Clustered single cellulosic fiber dissolution kinetics and mechanisms

Mar 27, 2018 - Data processing was performed using the Python language, utilizing available scientific libraries. The methods of processing the data i...
0 downloads 6 Views 1MB Size
Subscriber access provided by UNIV OF DURHAM

Clustered single cellulosic fiber dissolution kinetics and mechanisms, through optical microscopy under limited dissolving conditions Valtteri Mäkelä, Ronny Wahlström, Ulla Riikka Maria HolopainenMantila, Ilkka Antero Kilpeläinen, and Alistair William Thomas King Biomacromolecules, Just Accepted Manuscript • DOI: 10.1021/acs.biomac.7b01797 • Publication Date (Web): 27 Mar 2018 Downloaded from http://pubs.acs.org on March 29, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

Clustered single cellulosic fiber dissolution kinetics and mechanisms through optical microscopy under limited dissolving conditions Valtteri M¨akel¨a,∗,† Ronny Wahlstr¨om,∗,‡ Ulla Holopainen-Mantila,‡ Ilkka Kilpel¨ainen,† and Alistair W. T. King∗,† †Department of Chemistry, University of Helsinki, PO Box 55, FI-00014 University of Helsinki, Finland ‡VTT Technical Research Centre of Finland Ltd, PO Box 1000, FI-02044 VTT, Espoo, Finland E-mail: [email protected]; [email protected]; [email protected]

Abstract Herein, we describe a new method of assessing the kinetics of dissolution of single fibers by dissolution under limited dissolving conditions. The dissolution is followed by optical microscopy under limited dissolving conditions. Videos of the dissolution were processed in ImageJ to yield kinetics for dissolution, based on the disappearance of pixels associated with intact fibers. Data processing was performed using the Python language, utilizing available scientific libraries. The methods of processing the data include clustering of the single fiber data, identifying clusters associated with different fiber types, producing average dissolution traces and also extraction of practical parameters, such as, time taken to dissolve 25, 50, 75, 95 and 99.5 % of the clustered fibers. In addition to these simple parameters, exponential fitting was also performed

1

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

yielding rate constants for fibre dissolution. Fits for sample and cluster averages were variable, although demonstrating 1st order kinetics for dissolution overall. To illustrate this process, two reference pulps (a bleached softwood kraft pulp and a bleached hardwood pre-hydrolysis kraft pulp) and their cellulase-treated versions were analysed. As expected, differences in the kinetics and dissolution mechanisms between these samples were observed. Our initial interpretations are presented, based on the combined mechanistic observations and single fiber dissolution kinetics for these different samples. While the dissolution mechanisms observed were similar to those published previously, the more direct link of mechanistic information with the kinetics improve our understanding of cell wall structure and pre-treatments, towards improved processibility.

Introduction Cellulose is the most important structural constituent in plant fiber cell walls, typically composited together with lignin and hemicelluloses. Crystalline cellulose is a homopolymer of cellobiose. Its crystallinity and morphology, in the cell wall or often as a regenerated fiber, affords materials with a high degree of rigidity and strength. This is why cellulose historically has found wide application as a fibrous material in various products. It is expected that cellulosic materials are to again grow in importance as we move towards a renewable industrial economy. However, the serious limitation of technical celluloses (chemical pulps) in application is their lack of a melting point, precluding low-cost melt-processing. Cellulose, as a polymer, is also insoluble in most common non-derivatising molecular solvents, requiring more complex or derivatising solvents to plasticise or dissolve. Accordingly, the swelling and dissolving mechanism of cellulosic plant fibers has been studied widely by cellulose and fiber chemists for more than 150 years. Early reports established the non-homogeneous swelling of cellulose fibers and ’ballooning’ as a mechanism of swelling. 1 In the 1950s, Hock proposed that ballooning in cotton fibers is caused by swelling of the secondary cell wall (S1, S2), causing rupture roughly in 2

ACS Paragon Plus Environment

Page 2 of 35

Page 3 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

the middle and peeling back of the whole primary cell wall (PW), 2,3 along the fibre length. Restriction to swelling in the PW was suggested to result from the ’netlike arrangement of cellulose’ (fibrils). In S1 and S2 in cotton, the fibrillar angle is more lateral to the fiber length (25-45◦ ) which was thought to offer less restriction to swelling. 2 More recently still, ballooning has partly been explained through other mechanisms. The balloons have been proposed to consist of dissolved cellulose contained under pressure inside thin membranes. 4,5 It is thought that as long as the outer cell wall layer is present, the dissolution behaviour of natural fibers is at large similar, independent of the fiber source and more dependent on the quality of the cellulose solvent. 6 However, Hock originally proposed that more irregular ballooning was to occur for wood fibers due to changes in fibrillar angle between inner and outer cellulose layers. 2 Nevertheless, both wood chemical pulp and cotton fibers can be dissolved completely in a suitable solvent. By adjusting the solubilising power of cellulose solvents, simply by diluting the actual dissolving agent, or adding an anti-solvent, the swelling and dissolution efficiency of the solution can be tuned to an appropriate level to target different applications or studies. Controlled dissolution studies have been performed, e.g. to study the morphology of different fiber types, including regenerated fibers such as Lyocell fibers. 7 Interestingly, the swelling and dissolution mechanisms have also been found to be similar for cellulose derivative fibers, such as nitrocellulose or xanthate fibers. 8 In the seminal work by Cuissinat and Navard, five different modes of dissolution and swelling were identified when immersing a cellulose fiber in varying NMMO or sodium hydroxide water mixtures. This was performed to control the dissolution power and identify mechanisms of dissolution: 4,5 Mode 1) Rapid dissolution by disintegration into fragments, Mode 2) Swelling by ballooning followed by complete dissolution, Mode 3) Swelling by ballooning without dissolution, Mode 4) Homogeneous swelling with no dissolution, and Mode 5) No swelling or dissolution. Removal of the outer cell wall layers by, e.g. enzymatic treatments has been shown to reduce or remove ballooning and lead to faster and homogeneous dissolution. 9 Certain

