Current Issue

Journal of Korea Technical Association of the Pulp and Paper Industry - Vol. 54 , No. 2

[ Article ]
Journal of Korea Technical Association of the Pulp and Paper Industry - Vol. 54, No. 2, pp. 9-17
Abbreviation: J. Korea TAPPI
ISSN: 0253-3200 (Print)
Print publication date 30 Apr 2022
Received 07 Feb 2022 Revised 11 Mar 2022 Accepted 18 Mar 2022

Utilization of TEM with Automated Tile Scan Technique for Length Determination of CNF
Byeongcheol Min1, 2 ; Hee-Ae Kwon3 ; Ohkyung Kwon4, ; Soo-Jeong Shin5,
1Department of Wood & Paper Science, Chungbuk National University, Postdoctoral Fellow, Republic of Korea
2R&D Institue Pulp & Paper Team, Moorim P&P, Deputy General Manager, Ulsan, Republic of Korea
3Department of Wood & Paper Science, Chungbuk National University, Student, Republic of Korea
4National Instrumentation Center for Environmental Management (NICEM), Seoul National University, Research Scientist, Republic of Korea
5Department of Wood & Paper Science, Chungbuk National University, Professor, Republic of Korea

Correspondence to : †E-mail: (Address: Department of Wood & Paper Science, Chungbuk National University, Cheongju, 28644, Republic of Korea)
Co-corresponding Author : ‡E-mail: (Address: National Instrumentation Center for Environmental Management, Seoul National University, Seoul, 08826, Republic of Korea)

Funding Information ▼


Large field of view (FOV) imaging was introduced and evaluated to characterize length and width of cellulose nanofibril (CNF) by transmission electron microscopy (TEM) with automation capability. New imaging method provided an image area as wide as covering one hole on grid. The large FOV image provided 10×10 of tile image which can cover 2500 μm2 of sample area. Large FOV imaging has a great advantage to obtain massive information for average length of CNF. The imaging method enable to overcome the limitation of conventional TEM imaging to elicit average length of CNF. The average length and width of CNF was 375 nm and 7 nm, respectively. Length-weighted average length of CNF was 651 nm. Study to select the kind of support film resulted as carbon film was the best among silicon monoxide and carbon. Carbon film was more durable to high electron beam energy exposure and better to disperse CNF separately than silicon monoxide.

Keywords: Cellulose nanofibril, TEM, large field of view, width, nanocellulose

1. Introduction

Nanocellulose has been recently focused on as one of the most attractive material outstanding physical and chemical properties and the infinite potential of applications.[1,2] Nanocellulose is sustainable and environmentally friendly material isolated from cellulose which is known as the most abundant natural polymer on earth. Classification of nanocellulose such as cellulose nanofibril (CNF), cellulose nanocrystal (CNC), or bacterial cellulose (BC) depends on the production methods.[3] Nanocellulose is available to be applied to diverse fields of material science such as coating, composites, and films due to the unique properties including low thermal expansion coefficients, high elastic moduli, high surface area, and high aspect ratios.[2,3]

A nanocellulose is defined as a celluolosic fiber which has at least one dimension of 100 nm or less as regarding nanotechnology.[1] The cellulose of nanometer scales has fundamentally different properties from the cellulose in natural bulk state. Size reduction of cellulose down to nano-scale changes the phase of nanocellulose to obey thermodynamic and physicochemical laws due to the significant effect of high surface area. It is a contrast to cellulose of the large-scale phase following the traditional thermodynamic laws.[4] Application of nanocellulose for variety purposes is depends on the size of nanocellulose. Definition of average and range of size value is also required for proper applications. Therefore, a method of precise size measurement of size is fundamental and must be standardized for production and application of nanocellulose.

The nanostructure of CNC is relatively easy to determine average width and length, whereas precise characterization of CNF is limited to the width. Size distribution of CNF has more wide range especially in length. The average length of CNF has been approximately estimated since CNF has significantly high aspect ratios and wide range of length distribution.[3,4]

An average length of CNFs can be determined by a light scattering technique and viscoelastic evaluation.[5,6] A light scattering technique can provide estimation of particle size distribution for morphological characteristics of nanocellulose. Viscoelastic evaluation is another technique to determine an average length of nanocellulose by measurement of shear viscosity. However, both methods remain as an approximation not determining the precise diameter and length dimensions of nanocellulose due to the high aspect ratios.[6]

Microscope methods have been applied to determine CMF dimensions using a transmission electron microscope (TEM), a scanning electron microscope (SEM) or an atomic force microscope (AFM).[7-9] These electroscope methods have a good agreement on characterizing the size of CNF as direct measurement methods which can provide more precise value of dimension as compared to light scattering and viscoelastic methods.

However, SEM, TEM and AFM are also finite to measure the average length of CNF. The microscope methods have significant limitation due to the small field of view (FOV).[2] The FOV limitation enforces to visit a large number of imaging sites with CNF as many as possible if we need to collect a sufficient number of images for a statistically representative number of CNFs. This task is not easy to accomplish by manual operation of the microscopies.

