Fraction analyzed; Ar: aromatic residue; R: aliphatic chain.

For K-LF-0 fractionation, the results of the GC/MS analysis are reported in Table 4. Additionally, in this case, the efficacy of the fractionation procedure, expressed as percentages of total monomers in each sample, increased when proceeding through the processes: K-LFa-R5 and K-LFo-R5, constituting

the higher molecular weight retentates, contain only 21% of small molecules, which could be inferred from GC-MS; on the contrary, K-LFa-P2 and K-LFo-P2, constituting the fractions permeated from the finest membrane (2 kDa cutoff), were considerably enriched in small molecules (around 70%).

The chromatogram profiles of the different fractions and the structures of the most abundant identified compounds with their relative abundance are reported in Table 3 and Figure 6 for S-LF-0 and Table 4 and Figure 7 for K-LF-0. As discussed above (Tables 3 and 4), the enrichment in smaller molecular weight products clearly appeared when moving throughout the fractionation processes for S-LF-0 from fractions S-LFa/o-R5 to S-LFa/o-P2 (as shown in Figure 6), whereas for K-LF-0 the number of peaks in the chromatograms becomes much more relevant in K-LFa/o-P2 (in Figure 7). The peaks from 7 to 12 min are mainly due to the silylating agent used for the derivatization of the analyzed samples. The peak at 6 min is attributed to benzaldehyde, which was used as the internal standard.

**Figure 6.** GC/MS total ion chromatograms of delivered fractions from aqueous and organic processes using S-LFa/o-0. Selected component identification by MS library matching is reported for each fraction.

**Figure 7.** GC/MS total ion chromatograms of delivered fractions from aqueous and organic processes using K-LFa/o-0. Selected component identification by MS library matching is reported for each fraction.

#### *2.7. Determination of the Hydroxylation Levels in the Recovered Fractions*

The determination of hydroxylation levels in lignins constitutes an essential step for the effective valorization of the fractions obtained from the fractionation processes. The classical methods are the Folin–Ciocalteu titration method (FC), which allows the quantification of phenolic groups, and 31P-NMR analysis, which leads to a quantitative determination of aliphatic and aromatic hydroxyl groups as well as of carboxyl moieties.

#### 2.7.1. Phenolic Hydroxyl Group Determination

The FC assay was carried out to determine the total phenolic content in the original lignin and all process fractions. It is based on the reaction of phenolic hydroxyl groups with a specific redox reagent (FC reagent), which leads to the formation of a blue chromophore, which is, however, sensitive and unstable in strong bases. Therefore, based on an analytical protocol previously demonstrated by our group [41], DMSO was used as solvent for the samples in order to obtain their complete solubilization in neutral conditions.

The phenolic content results are reported in Tables 5 and 6 as vanillin equivalents (mmol/g of dry lignin). In the case of soda lignin, S-LFa/o-R2 and S-LFa/o-P2 appeared to be the fractions with a higher content in phenolic hydroxyl groups, in agreement with the trends observed with 31P-NMR, as will be discussed later in the text.



**Table 6.** Results of the determination of phenolic hydroxyl groups expressed as vanillin equivalents in the lignin fractions recovered from aqueous and organic processes applied on K-LF-0 (expressed in mmol/g) in the dry lignin. Estimated standard errors ±0.2 mmol/g vanillin equivalent (1 σ, from calibration data).


If we compare the two starting lignins, K-LF-0 appears to contain an increased amount of phenolic hydroxyl groups compared to S-LF-0, probably due to the extensive cleavage of β-aryl group bonds during the cooking process. In S-LF-0 fractionation, a very high enrichment in phenolic groups appears in the fractions R2 and P2 as reported in Table 5, in excellent agreement with the GC/MS analysis reported in Figure 6. For K-LF-0 fractionation, as reported in Table 6, the highest value of phenolic content appears to be associated with K-LFa-P2 for the aqueous process, whereas in the organic one, the R5 fraction appears to be enriched in phenolic groups.