3

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

chemical treatments, such as TEMPO oxidation, have also been found to reduce ballooning, possibly by derivatizing the outer cell wall layers to a more soluble form. 10 Paper-grade pulps typically still have their PW and S1 remaining whereas these layers are removed during the production of dissolving pulps, which reflects on the ballooning behavior during swelling and dissolution of the two pulp grades. 11 Fiber swelling and dissolution mechanisms, and to a smaller extent kinetics, have been studied by light microscopy while dissolving fiber samples using solvents with limited cellulose dissolution capacity. Typically, only a few fibers have been studied at a time, on the microscope slide, to avoid too much dilution at the fiber surface (with never-dried pulp samples), or other concentration gradients that might limit the kinetics of dissolution, efficacy of the solvent or repeatability of the measurement. In the NMMO-water system it has been proposed that the dilution of the solvent close to the fiber, by water released from never-dried pulp, is the main limiting factor of the dissolution. 12 NMMO-water has been a popular solvent system for fiber dissolution studies, 7,10 but parallel studies have been conducted with different ionic liquids, 8,13 cupriethylene diamine (CED) solutions, 14 and alkaline aqueous systems with various added chemicals to adjust cellulose solubility. 4,5,9 The approach to study only one or a few fibers has been criticized for not giving a representative picture of the whole pulp and its behavior under swelling and dissolution conditions. 14 The analysis of single fibers typically give only qualitative data, and as many pulps contain several different fiber types, large errors might thus follow from using this approach. As an improved method to study fiber swelling, Arnoul-Jarriault, et al., proposed combining a short pulp treatment in low concentrations (< 0.2 M) of CED followed by characterization of the pulp swelling using a ’MorFi’ fiber analyser. 14 In this approach, a set of several thousand fibers are analysed and the result can be said to be truly representative for the bulk pulp. However, the main limitation is that the method is suitable for studying fiber swelling but dissolution only to a lesser degree. A few studies on dissolution kinetics have been published, with focus on fiber dissolution time (complete dissolution) under limited

4

ACS Paragon Plus Environment

Page 4 of 35

Page 5 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

dissolving conditions. 7 Fiber swelling kinetics has also been experimentally described through the use of three different parameters, determined by microscopy, as a function of swelling time: the total fiber radius, the radial position of the swelling front, and the thickness of the swollen region. 7 An overview of the different methods used and materials analysed is given in Table 1. Table 1: Review of different fiber dissolution analysis methods using modern microscopy equipment. Used abbreviations: NMMO (N -methylmorpholine-N -oxide), CED (cupriethylene diamine), NaOH (sodium hydroxide), ZnO (zinc oxide), ILs (Ionic Liquids), [emim][OAc] (1-ethyl-3-methylimidazolium acetate, an ionic liquid), Ph-K (Prehydrolysis Kraft pulp), K (Kraft pulp), CCE (Cold Caustic Extracted), TEMPO (TEMPO oxidized), SAQ (soda anthraquinone pulp, totally chlorine free), Ac (acetate grade), H (sodium hypochlorite treatment), B (sodium borohydride treatment), enz (cellulase treatment). Method NMMO, Xanthation, CED NMMO CED

Cellulose Type Ph-K, K-CCE, TEMPO-K-CCE Lyocell Fibers Ph-K, Ph-SAQ, SAQ-CCE, K-CCE NaOH, Urea, ZnO, NMMO Cotton, Wood Cellulose Fibers NaOH Sulfite Pulp ILs Cellulosic Pulps NaOH, NMMO, ILs, Lyocell Fibers Etherification, Xanthation [emim][OAc]-Water Ph-K, Sulfite Pulp, MCC CED with MorFi Ph-K, Ph-K-CCE, Ph-K-ECF, Ph-K-TCF, Ac-Ph-K, K, Ac-Sulfite, Ac-Sulfite-H-B Ac-Sulfite-enz

Reference Gehmayr et al. 10 Chaudemanche et al. 7 Schild et al. 11 Cuissinat et al. 5 Le Moigne et al. 9 Parviainen et al. 13 Cuissinat et al. 8 Olsson et al. 15 Arnoul-Jarriault et al. 14

Therefore, we wish to detail herein a new microscopy-based method of determining fiber dissolution kinetics and mechanisms, based upon dissolving small samples of fibers (7-23 fibers per sample) from each material into a limited dissolution efficiency CED solution. The dissolution is performed by dissolving 1-3 fibers each time and followed by recording a moderate resolution and magnification video on the optical microscope stage. Frames are extracted from the video and the kinetics of dissolution are parameterised by loss of fiber pixels. Partially scripted processing allows fast and robust acquisition, and a comparison of simple kinetic parameters. 5

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 35

Experimental Characterisation of fiber materials A bleached softwood kraft pulp (BS-K) and a bleached hardwood dissolving (pre-hydrolysis kraft) pulp (BH-PhK), which were used as base pulps for all the samples, were obtained from two large Finnish pulp mills. Several standard analyses were performed to acquire the key characteristics of the pulps (Table 2). Viscosity was measured according to the standard procedure ISO 5351. Lignin and carbohydrates were analyzed according to the NREL method TP-510-42618. Fiber type analysis was performed according to the standard ISO 9184. The earlywood and latewood fibers could not be distinguished by this method in the BH-PhK sample, so the fiber type distribution could only be obtained from BS-K. For both base pulps and treated samples the weight average molecular weight (Mw), number average molecular weight (Mn) and polydispersity index were measured in the LiCl/DMAc solvent system by size-exclusion chromatography (SEC), with sample preparation and instrumentation as described by Wahlstr¨om et al., 16 with ethylisocyanate (10 vol-%) added into the 8 % LiCl/DMAc to facilitate the pulp dissolution. In addition, the fiber dimensions were analyzed by automated fiber analysis using an STFI Fibermaster device by Lorentzen & Wettre. Table 2: Basic characterisation of the analyzed pulps Analysis Pulp type Wood type Earlywood fibers (%) Latewood fibers (%) Lignin Content (wt%) Cellulose Content (wt%) Xylan Content (wt%) Glucomannan Content (wt%) Intrinsic Viscosity mL/g

BS-K BH-PhK Method bleached softwood bleached hardwood kraft prehydrolysis kraft pine, spruce birch, aspen 64 ISO 9184 36 0.9 1.2 80.3 93.4 NREL/TP-510-42618 10.4 5.4 8.4 0.0 810 460 ISO 5351

6

ACS Paragon Plus Environment

Page 7 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

Pulp treatments The dry pulp sheets were homogenized by soaking them in water (1 g pulp/ 66 g of water) for 4 h and disintegrating the wet pulp using an electric kitchen mixer, followed by dewatering by vacuum filtration on a 60 µm wire cloth. The exact dry matter content was determined as the mass loss of three parallel pulp samples when kept at 105 ◦ C overnight. The enzymatic treatment was performed with a glycosyl hydrolase (GH) family 5 endoglucanase, obtained from ROAL Oy (Rajam¨aki, Finland). The homogenized pulp was redispersed in 50 mM citrate buffer (pH 5, adjusted with NaOH) to a final consistency of 2.5 % (wt.). The slurry was pre-heated at gentle stirring to 50 ◦ C, the enzyme was added (dosage 7 mg protein/g of dry pulp) and the enzymatic treatment was continued for 3 h at 50 ◦ C, whereafter the slurry was vacuum filtered and the residual enzyme was inactivated in the pulp by heating to 95 ◦ C for 20 min.