A modern TEM equipped with a motorized goniometer and control software can automatically collect high magnification images according to pre-defined sites by a user. The automatically collected images cover a large field of view including sufficient number of CNFs for the determination of length distribution. With a grid method of non-overlapping CNFs, the large FOV capability enables us to measure the average length of CNFs and their distribution. In the present study, we evaluated an automated TEM with large FOV capability to determine average and distribution of CNF length.

2. Materials and Methods
2.1 CNF preparation

Mixed bleached kraft hardwood pulp was selected as the material of CNF. The pulp was modified by carboxymethylation treatment to improve productivity of CNF and to reduce demand of energy. 75 g of pulp was suspended in 2 L of isopropyl alcohol. 2.5 M of sodium hydroxide solution in 400 ml of methanol and 0.88 M of chloroacetic acid solution in 600 ml of isopropyl alcohol were supplied in addition to the pulp suspension. The container of the pulp suspension was sealed and placed into an oven at 95°C for 150 minutes. Carboxymethylated pulp was filtrated and washed with water.

Carboxymethylated pulp was diluted as 2% (w/w) in water and subjected to high pressure homogenizer (Panda plus, GEA, Italy). The first homogenizing passed the pulp in the pressure of 100 bar and then additional five passes were conducted in the condition of 700–800 bar.

2.2 TEM imaging for CNF characterization

In previous study,[10] three different types of support films were selected to evaluate the fitness for TEM imaging of CNF. The selected support films were made with different materials such as silicon monoxide, carbon and formvar. Grids were modified with glow discharge treatment at 15 mA for 25 seconds to evaluate the effects of glow discharge process on different support films. CNF was diluted as to 0.001% and dispersed on the grids. CNF was negatively stained using uranyl acetate. We qualitatively evaluated each grid in the aspect of CNF dispersion and attachment properties by utilizing EF-TEM (120 kV, Libra 120, Carl Zeiss, Germany) with CCD camera (SC200W, Gatan, US). The support film of silicon monoxide resulted as the best support film for CNF imaging due to the dispersion and fixation of CNF fibers on the grid. The carbon support film showed moderate and acceptable results for dispersing and fixing CNF fibers.[10] Silicon monoxide and carbon support film were selected for CNF characterization by automated TEM imaging. Both support films were subjected to glow discharge treatment for 15 seconds.

A brand new TEM (120 kV, Talos L120C D5375 CyroTwin, FEI, Czech) with CMOS camera (Ceta 16K, FEI, Czech) was used for CNF characterization. The TEM equipped automation capability, a motorized goniometer, and MAPS software for automated image collection and stitching capability. Quantitative determination of width and length distribution of CNF was performed with TEM images by a large field of view imaging. The automated TEM could provide 10×10 tiles which were almost covering the size of one mesh (50×50 μm) on a grid. The size of one tile was 5×5 μm at 8,500 magnification and the tile image had 4096 pixels for each dimension. The actual pixel size was determined as 1.22 nm for each dimension by calibration results. The pixel size and number of pixels were multiplied to determine the length and width of each CNF fibers in TEM images. An Image software, ImageJ, was used for counting the number of pixels of the CNF fibers in the images.[11]

3. Results and Discussion
3.1 Comparison between silicon monoxide and carbon support film for large FOV images

Silicon monoxide supporting film was expected as the most suitable among silicon oxide, carbon and formvar films for TEM imaging since the results of previous study with EF-TEM showed silicon monoxide was the best for dispersion and fixation of CNF.[10] Carbon support film, however, actually showed better results for the automated TEM imaging than the silicon monoxide film. The main reason was durability of the films. Silicon monoxide film got cracks or damages more readily than carbon film which made difficult to take a good large FOV image (Fig. 1). We frequently observed silicon monoxide support film cracked after automated TEM imaging. Large field imaging system of automated TEM required the grids exposed on high beam energy. Less thermo-stability of silicon monoxide film may not be suitable to endure the high beam energy.[10] Otherwise, carbon film was stable to exposure of high beam energy for large FOV imaging (Fig. 2).

Fig. 1. 
Large FOV image of CNF on the silicon monoxide support film. The film damaged by electron beam of TEM with automation capability.

Fig. 2. 
Automated TEM image of CNF on carbon support film. Red color represents countable CNF and black color CNF was not counted due to crossing (a), twisting (b), kinking (c) and combined effect (d).

The tile images on the silicon monoxide film revealed a number of aggregated CNF fibers (Fig. 1). The aggregated fibers were produced by the interactive formation of fibers including crossing, twisting, and overlapping. Silicon monoxide film resulted the highest fiber retention capacity due to the high hydrophilicity in the previous study.[10] The high retention capability of silicon monoxide increased the chance of interaction and aggregation of CNF fibers. The aggregated CNF was not proper to be counted as one fiber to measure the length of fibers.