#### 2.7.2. Total Hydroxyl Groups' Quantification with 31P-NMR

The different types of hydroxyl groups present in the starting lignins and fractionated samples were deeply investigated by 31P-NMR. The 31P-NMR spectra of the phosphitylated lignin fractions are shown with chemical shift assignments in Figures 8 and 9, where the parent materials S-LF-0 and K-LF-0 are also shown for easy comparison. In detail, the following spectral regions were integrated to acquire the information about the chemical nature of the different hydroxyl groups: 150–147 ppm for the signals associated with aliphatic hydroxyl groups, and 145–138 ppm for the signals associated with aromatic hydroxyl groups. The signals of carboxylic acids groups were centered at 136 ppm. The peak integration in the three main portions of the spectrum led to the quantification of the total hydroxyl groups expressed in mmol of functional groups per gram of dry compound, as reported in Tables 7 and 8 for S-LF-0 and K-LF-0, respectively.

**Figure 8.** 31P-NMR spectra of lignin fractions recovered from the aqueous and the organic process applied on S-LF-0.

**Table 7.** Detailed hydroxyl/carboxyl quantification by 31P-NMR for the fractions recovered from aqueous and organic processes applied on S-LF-0.


**Table 8.** Detailed hydroxyl/carboxyl quantification by 31P-NMR for the fractions recovered from aqueous and organic processes applied on K-LF-0.


\* 31P-NMR measurements could not be performed for this fraction due to extensive precipitation of the sample in the conditions used for the analysis; \*\* possible underestimation of these values due to the limited solubility of the sample in the solvent used for the analysis.

**Figure 9.** 31P-NMR spectra of lignin fractions recovered from the aqueous and the organic process applied on K-LF-0.

Both FC assay and 31P-NMR data indicated a slight increase in phenolic hydroxyl groups in the lower molecular weight fractions B-LFa/o-R2 and B-LFa/o-P2. A similar trend was also reported from 31P-NMR analysis on aliphatic hydroxyl and carboxyl groups, in agreement with recent works on analogous technical lignins [43–45]. These results have great interest in terms of further selection of an appropriate fraction for a specific application, where the relative abundance of the functional groups is essential for the further processing.

#### *2.8. Fourier-Transform Infrared Spectroscopy (FTIR)*

FTIR spectroscopy was employed to analyze the chemical characteristics of the lignin fractions recovered after solvent extraction and subsequent membrane-assisted ultrafiltration. FTIR spectra of both the soda and kraft lignin fraction systems are presented in Figure 10, along with the FTIR spectrum of each parent lignin also shown for comparisons. All fractions presented a broad absorption band in the 3600–3100 cm−<sup>1</sup> range that could be attributed to stretching vibrations of hydrogen-bonded phenolic and aliphatic O−H groups present in lignin, with a peak at 3390 cm−<sup>1</sup> and a shoulder at 3200 cm−1, respectively [47]. Signals observed in the 3050−2800 cm−<sup>1</sup> region were attributed to the symmetrical and asymmetrical CH stretching of the methyl and methylene groups. At around 1700 cm−1, a clear peak was observed in all spectra with the exception of K-LF-0, where more of a shoulder was observed, attributable to the stretching vibration of C-O bonds in conjugated aldehydes and carboxylic acids. In particular, it was observed that after solvent extraction (in both kraft and soda lignin streams), an increase in this signal was registered compared with the pristine samples (especially in low molecular weight fractions S-LFa/o-P2 and K-LFa/o-P2), likely indicating a higher concentration of carbonyl and carboxylic groups. The presence of the intense peak at 1515 cm−<sup>1</sup> was associated to the pure aromatic skeletal vibrations in lignin. In the 1400−1000 cm−<sup>1</sup> spectral region, signals of variable intensity were observed, which could be attributed as follows: Bending vibrations of phenolic O−H and aliphatic C−H in methyl groups (1370 cm−1); C−O, C−C, and C−O stretching vibrations (1270 and 1210 cm<sup>−</sup>1); C−H in plane deformations (1125 cm<sup>−</sup>1); and C−O deformations in primary (1030 cm<sup>−</sup>1) alcohols.

