. This is probably due to the higher content of end-groups in DP as compared to the long-chained cellulose in KP. The yields of the lignin: cellulose-derived CFs were about 69% of the theory, illustrating the positive contribution of lignin to the overall CF yield, but no difference between the three lignin types (KL, RL5 and RL15) was seen.