Imaging of fibers during CED treatment The pulp sample fibers were first separated in distilled water in a microtube by manual shaking. A drop of water with separated fibers was added on a microscopy slide and diluted and washed further with distilled water in order to have a few individual fibers on each microscopy slide. The fibers were orientated with a metal spike under a stereomicroscope and left to dry in +30 ◦ C for 5-10 min. Then the fibers were detached from the surface of the slide and re-orientated with a metal spike. The cover slip was added on top of the fibers and it was attached with nail polish from the corners, thus, leaving a gap for influx of the cellulose solvent. The microscopy slide with covered fibers was positioned for imaging with confocal laser scanning microscopy (CLSM) equipment consisting of a Zeiss LSM 710 (Zeiss, Jena, Germany) attached to a Zeiss Axio Imager.Z microscope. A drop of aqueous 0.2 M cupriethylenediamine (CED; Oy FF-Chemicals Ab, Haukipudas, Finland) was added to the microscopy slide and the solution reached the fibers due to capillary effect caused by the space 7

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

between the microscopy slide and the cover slip. A transmission image was captured after the CED solution addition at time intervals of approximately 1.5 s using time series mode until the fiber finally was dissolved. The images were taken with a 5x objective (EC Plan-Neofluar, numerical aperture of 0.16) using a 405-nm diode laser line, a transmission photomultiplier tube (T-PMT) and a polarizer in the angle of 75◦ . The resolution of the images was 512x512 and pixel size 3.3 µm. When scanning, each pixel was shot by the laser for 2.55 µs and the scanning of the whole image took 0.67 s. At least ten fibers per sample should be analyzed on separate microscopy slides to get average dissolution traces for the pulp but more detailed analysis may require up to fifty fibers per sample.

Processing of the imaging data The overall workflow for the analysis of the imaging data is described in Scheme 1. The image sequences produced by microscopy were analyzed with ImageJ 2 software (Fiji package), 17 with the aim of extracting the fiber area in pixels from each image. This was achieved using a threshold function, which classifies each pixel into either pure white or black, and then calculating the number of black pixels. The overall processing sequence for each video was as follows: 1. Loading and cleaning the video: The video/image stack was opened, and the frames before the introduction of the solvent and the last frame containing a black bar artifact due end of scanning were discarded. 2. Fiber crop/cleaning fiber surroundings: A single fiber was selected with the polygonal selection tool, and the pixels outside of the approximate region were filled with pure white (Figure 1). The fiber area was selected so that the slight movement of the fiber during dissolution was contained. 3. Fiber area extraction: A threshold function was applied to the image stack and the resulting data saved to the disk in CSV format (Figure 1). A maximum entropy thresh8

ACS Paragon Plus Environment

Page 8 of 35

Page 9 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

1

Microscopy

512x512 grayscale video 1.5s interval, AVI container 1-3 fibers/video

2

Extraction

ImageJ (Fiji package)

CSV file 1 file/fiber

3

Processing & analysis solna (custom python)

Dissolution trace plots Extracted parameters (png/pdf figures) .csv files for spreadsheet etc.

Scheme 1: The overall workflow for the analysis imaging data. The microscopy images are analyzed as individual frames in the ImageJ software yielding the area of the fiber in each frame. Data from several fibers is then processed and combined to yield images and various kinetic parameters.

9

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 1: Fiber area selection for cleaning surroundings (top) and applying a threshold function to calculate fiber area (bottom) in the processing of microscopy imaging data using ImageJ software. 17

10

ACS Paragon Plus Environment

Page 10 of 35

Page 11 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

old function was used, 18 or alternatively a Yen function if too much of the background was picked by the function. 19 The processing was aided by creation of several ImageJ macros to allow quick access to the needed features via function keys and saving of the extracted data, which are available in a public BitBucket repository. 20 Using these aids, an experienced person could process a single video in less than minute. The processing workflow is also illustrated in a supplemental video. 21

Data analysis and plotting The resulting CSV files were analyzed with a custom program, solna, specifically created for the task, written in Python and utilizing NumPy, SciPy, Pandas and Matplotlib libraries. 22–25 The processing was completed in two stages with two separate tools: solna extr and solna proc. Both tools along with instructions are available at the BitBucket repository with a BSD license. 20 A summary of the function and capabilities of the tools are presented below, with more details about the processing available in the supplementary information. Stage 1: solna extr The solna extr is a simple preprocessing tool for the csv output files produced by ImageJ, designed to transform the dissolution data into a neat and consistent format. The preprocessing consists of simple operations, such as, filename filtering, adjusting the frame rate (if needed) and calculating normalized relative areas. All fiber dissolution traces from a single sample are then stored in a single NumPy NPZ file, which makes it easy to handle the data for each sample. Stage 2: solna proc The solna proc can be used to read one or several NPZ files, do additional processing, calculate several parameters and render different types of figures. A brief summary of the 11

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

processing pipeline is given below. An extended summary is given in the supplimentary information: 1. Preprocessing: Resample the time axis, smoothing (moving average, optional) etc. 2. Clustering (optional): Grouping of the dissolution traces automatically (k-means using cumulative sum), or for example separation into two clusters (fast/slow), by a given hard time limit. 3. Trace calculations: Dissolution trace averages (for all dissolution traces and traces in the same cluster), cumulative sum for each trace, discrete gradients (1st order difference) for trace averages etc. Average is calculated by taking the average of each trace in each (resampled) time point. 4. Fitting and parameter calculations: Initial dissolution speed and dissolution time limits from the trace and cluster averages. The initial speed is calculated by a linear fit of data points in which the first 20% of the total area is dissolved. An exponential decay fitting, using the Levenberg-Marquardt algorithm, is implemented in the SciPy library. 25 Dissolution time limits, calculated by determining the time point in which 25, 50, 75, 95 and 99.5% of the total area is dissolved, using the average or cluster average dissolution trace. Individual fiber dissolution times are based on 95% dissolution. All calculated parameters are printed and written into CSV files for further analysis. 5. PNG/PDF plots (optional): Fiber dissolution traces/averages, dissolution traces with cluster averages etc. All of the dissolution kinetics figures presented in this paper are created with solna proc. 6. Data export (optional): The data can also be exported in Matlab or CSV format, for plotting in spreadsheet or other data-analysis software. For the data presented in this article, default moving average smoothing of 1201 pts (12 sec) was used on all data. Any other relevant configuration options used in the processing, 12

ACS Paragon Plus Environment

Page 12 of 35

Page 13 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

such as the used clustering method, are mentioned in the conjuction of the figure and/or tabulated data.