**Figure 10.** FT-IR spectra of soda (**a**,**b**,**c**,**d**) and kraft (**e**,**f**,**g**,**h**) lignin fractions recovered from the aqueous process (**a**,**b**,**e**,**f**) and the organic process (**c**,**d**,**g**,**h**). Different regions of the infrared spectrum are reported for clarity.

In the kraft fractions, both in aqueous and organic solvent, a clear peak at 1154 cm−<sup>1</sup> was observed, which could be assigned to C-O deformations in conjugated ester groups present in guaiacyl (G), syringyl (S), and *p*-hydroxyphenyl (H) groups. This signal indicated a higher abundance of such moieties, not present in soda fractions.

#### *2.9. Di*ff*erential Scanning Calorimetry (DSC)*

The thermal transitions in all lignin fractions, in particular the glass transition temperature (Tg), were evaluated by means of DSC analysis and compared with the pristine samples S-LF-0 and K-LF-0. DSC traces of both soda and kraft lignin fraction systems are presented in Figure 11. As it can be seen in

the plots, in the case of organic solvent fractionation, the solvent extraction step led to a decrease in Tg of all the recovered materials, with samples S-LFo-P2 (Tg = 58 ◦C) and K-LFo-P2 (Tg = 44 ◦C) showing the lowest values, in accordance with the trends observed on molecular weights in the GPC results and in line with what was recently reported on analogous systems [43,44]. In the case of aqueous fractionation, a similar trend was observed on S-LFa-P2 and K-LFa-P2, with comparable values of Tg (Tg = 50 ◦C and 46 ◦C for S-LFa-P2 and K-LFa-P2, respectively). Sample S-LFa-R5 showed a slightly higher Tg than the parent sample S-LF-0, likely due to the slightly higher value of Mn observed for this fraction in GPC measurements. Similarly, the Tg of sample K-LFa-R5 was not easily detectable from the DSC trace but was higher (Tg = 180 ◦C) than the corresponding parent material K-LF-0 (Tg = 155 ◦C), in line with the high molecular weight inferred for this fraction from GPC analysis (Table 2).

**Figure 11.** DSC traces of soda (**a**,**b**) and kraft (**c**,**d**) lignin fractions recovered from the aqueous process (**a**,**c**) and the organic process (**b**,**d**).

The similar trends observed on the resulting Tg of the two different lignin systems (soda vs. kraft) upon fractionation in both aqueous and organic phase further demonstrated the versatility of the proposed process, which did not appear to be significantly affected by the solvent used for the extraction operation. Moreover, they provided additional proof of the ability of this approach to deliver consistent and reproducible fractions of well-controlled and predictable thermal characteristics irrespective of the biomass origin, with this aspect being particularly useful for potential use in the development of lignin-based macromolecular materials.

#### **3. Materials and Methods**

#### *3.1. Materials*

All chemicals and analytical grade solvents, such as tetrahydrofuran (THF), methanol (MeOH), ethyl acetate (EtOAc), n-butyl acetate (BuOAc), and *tert*-butyl methyl ether (MTBE), were purchased from Sigma-Aldrich. 2-Butanone (methyl ethyl ketone, MEK) was provided by BCD Chemie GmbH (Hamburg, Germany). Lignin S-LF-0 (Protobind 1000, a mixed wheat straw/Sarkanda grass lignin from soda pulping of non-woody biomass) was provided by Tanovis (Alpnach, Switzerland). Lignin K-LF-0 (softwood Kraft lignin, Indulin AT) was provided by Meadwestvaco (Charleston, SC, USA). Lignins

were used as received. All solutions were prepared in Milli-Q water (Elix Millipore Purification System, Molsheim, France). All analyses were carried out at least in duplicate unless otherwise stated.

#### *3.2. Lignin Solubility in Organic Solvents*

Lignin solubilities in organic solvent were determined by treating 10 g of starting lignin with 100 mL of the different solvents stirring at 400 rpm. Each test was carried out overnight at room temperature. The suspensions were then filtered, and the solvents were evaporated at reduced pressure and the final residues were dried until a constant weight was achieved prior to quantification.