Results and discussion Method Development Only very few studies focus on dissolution kinetics of cellulosic fibers. Chaudemanche et al. 7 studied the dissolution of Lyocell fibers using NMMO-water mixtures by measuring their total dissolution time. The transition between swelling and dissolving conditions, as function of water content, was sharp and there was, thus, no concentration region of partial dissolution. 7 In the optimum case, the fibers dissolved rather rapidly (< 1 min) and homogeneously. However, rapid and homogeneous dissolution in this case is to be expected as the fibers are regenerated cellulose, with a far more primitive morphology (low crystallinity, cellulose II and mainly parallel orientation of both micro and macrofibrils 26 ), than native plant cell fibers which have evolved to be recalcitrant. In our work, the paper-grade BS-K fibers in particular dissolved slowly and had poorly soluble regions. However, we require faster dissolution kinetics to avoid prolonged dissolution times with recalcitrant fibers but with slow enough dissolution, to allow for resolution of the key kinetic features that limit dissolution, yet allowing for complete dissolution to observe the full fiber dissolution kinetics. Ionic liquids or NMMO monohydrate are obvious choices as they are powerful cellulose solvents, leading to homogeneous mixtures. However, the melting point of NMMO monohydrate is 74 ◦

C, 4,5 necessitating heating and other special equipment. Ionic liquids, like NMMO mono-

hydrate, also have high viscosity and surface tension. For the cellulose dissolving varieties, viscosity typically ranges between 100-10000 cP at room temperature. 27 Even for the lowest viscosity ionic liquids, this would limit dissolution kinetics, requiring high temperature to have any reasonable rate of dissolution. Higher viscosities could also induce concentration gradients reducing local swelling, when in contact with never-dried pulps. This could be 13

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

limited when using an aqueous based solvent as diffusion of and in water is much faster, with viscosities below 1 cP at room temperature. The high surface tension of ionic liquids would also be problematic, as soaking of the fibres would be slow and thick layers of ionic liquid would provide significant refraction of light, which would make focusing on the fibres difficult. Thus, the next logical choice was application of aqueous solutions of CED, which can be utilised at RT with different concentrations to control the kinetics. 10 Both viscosity and surface tension are low, allowing for rapid soaking of the sample. Of course CED has also been well studied as a media for determining ’intrinsic viscosity’, linked to molecular weight, of chemical pulps. Arnoul-Jarriault et al 14 developed a test for dissolving pulp reactivity in which swelling in diluted CED solution was combined with fiber analyser analysis to determine the degree of swelling. The method was well correlated to the typically used Fock analysis but was only suitable for assessing pulp swelling and not dissolution, as dissolving fibers cannot be analysed by the fiber analyser. The Fock test is a quite tedious laboratory method in which the viscose production process is simulated, and involves the use of CS2 . The above mentioned authors also found that pulp fibers dissolved slowly, with the common ballooning behaviour, at 0.2 M CED molarity, whereas at 0.25 M CED molarity fibers dissolved rapidly without ballooning. Thus, there seemed to be a fine CED molarity window (between 0.2 and 0.25 M) where high purity and molecular weight cellulosic fibers will dissolve in a slow and controllable manner. In our work, these results were mostly confirmed in independent experiments and actually the lower concentration of 0.2 M CED at high dilution (a few fibers dispersed in bulk CED solution) was chosen as fiber solvent. This allowed for enough dissolving power to differentiate fiber types kinetically, yet achieve complete dissolution in all cases. As dissolution of cellulose in higher molarities of CED is rapid, the lower molarity allowed for a prolonged dissolution period where different kinetic regimes could be more easily resolved. This better resolution also prompted us to introduce the 95 % dissolution time as a key parameter to assess the dissolution speed. This helps to avoid bias in total

14

ACS Paragon Plus Environment

Page 14 of 35

Page 15 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

dissolution times due to the presence of a small proportion of fibers that contain recalcitrant regions that take much longer to dissolve than the bulk. These poorly soluble regions may have a large impact on the pulp filtration value and likely constitute the insoluble fraction in many cellulose solutions. 6 It should be kept in mind that ideally this method could be optimised for a range of swelling, direct-dissolution and reactive dissolution solvents as each may have different mechanisms of dissolution.

Observed Mechanisms As determined in previous publications 1–5 there are several distinct mechanisms described, e.g. ballooning, which are also separated by dissolution or incomplete dissolution, depending on the solvent dissolution power. We have observed similar mechanisms for our base pulps BS-K, BH-PhK and enzyme-treated samples but would also like to comment on a few additional features: 1. Ballooning: Ballooning, with different degrees of symmetry, is a common feature to both the BH-PhK and BS-K samples (Figure 5). However, it is much more prevalent with the BS-K pulp. Hock 2 previously suggested that the differences in the symmetry of the ballooning were due to more extreme differences in fibrillar angle between different cell wall components. Ballooning is thought to occur when there is very minimal but uniform restriction in swelling of the cell wall, after splitting of the primary cell wall orthogonal to the fiber axis. This explanation was originally proposed by Siu 3 and Hock. 2 Le Moigne et al. 9 have nicely illustrated this, showing that when the primary cell wall splits, bands of recalcitrant material, described as ’disks’, form sequentially along the fiber axis. This allows the S1 and S2 secondary cell wall layers to swell, leading to the ballooning effect. As these disks accumulate in solution they offer a stepwise dissolution profile for slower dissolving fibres. This can be observed most plainly for the charted kinetics of the slowest dissolving fibres from the BS-K (Figure 2c). The mechanism causing the step-wise dissolution kinetics is also confirmed by inspection of 15

ACS Paragon Plus Environment

Biomacromolecules

BS-K

BS-K (clusters)

C 99.5% dissolved

A

Trac e

avg .

Slow

Fa s

t

Initial speed

BH-PhK

BH-PhK (clusters)

D

99.5% dissolved

B

s Fa

Trace avg.

t

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 35

Slow

Initial speed

Figure 2: Dissolution traces from the untreated base pulp samples, with all dissolution traces (A, B) and the traces clustered into two clusters (C, D). In the left plots (A, B), the average of the traces with a linear fit for initial speed and 99.5% dissolution time are also shown. In the right plots (C, D), the traces are clustered into two clusters (fast/slow dissolving fibers), with the averages of the clusters shown emphasizing the difference in dissolution speed.