#### *3.3. Fractionation via Aqueous Process*

The aqueous process fractionation comprised a microfiltration step followed by a two-membrane cascade ultrafiltration (shown in Figure 1 as part of an idealized lignin valorization scheme) and a final fraction recovery step, as described below (see also the discussion section).

#### 3.3.1. Microfiltration

The lignin solution in ethanol/ water was prepared by dissolution of 150 g of lignin per L of a mixture of ethanol and NaOH aqueous solution at pH 10. The insoluble material was eliminated by a microfiltration on a 0.7-μm fiberglass filter under vacuum.

#### 3.3.2. Membrane-Assisted Fractionation

Lignin fractionation was performed by means of an ultrafiltration apparatus (Sartoflow Advanced filtration module purchased from Sartorius Stedim, Aubagne, France) equipped with flowmeters and pressure sensors to control the permeate flow, trans-membrane pressure, and filtration time. The membranes (Hydrosart membranes, Sartorius Stedim) in stabilized cellulose had nominal molecular weight cutoffs of 5 and 2 kDa and a filtration area of 0.1 m2 each. The membrane regeneration and storage were performed at 40 ◦C using 1 M and 0.1 M NaOH solution, respectively.

These membranes were compatible with a high range of different solvents, allowing to switch easily from an aqueous solvent to an organic one. The procedure required the cleaning of the membrane with a solution of 1M NaOH at 40 ◦C, and then the washing and conditioning of the membrane in the appropriate solvents for the successive fractionation sequence. The membranes could be also stored without any problems in 0.1 M NaOH.

#### 3.3.3. Fraction Extraction Procedure

The recovered fractions from the aqueous process were evaporated under reduced pressure to remove ethanol. The aqueous solutions were acidified to pH close to 1 with 37% HCl, and extracted three times with ethyl acetate. The combined organic phases were dried over sodium sulfate and evaporated under reduced pressure. The final solid residues were dried until a constant weight was achieved prior to analysis and quantification.

#### *3.4. Fractionation via Organic Process*

The organic process fractionation comprised a Soxhlet extraction followed by a two-membrane cascade ultrafiltration (shown in Figure 1 as part of an idealized lignin valorization scheme) and a final fraction recovery step, as described below (see also the results and discussion section).

#### 3.4.1. Soxhlet Extraction

Soxhlet extractions were performed using a standard glass apparatus (Buchi extraction system B-811, Villebon sur Yvette, France), which allowed 4 extractions to be achieved at the same time. Each extraction unit was composed of a 150-mL working volume bottom solvent flask, a 330-mL-capacity thimble holder and a water-cooled condenser. For each extraction process, 150 mL of MEK were placed

in the solvent flask and about 15 g of lignin were inserted inside a 41-mm diameter and 123-mm height cellulose paper thimble (FiltraTECH, Saran, France). The solvent reflux was kept for 8 h, adjusting the heating power in order to have 4 extraction cycles/h.

#### 3.4.2. Membrane Ultrafiltration

Lignin fractionation was performed in MEK directly on the Soxhlet extraction solution by means of the same apparatus described in Section 3.3.2 with the membranes conditioned and used only in MEK.

#### 3.4.3. Fraction Recovery Procedure

The recovered fractions from the organic process were evaporated under reduced pressure without further processing. The final solid residues were dried until a constant weight was achieved prior to analysis and quantification.

#### *3.5. Gel Permeation Chromatography (GPC)*

A Waters 510 HPLC system equipped with a refractive index detector was used for GPC analyses. Tetrahydrofuran (THF) was used as eluent. The analyzed lignin sample (volume 200 μL, concentration 1 mg/mL in THF) was injected into a system of three columns connected in series (Ultrastyragel HR, Waters–dimensions 7.8 mm × 300 mm) and the analysis was performed at 30 ◦C at a flow rate of 0.5 mL/min. The GPC system was calibrated against polystyrene standards in the 102–104 g/mol molecular weight range. To allow complete solubility in the THF eluent, before the analysis, the parent lignin and the fractions were acetylated following a standard literature procedure. The estimation of the number-average and weight-average molecular weights of the obtained lignin fractions was performed excluding the signals related to the solvent (THF) and the solvent stabilizer (butylated hydroxytoluene), visible at long elution times (>29.5 min).