16

ACS Paragon Plus Environment

Page 17 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

BH-PhK vs BS-K

BH-PhK vs BS-K (only averages)

BS-

K BH-PhK

Figure 3: Comparison of the untreated base pulp samples (BH-PhK and BS-K), with the fiber dissolution traces and averages (top) and only the dissolution trace averages with 95% confidence interval shaded (bottom). The faster dissolution speed of the dissolving pulp sample (BH-PhK) is evident from both plots, however the differences in the dissolution profile can be more clearly seen when comparing only the average dissolution traces (bottom plot). The confidence intervals for the dissolved area (y-axis) can be optionally plotted to visualize the spread and uncertainty associated with the average dissolution traces.

17

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 35

BH-PhK / BS-K, exponential fit

BH

BS-

K

-P

hK

BH-PhK / BS-K exponential fit, log scale

BS-

K

BH

-P

hK

Figure 4: The 1st order reaction kinetics (exponential decay) fitted to the BS-K and BHPhK samples with linear (top) and logarithmic (bottom) y-axis. The the exponential fits are marked with solid black lines. The logarithmic y-axis emphasizes the errors in lower part of the dissolution trace. The curve fitting was done in the linear space to avoid disproportionate weighting of the lower part of the trace.

18

ACS Paragon Plus Environment

Page 19 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

A1

A2

B1

B2

C1

C2

D1

D2

Figure 5: Main mechanisms of fiber dissolution: A) ballooning (BS-K), B) helical unwinding (BS-K), C) helical unwinding (BH-PhK), D) homogeneous dissolution (BS-K)

19

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

the relevant microscopy video over the appropriate frame (time) range. In this case, the slowest dissolving fibre forms a plateau at a relative area level of 0.2, beginning and ending at approximately 480 and 520 s respectively, after which it rapidly goes to zero. This initial plateau is due to accumulation of disks, after ballooning, which all seem to disperse at once, around 520 s. The reason for this concerted dispersal of residual disks may be statistical or there may be some underlaying kinetic mechanism, as yet not understood. In our analyses, these disks take much longer to disperse than the remaining contents of the cells. This can be observed from both microscopy images (Figure 5) and the kinetic data (Figure 3) where especially the fibers in the slow dissolving cluster can be observed to disappear in a step-wise fashion, towards the end of the dissolving period. Le Moigne et al. 28 have also determined that there is a higher ratio of mannans in the insoluble fractions, which suggests a possible complexation with hemicelluloses also restricting dissolution. This may contribute to why kraft pulps are more difficult to dissolve than sulfite, pre-hydrolysis kraft and other caustic or acid-treated pulps. Further detailed fiber fractionation studies are required to understand the effect of fiber orientation and hemicellulose (and possibly residual lignin) content. 2. Helical Unwinding: Formation of helicies is also common to both pulps (Figure 5). Helical unwinding can be described as a more asymmetric ballooning of fibers. The helical structure is clearly influenced by the forces asserted by the swelling of the underlaying cell walls, with fibrillar angles of the S1 layer least parallel to the fiber axis and S2 layer almost parallel with the fiber axis. Longitudinal splitting of the primary cell wall may also contribute to the helical unwinding effect. 3. Homogeneous dissolution: There are a few cases where the fiber simply swells and disperses, with minimal ballooning. This occurs mainly from the fiber ends and seems to be linked to the cell wall thickness. This was the most common dissolution feature

20

ACS Paragon Plus Environment

Page 20 of 35

Page 21 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

after the cellulase treatments. 4. Dissolution from the Fiber Ends: In all cases, swelling and dissolution seems to occur predominantly from the fiber ends (additional figures in supplementary material). It initially looks like the ends of the fibers are the main point of entry for the CED reagent. Swelling and dissolution proceeds from the ends to the middle, as does ballooning or helix formation. This would assume that the CED solution would rapidly diffuse through the lumen. However, there is no direct evidence for this, except for the more rapid swelling from the fiber ends. An alternative explanation is that this may also be due to thinner cell walls at the fiber ends, with reagent partially diffusing through the cell wall. Further studies need to be done to trace the kinetics of diffusion of the reagent through the fibers and this may vary from reagent to reagent. One must also remember that fibers may be cut during prior processing which may or may not allow for more rapid penetration at the ends. Clearly this should be studied systematically. 5. Fiber Fragmentation: In the cases where the fibers are enzymatically treated there are some fibers that fragment homogeneously along the fiber axis (additional figures in supplementary material), i.e. no swelling from the fiber ends. This would be strong evidence for removal of some cell wall component, allowing for direct diffusion of the CED reagent through the fiber cell wall and a more homogeneous dissolution. Likely, other factors such as the reduced fiber dimensions, cellulose molecular weight and changes in the fiber cell wall morphology might play their own part in the change of dissolution mechanism (Table 3). As the ballooning effect seems to occur where the secondary cell wall is not restricted, this fragmentation might suggest that the molecular weight of the cellulose fraction is also significantly reduced, within the secondary cell wall fibrils.

21

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 35

Table 3: SEC and Fibremaster fiber analyzer results. PDI is the polydispersity index. Sample BS-K Treatment Enzyme Mw (g/mol) 775700 362000 Mn (g/mol) 45800 46600 PDI 16.9 7.8 Fiber length (mm) 2.01 0.69 Fiber width (µm) 29.45 30.25 Fines (%) 6.45 14.53