#### *3.6. Gel Filtration Chromatography (GFC)*

GFC analyses were performed using a Merck-Hitachi L4000 apparatus (Tokyo, Japan) equipped with a UV detector set at λ of 254 nm. A hydroxylated acrylic polymer GFC column (Polysep-GFC-P 2000, 300 × 7.8 mm, Phenomenex, Aschaffenburg, Germany) was used with a water-based eluent (sodium borate pH 10 buffer 10 mM plus 300 mM NaCl) at 0.5 mL/min flow rate.

The samples (5 mg) were completely dissolved in a small amount of NaOH. The pH of the solution was adjusted to pH 10 with HCl 6 M, and diluted to the final concentration of 2 mg/mL with sodium borate buffer 10 mM pH 10. The sample was then centrifugated, filtered, and analyzed.

#### *3.7. Gas-Chromatography*/*Mass Spectrometry (GC*/*MS)*

The GC/MS apparatus used was an Agilent GC System 7890A, with an inert MSD with Triple-Axis Detector 7975C (Cernusco sul Naviglio, Italy). The separation was performed on a DB-5MS column (30 m × 250 μm × 0.25 μm, Phenomenex) with a helium flow rate of 1.18 mL/min, a temperature program of 50 ◦C (1 min) to 280 ◦C at 10 ◦C/min, 280 ◦C at 15 min (total run time 39 min, temperature of the injector 250 ◦C, injection volume 1.00 μL, injection mode split, split ratio 5:1). A solvent delay of 4 min was selected. The samples were prepared by derivatization and dissolved in methanol or acetone in a concentration around 0.5-1 mg/mL as previously described [44]. Compound identification was performed by means of an NIST 2008 mass spectral library search.

#### *3.8. Folin–Ciocalteu Assay (FC Assay)*

Total phenolic contents of the different fractions were determined by the classical Folin–Ciocalteu method with some modifications in the sample preparation step. The samples were prepared by dissolving lignin in dimethylsulfoxide (DMSO) with a final concentration of 1 mg/mL. DMSO was

chosen because, being completely miscible in water, it allowed a complete lignin solubilization and did not interfere with the FC assay.

For each determination, 5 μL of the working solution (or the standard solution) were then mixed with 120 μL of deionized water, 125 μL of FC reagent (Sigma 47641), and kept for 6 min at r.t. after 30 s of vortex stirring. Then, after the addition of 1.25 mL of 5% sodium carbonate and mixing, the vial was incubated on a thermoshaker at 40 ◦C for 30 min. The reaction mixture absorbance was measured using a UV/Vis spectrophotometer (Jasco V-560) equipped with a temperature-controlled cuvette holder and a thermostatic water bath (Haake K10, Karlsruhe, Germany). All spectrophotometric measurements were carried out at 760 nm, 25 ◦C, using a 1-cm optical path cuvette. Vanillin was chosen as the reference standard. The calibration curve was constructed with nine different vanillin solutions in DMSO with the concentration in the range 0–800 μg/mL. Each FC assay determination was carried out in triplicate.

#### *3.9. 31P-NMR Analysis*

31P-NMR spectroscopic analyses were recorded on a Bruker Instrument AVANCE400 spectrometer (Milano, Italy). Acquisition and data treatment were performed with Bruker TopSpin 3.2 software (Milano, Italy). The spectra were collected at 29 ◦C with a 4-s acquisition time, 5-s relaxation delay, and 256 scans. Prior to analysis, samples were dried for 24 h under vacuum and then derivatized according to the following procedure.