22

BH-PhK Enzyme 391000 306800 62100 47200 6.3 6.5 0.74 0.45 19.80 20.75 7.71 25.71

ACS Paragon Plus Environment

Page 23 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

Dissolution Kinetics for the Untreated Pulps When we compare the dissolution kinetics for the BS-K and BH-PhK pulps (Figure 2 and 3) it is obvious that the BH-PhK fibers dissolve faster and more uniformly (similar kinetics) than the kraft fibers. Considering the prevalence of the observed mechanisms for each pulp (more ballooning with the kraft pulp), this may be due to a stronger retention of the primary cell wall when the prehdyrolysis treatment is absent, allowing for restricted dissolution. Alternatively, it may be the additional contribution of the higher molecular weight or the larger fiber dimensions of the BS-K pulp or even the presence of hemicellulose (Table 3). From the clustering of the data for each pulp, we see that the BS-K pulp nicely forms two clean clusters using the automatic function (K-means clustering), whereas the BH-PhK requires manual clustering (time clustering with 175 sec cutoff, fibers with at least 10 % undissolved at cutoff are classified as slow) to give more meaningful clusters as the few outlier fibers are not sufficiently distinct from the main group (Figure 2). Fiber analysis (standard method ISO 9184) was performed for the BS-K showing 64 % earlywood and 36 % late wood (Table 2). Calculating the fractions of fibers in each of the clusters yields very similar values of 67 % for cluster 1 (fast dissolving) and 33 % for cluster 2 (slow dissolving). This gives a very good indication that clustering can separate the fiber types. Clearly, part of the faster dissolution kinetics seems to be related to the faster dissolution of the BS-K earlywood vs latewood. This is supported by the report that thin-walled springwood fibers dissolve faster than thick-walled latewood fibers in ionic liquids, as observed by microscopy. 13 Furthermore, analysis of the videos uncovered slight mechanistic differences between the clusters. Some degree of ballooning or helix formation was observed for both fiber-types but for the latewood fibers, the initial swelling of the cell walls, at the early stages of dissolution, was much slower than for the earlywood fibers. This was then followed by mainly slow ballooning of the fibers and slow stepwise dispersion of the remaining disks. The earlywood fibers tended to display a rapid swelling of the cell wall, followed by rapid dispersion via all of the observed mechanisms. It was not possible to do the fiber analysis for the BH-PhK to 23

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

yield the proportion of earlywood and latewood. This is not surprising as a typical Finnish hardwood pulp, as is BH-PhK in this study, is mainly birch (Betula pendula), with aspen as a minor component (< 5 wt. %). Birch is a diffuse porous hardwood, which typically has little visible difference between the earlywood and latewood. This would also explain why the kinetics for BH-PhK does not exhibit well separated clusters; the BH-PhK has some slow dissolving fibers but they do not cluster to the same extent as for the kraft pulp and seem to be fewer in number. This also clearly demonstrates that the amount of analyzed fibers per sample should be large enough to have a good statistical representation of fibers from all types and dissolving rates. Looking at the individual fiber dissolution traces allows for rationalization of certain mechanisms, but ultimately comparing the average dissolution traces is the clearest way to visualize the differences between the pulps (Figure 3). For practical purposes, the average dissolution trace can be represented by a selection of dissolution times, which express the time it takes to dissolve a certain fraction of the sampled fibers, forming a coarse profile of the average dissolution trace. In our case, we decided to report values for 25, 50, 75, 95 and 99.5 % dissolution (Table 4). Of these, 95 % dissolution time seems to be the most robust single value describing the dissolution speed, including both fast and slow dissolving fibers. The 25 and 50 % values are mostly affected by the initial speed and fast dissolving fibers, while the 99.5 % value can represent the expected ”total” dissolution time more accurately. However, the 99.5 % dissolution time is easily affected by the number and dissolution profile of the few slowest fibers, so it should be used with caution. Of course, obtaining as large datasets as practical for each pulp is the best way to minimize these errors. Measuring 25-30 or more fibers per pulp should provide a robust dissolution profile, while even as few as 10 fibers seem to already distinguish large differences between pulps. Ultimately the number of fibers needed is dependent on their uniformity: if the pulp contains fibers with a greatly varying dissolution profile, more fibers are needed to measure an average dissolution profile and perhaps see the clustering of distinct fiber types. For

24

ACS Paragon Plus Environment

Page 24 of 35

Page 25 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

example, by using a binomial distribution it can be easily calculated that from a material consisting of 30 % slow or otherwise exceptional fibers, just 9 fibers are needed to end up with a sample which contains at least one slow/exceptional fiber with over 95 % probability, while for a material with only a 10 % portion of exceptional fibers, already 29 fibers are needed to achieve a similar probability. The variance in fiber dissolution profile is usually evident already from the individual dissolution traces but can also be illustrated by plotting confidence intervals for the average dissolution plots (Figure 3). Additionally, by comparing the standard errors of the mean (SEM) and confidence intervals (CI) in individual fiber dissolution times (Table 4), a more quantitative measure can be achieved. These tools can also be used to evaluate the differences between various samples, e.g. based on the mean and CI for the average dissolution time of individual fibers, BS-K and BH-PhK can be clearly separated while the enzyme treatments yield smaller differences. Another way of describing the fiber dissolution speed is by exponential fitting, assuming 1st order reaction kinetics. The dissolution process seems to adhere well to 1st order kinetics (Figure 4), at least until the majority of the fibers have dissolved. Deviation at the tailend of the log plot may indicate a mechanistic change, however, any deviations from perfect exponential decay are greatly exaggerated due to the logarithmic y-axis, with 0.1% fiber area fluctuations easily visible in the lower part of the plot. This, combined with the fact that only a few of the slowest dissolving fibers contribute to the tail-end of the average dissolution trace, practically ensures some level of deviation. One should thereby compare the exponential fitting for both the linear and log plots (Figure 4 and supplementary information) to get a balanced representation of where the errors are. In addition, while it seems that individual and even fiber clusters may not be fitted very well with an exponential decay, the average dissolution traces of the whole samples seem to adhere to 1st order kinetics to a reasonably high degree. Using this kind of model provides a very concise and rational description of the dissolution speed (Table 5), but of course it is a rigid model which can not deal with more

25

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 26 of 35

complex dissolution profiles easily, such as some potential features rising from the different dissolution mechanisms. For this, observing the linear and log fits will be invaluable. The quality of the fits can be estimated from the standard error of regression (Table 5). From this fitting model, the most ideal parameters to assess the kinetics of dissolution of different samples are rate constants k or t0 for fiber dissolution, with k representing the rate constant for dissolution, or alternatively its reciprocal, t0 , as the time taken for 63 % dissolution (e−1 fraction remaining). Table 4: A) The dissolution times determined from the average dissolution trace of the sample and B) average dissolution times for individual fibers. In A), each time corresponds to the time it takes to dissolve the indicated fraction of fiber area in the microscopy image, on average, while B) lists the average time for complete dissolution of individual fibers. The average times for individual fibers in B) were calculated using 95% dissolution as a threshold for complete dissolution. The standard error of mean (SEM) and 95 % confidence interval (CI) for the mean are also listed. Sample Treatment 25 % BS-K 23.6 BS-K enzyme 53.8 BH-PhK 14.0 BH-PhK enzyme 13.7

A) Dissolution 50 % 75 % 75.4 174.9 90.5 140.1 35.6 66.2 24.0 43.5

time (s) B) 95 % 99.5 % mean 411.2 590.4 270.1 269.6 466.0 202.2 157.4 328.4 102.2 85.6 128.8 63.6