A lignin sample (40 mg) was completely dissolved in 300 μL of *N*,*N*-dimethylformamide. To this solution, the following components were added: 200 μL of dry pyridine, 100 μL of solution of internal standard (10 mg of Endo-*N*-hydroxy-5-norbornene-2,3-dicarboximide (Sigma 226378) dissolved in 0.5 mL of a mixture of pyridine and CDCl3 1.6:1 *v*/*v*), 50 μL of solution of relaxation agent (5.7 mg of chromium (III) acetylacetonate (Sigma 574082) dissolved in 0.5 mL of a mixture of pyridine and CDCl3 1.6:1 *v*/*v*), 100 μL of 2-chloro-4,4,5,5-tetramethyl-1,3,2-dioxaphospholane (Sigma 447536), and at the end 200 μL CDCl3. The solution was centrifuged if necessary. All chemical shifts reported were related to the reaction product of the phosphorylating agent with water, which gave a signal at 132.2 ppm.

#### *3.10. Fourier-Transform Infrared Spectroscopy (FT-IR)*

FT-IR spectra of all lignin fractions were recorded in transmission mode on lignin-containing KBr pellets (amount of lignin was approximately 5 mg). The analysis was performed by means of a Nicolet 760-FTIR spectrophotometer (Thermo Fisher Scientific, Rodano, Italy) at room temperature in air in the <sup>4000</sup>−700 cm−<sup>1</sup> wavenumber range with 64 accumulated scans and a resolution of 2 cm<sup>−</sup>1.

#### *3.11. Di*ff*erential Scanning Calorimetry (DSC)*

DSC was performed on solid-state samples (about 10–15 mg) by means of a Mettler-Toledo DSC/823e instrument (Milano, Italy) at a scan rate of 20 ◦C/min under nitrogen flux.

#### **4. Conclusions**

Lignin constitutes a byproduct accumulated worldwide in significant quantities in the pulp and paper industry, despite possessing a great potential for generating novel biobased platform chemicals and polymeric materials alternative to those currently derived from petrochemical routes. One of the most important tasks when targeting the establishment of any lignin valorisation process is the ability to set up robust and reliable fractionation processes enabling the recovery of lignin streams with controllable and predictable composition and properties. Indeed, the great variability in its chemical composition, structural characteristics, molecular weight, and reactivity, intimately linked to the vast geographical and seasonal distribution of lignocellulosic biomass feedstocks, makes lignin processing and valorization particularly challenging.

In an attempt to address this issue, in this work, a straightforward fractionation method was demonstrated combining solvent extraction and membrane-assisted ultrafiltration. This approach was applied to two commercially available technical lignins with different origins (non-woody vs. woody biomass) obtained from two distinct delignification routes (soda vs. kraft pulping), starting from an extraction process using two different solvent (aqueous vs. organic) systems. Based on this strategy, a comprehensive method was demonstrated for the tunable and predictive control over the characteristics of the recovered lignin fractions. In particular, the key role played by the ultrafiltration process in providing consistent and reproducible fractions of narrowly controlled chemical, structural, molecular, and thermal characteristics was shown. In addition, the performance of the ultrafiltration membranes was found to be relatively unaffected by the solvent used (either aqueous or organic), indicating a relative independence between the extraction operation (controlled by the solvent) and the downstream fractionation (controlled by the ultrafiltration membranes). This aspect is a very relevant feature from an industrial scale-up perspective, because it represents a clear advantage in terms of process tunability and optimization when tackling different types of biomass as input feedstock.

This versatile, robust, and easy to implement fractionation approach paves the path for the delivery of consistent and reproducible lignin fractions of well-controlled and predictable properties irrespective of the biomass origin, with this aspect being of particular interest in view of their potential use in the development of lignin-based macromolecular materials.

**Author Contributions:** Conceptualization, P.D., G.G., S.T., J.T. and A.S.; investigation, C.A., O.B., L.R. and J.T.; writing, C.A., P.D., G.G., A.S. and J.T.; supervision, P.D. and G.G.; funding acquisition, P.D. and G.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received funding from Regione Lombardia and Fondazione Cariplo, through grant number 2018-1739 (project POLISTE).

**Acknowledgments:** The authors kindly acknowledge Maria Elisabetta Brenna for instrumental support in GC-MS analyses and Francesco Gilberto Gatti for help in NMR analyses optimization.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


**Sample Availability:** Samples of the compounds are not available from the authors.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