Avg. time (s) SEM CI (95%) 56.0 147-393 38.9 110-294 17.9 65-139 13.0 32-95

Table 5: The 1st order reaction kinetics (exponential decay) fitting to the average dissolution trace. The formula y = ae−t/t0 = ae−kt was used for the fitting, where y is the fiber area at time t, a is the initial fiber area and k is the rate constant. The t0 is an alternate and perhaps more intuitive version of the rate constant, indicting the time in which the 1 − e−1 fraction or approx. 63% of material is dissolved. The SEreg is the standard error of the regression or estimate, which indicates the standard deviation of the difference between the exponential decay model and the average dissolution trace. Sample Treatment a t0 (s) BS-K 0.90 136.7 BS-K enzyme 1.33 86.1 BH-PhK 1.01 48.9 BH-PhK enzyme 1.14 29.0

26

k (s−1 ) 7.3 * 10−3 11.6 * 10−3 20.4 * 10−3 34.4 * 10−3

ACS Paragon Plus Environment

SEreg 0.95 * 10−2 4.30 * 10−2 0.53 * 10−2 0.99 * 10−2

Page 27 of 35

BS-K, enzyme treatment

Refe ren ce

En

z

BH-PhK, enzyme treatment

Refe

renc e

z

En

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

Figure 6: The effect of enzymatic treatment in bleached softwood kraft pulp (BS-K, top) and bleached hardwood pre-hydrolysis kraft pulp (BH-PhK, bottom) material. Only the average dissolution traces are plotted. In both cases the enzyme treatment (’Enz’) clearly yields a faster dissolution profile than the untreated pulp (’Reference’). In the bleached softwood kraft pulp (BS-K, top), the initial swelling of the fibers can be seen as a hump in the treated dissolution profile.

27

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Dissolution Kinetics for the Enzyme-Treated Pulps Both pulps were treated with a cellulase (Cel5, endoglucanase of family GH5) to give enzymetreated bleached softwood kraft (enz-BS-K) and enzyme-treated bleached hardwood prehydrolysis kraft (enz-BH-PhK) pulps. A high dose of cellulase (7 mg protein/g pulp) was chosen to allow us to observe the effects on dissolution of different fiber regions and ultimately on the overall kinetics (Figure 6). One interesting feature was observed with the enzyme-treated BS-K sample. The relative area for the initial phases of the dissolution seems to even increase slightly before rapidly decreasing. According to the dissolution videos there was an initial swelling phase where the fiber seemed to expand in diameter, with little or no ballooning present and less dissolution occurring from the fiber ends but rather initial swelling over the whole fiber. Most of the fiber proceeded to fragment and dispersed rapidly. Those fibers that showed the slowest rate of dissolution (assumed to be latewood) also showed faster dissolution from the ends but again with minimal ballooning. In the context of our previous discussion on the mechanisms it seems that the primary cell wall was absent or weakened, providing no inhibition to swelling of the fiber before final fragmentation of the parts and dispersion into the medium. This initial swelling phase was not observed in the kinetics or from the videos for the enzymetreated BH-PhK sample. In general the dissolution occurred a little faster from the fiber ends. The enzymatic treatment did not change the ballooning behavior of the dissolving BHPhK fibers because ballooning was not distinct, even for the untreated samples of this pulp. Thus, this swelling may be a specific feature to the softwood kraft pulp. These results might indicate that the initial observed swelling action for enzyme treated BS-K is somehow related to the higher molecular weight or hemicellulose content in the original BS-K sample or that the enzyme treatment is somehow less effective for the BH-PhK sample. The faster overall dissolution of the enzyme-treated fibers might be due to their measured lower molecular weight (Table 3, SEC plot in supplementary material), or the fragmentation of the fibers 28

ACS Paragon Plus Environment

Page 28 of 35

Page 29 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

which led to shorter fibers and higher amounts of fines. However, decreases in fiber width were on the other hand not observed (Table 3). The fiber length correlated with the fiber dissolution times but the correlation was not very strong meaning that factors, other than fiber length, clearly had an impact on the dissolution process. As PhK pulping affords a much more weakened primary cell wall, the observation that the mechanism of dissolution did not change is logical as the primary cell wall seems to be implicated in ballooning and helix formation. Thus, the overall reduction in molecular weight of BH-PhK after the enzyme treatment is the likely explanation for increased dissolution kinetics. However, for both samples the overall kinetics of dissolution is clearly faster after enzymatic treatment. In the case of BH-PhK, the dissolution speed is nearly doubled (Table 5). The results also show that the initial 25 % and 50 % dissolution times of the enzymatically treated pulps are similar or even longer than for the respective untreated base pulps. Thereafter, the dissolution speed changes and the residual enzyme treated samples dissolve faster, compared to the untreated base pulps (at 75 % dissolution times and higher). This illustrates the value of the kinetic analysis. The apparent initial swelling of the enzyme-treated BS-K sample obviously highlights the sensitivity of the kinetics method towards swelling of the sample, which will be an additional valuable feature for assessing the efficacy of a pre-treatment. The swelling is also evident from the notably worse exponential fit (Table 5). The kinetic parameters for the cellulase-treated samples are summarised in Tables 4 and 5.

Future Work It has not escaped the authors minds that there is great potential for automation of the extraction of the dissolution data directly from the fiber dissolution videos, in addition to the data analysis, plotting and parameter calculations presented above. The automation should also be quite reasonably accomplished with modern tools, as the video data is fairly clean and easy to interpret, at least if care is taken when acquiring the videos. If designed correctly, automation might also have the additional benefit of removing human error in 29

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

processing the videos. However, it may also allow for further introduction of error if not tested correctly. Some preliminary work has been already conducted on the matter with a simple (experimental) extraction tool available in solna, based on adaptive Sauvola Thresholding. 20,29 Currently, identification of the starting frame is important for the accuracy of the parameters. Because the exponential fittings for the averages and cluster averages are quite good, rate constants (t0 and k), derived from exponential fitting would have the benefit that they would not be so dependent on accurately placing the initial frame, provided that the delay before dissolution is roughly the same in each sample. Thus, even with a very straightforward analysis of the dissolution imaging the rate constants could be used to compare kinetics of the dissolution of pulps. However, it should be noted, that currently the bottleneck of the method is not the processing but rather the acquisition of the microscopy images themselves: the dissolution of some fibers can already take close to 10 minutes and preparation of the fibers for imaging also takes a significant amount of time. This can be compared to the fairly quick review and processing of the resulting video, taking a minute or two. The time taken for generation of usable data may be reduced by incorporating many fibers in one video, as long as they do not overlap; the ImageJ data extraction method allows for separation of the kinetics for selected fibres, in one video, by applying a white background to remove the fibers or other objects that you do not wish to include in the quantification. Reprocessing the frames can then be used to analyze separate fibers in sequence. In any case, streamlining and automating the method as a whole remains as an important future work.

Conclusions A new method of assessment of the kinetics of dissolution of single fibers has been demonstrated. Limited dissolving conditions (0.2 M CED) are used to allow for sufficient resolution

30

ACS Paragon Plus Environment

Page 30 of 35

Page 31 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

of dissolution mechanistic events, e.g. ballooning and disk formation. Clustering of the data has demonstrated the ability to differentiate between fiber types (early and late wood in the case of the softwood kraft pulp). By analysing a sufficient number of fibers from a sample, the methods sensitivity to fiber variability can be controlled. Importantly the method has also yielded the average dissolution traces, which can be used to differentiate between pulps and treatments visually. The 25, 50, 75, 95 and 99.5 % dissolution time parameters are also proposed to give a good practical overall assessment of the recalcitrance of large sets of different pulps. 1st order kinetics can be used to provide a more quantitative but rather rigid model, and associated kinetic parameters, of the dissolution process. Differences in recalcitrance of the pulps were easily observed using these methods. The bleached hardwood prehydrolysis kraft pulp dissolved more rapidly than the bleached softwood kraft pulp. Ballooning was also reduced for the former pulp, presumably due to a more complete removal of the primary cell wall. The cellulase-treated pulps dissolved even faster still, with the softwood pulp exhibiting an interesting pre-swelling phenomenon before rapid dissolution. Ballooning was almost completely absent in the enzyme-treated samples. Overall, the method shows promise for parametrization of a wide range of pulps towards non-derivatising dissolution, which also seems to be closely linked to chemical reactivity.

31

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Supporting Information 1. Additional details on solna processing 2. Additional microscopy images of the discussed dissolution mechanisms 3. SEC plots of untreated and treated pulps 4. Additional dissolution traces with confidence intervals 5. Additional dissolution traces with exponential decay fits

Acknowledgement The authors thank CLIC Innovation Oy for funding provided under the Advanced Cellulose to Novel Products (ACel) program and the University of Helsinki Science Faculty for funding to support the development of novel research areas. Ritva Heinonen (VTT) is thanked for skillful technical work with the recording of fiber dissolution videos.

References (1) Nageli, C. Sitzber. Bay. Akad. Wiss. M¨ unchen 1864, 1, 282–323. (2) Hock, C. W. In Cellulose and cellulose derivatives (Part 1): IV Structure and Properties of Cellulose Fibres, C Microscopic Structure; Emil Ott, H. M. S., Grafflin, M. W., Eds.; Interscience, New York, 1954. (3) Siu, R. G. In Cellulose and cellulose derivatives (Part 1): III Chemical Nature, 5. Microbiological Degradation; Emil Ott, H. M. S., Grafflin, M. W., Eds.; Interscience Publisher, New York., 1954. (4) Cuissinat, C.; Navard, P. Macromol. Symp. 2006, 244, 1–18.

32

ACS Paragon Plus Environment

Page 32 of 35

Page 33 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

(5) Cuissinat, C.; Navard, P. Macromol. Symp. 2006, 244, 19–30. (6) Le Moigne, N.; Montes, E.; Pannetier, C.; H¨ofte, H.; Navard, P. Macromol. Symp. 2008, 262, 65–71. (7) Chaudemanche, C.; Navard, P. Cellulose 2011, 18, 1–15. (8) Cuissinat, C.; Navard, P.; Heinze, T. Cellulose 2008, 15, 75–80. (9) Moigne, N. L.; Jardeby, K.; Navard, P. Carbohydr. Polym. 2010, 79, 325 – 332. (10) Gehmayr, V.; Potthast, A.; Sixta, H. Cellulose 2012, 19, 1125–1134. (11) Schild, G.; Sixta, H. Cellulose 2011, 18, 1113–1128. (12) Spinu, M.; Dos Santos, N.; Le Moigne, N.; Navard, P. Cellulose 2011, 18, 247–256. (13) Parviainen, H.; Parviainen, A.; Virtanen, T.; Kilpel¨ainen, I.; Ahvenainen, P.; Serimaa, R.; Gr¨onqvist, S.; Maloney, T.; Maunu, S. L. Carbohydr. Polym. 2014, 113, 67 – 76. (14) Arnoul-Jarriault, B.; Passas, R.; Lachenal, D.; Chirat, C. Holzforschung 2016, 70(7), 611–617. (15) Olsson, C.; Idstr¨om, A.; Nordstierna, L.; Westman, G. Carbohydr. Polym. 2014, 99, 438 – 446. (16) Wahlstr¨om, R.; Rovio, S.; Suurn¨akki, A. RSC Adv. 2012, 2, 4472–4480. (17) Schindelin, J. et al. Nat. Methods 2012, 9, 676–682. (18) Sahoo, P.; Soltani, S.; Wong, A. Comput. Vision Graphics Image Process. 1988, 41, 233 – 260. (19) Yen, J.-C.; Chang, F.-J.; Chang, S. IEEE Trans. Image Process. 1995, 4, 370–378.

33

ACS Paragon Plus Environment

Biomacromolecules 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(20) https://bitbucket.org/vltrrr/solna/. (21) https://youtu.be/O4pcDtuUHE4. (22) van der Walt, S.; Colbert, S.; Varoquaux, G. Comput. Sci. Eng. 2011, 13, 22–30. (23) McKinney, W. Data Structures for Statistical Computing in Python. Proceedings of the 9th Python in Science Conference. 2010; pp 51 – 56. (24) Hunter, J. D. Comput. Sci. Eng. 2007, 9, 90–95. (25) Millman, K. J.; Aivazis, M. Comput. Sci. Eng. 2011, 13, 9–12. (26) Lange, T.; Berger-Nicoletti, E.; Kosma, P.; Potthast, A.; Sixta, H. Lenzinger Ber. 2003, 82, 102–106. (27) Parviainen, A.; King, A. W. T.; Mutikainen, I.; Hummel, M.; Selg, C.; Hauru, L. K. J.; Sixta, H.; Kilpel¨ainen, I. ChemSusChem 2013, 6, 2161–2169. (28) Le Moigne, N.; Navard, P. Cellulose 2010, 17, 31–45. (29) Sauvola, J.; Pietik¨ainen, M. Pattern Recognition 2000, 33, 225 – 236.

34

ACS Paragon Plus Environment

Page 34 of 35

Page 35 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biomacromolecules

Graphical TOC Entry

Cellulose pulp fibers

Fiber dissolution Optical microscopy

35

Semi-automated data-analysis

ACS Paragon Plus Environment