Large FOV image by automated TEM provided well distributed and isolated CNF fibers which makes convenient to find more countable fibers in Fig. 2. The fibers with red color presents counted mono fibers. Fibers marked with dark color are uncounted for length measurement due to the aggregated or distorted structures. TEM image on the carbon support film also had some aggregated fibers by crossing, twisting and kinking of fibers. The number of aggregated CNF is much less than silicon monoxide film. This well distribution of CNF is due to the less hydrophilicity of carbon support film than silicon monoxide support film.[10]

In the present study, sizes of particle-like structures in the background images of carbon and silicon monoxide support film were different (Figs. 3 and 4). The different quality of the background is hypothetically due to the grain size of composites of the films. Different grain size of two support film is related to the chemical structures of silicon monoxide and carbon and/or formation of amorphous region of the materials in films. The different structures and the amorphous formation of film might give us chance to explain the difference in durability and hydrophilicity related to the different degree of dispersion.

Fig. 3. 
Background image of silicon monoxide support film.

Fig. 4. 
Background image of carbon support film.

The bigger grain size in the background image from the silicon monoxide film could be another reason to make silicon monoxide film less suitable for automated TEM imaging. TEM imaging technique basically requires a good contrast to distinguish between samples and background. CNF may not be characterized precisely due to the less definitive edges formed by the bigger grain. Characterization of CNF width is more vulnerable to be alteration by the grain size since the width size of CNF is much smaller than length with the range of 3–15 nm.[4]

3.2 Determination of length and width of CNF

Results of CNF size measurement shows in Table 1 including an average length and width. Range of sample size and the number of measured samples are also described. The average length of CNF resulted in 375 nm and standard deviation was 323 nm. The length range of CNF was between 53 nm and 1459 nm out of the 165 measured samples. The average width of CNF was 7 and standard deviation was 2. The minimum value of width was 4 nm and the maximum value was 13 nm among the 180 of measured samples. The aspect ratio was 53. The distribution of CNF length had right-skewed shape in Fig. 5 while CNF width showed bell-shaped distribution in Fig. 6. Length-weighted average length (Lw) was calculated (Eq. 1) as 651 nm which increased almost as two times as number average length (Ln). Width-weighted average width (Ww) was 8 nm which was not far different from number average width (Wn).

  • where Ni = number of Li and Li = length of a CNF
Table 1. 
Characteristics of CNF measured by TEM with automation capability showing average and standard deviation of length and width
Average (nm) Standard deviation (nm) Sample range (nm) Number of samples measured
Length 375 323 53–1459 165
Width 7 2 4–13 180

Fig. 5. 
Length distribution of CNF as number average length.

Fig. 6. 
Width distribution of CNF as number average length.

It is required to consider a length-weighted average length (Lw) instead of number average length (Ln) to represent average length of CNF connected to the specific properties and application of CNF. The first reason to stand the hypothesis is that fibers have different values to impact on the average CNF property. In Tanaka’s study,[2] average length of CNF was determined by shear viscosity measurements. The study found a linear relationship between viscos average length (Lvis) and Lw rather than Ln. The result is understandable considering the higher effect of long fibers on the viscosity variation than short fibers.

The second reason to support preference of using Lw is that the possibility of long fibers is lower to be counted in CNF characterization with TEM imaging. Longer fiber has more chance to be lost or aggregated with other fibers. The image of long fibers has more chance to be cut by boundary of TEM image near the edges. Long fibers also have higher chance for self-aggregation or distortion. These cases make low chance of reflecting long fibers onto the calculation of average length. Considering these two reasons, application of Lw is more suitable to determine the property of CNF of the wide range of length distribution.

Sufficient number of measurement is required to provide a value as a representative size, which is a characteristic size for a certain volume of sample. Many scholars do not accept a characteristic width and length of CNF if there is no sufficient number of measurement. For the width and length measurement, especially for the width, a high magnification pictures of the CNF are necessary, but this decreases the area of FOV significantly. Consequently, the number of CNFs in a FOV at the high magnification are very limited. To overcome this kind of limitation, we need to take pictures from many different sites. Thus, the imaging of a large FOV is a systematic way to obtain many high magnification pictures containing sufficient number of measurable CNFs. Large FOV image with automated TEM enabled to scan 2500 um2 of sample area which is much larger than AFM scanning area with high magnification.[8] The large FOV images encourages to determine a characteristic value close to “trueness” and more strictly following the way (which means to measure many samples) to produce statistically representative value for the sample.

Another fact we need to consider for determination of a representative size for a volume of CNF is that degree of dispersion or dispersion characteristics of CNFs on a TEM grid. If the dispersion of the CNFs is good, we can select few or several sites to obtain CNF pictures and use them to determine the width and length of the CNFs. However, if the dispersion is not good, parts of the TEM grid show well dispersed CNFs which other parts does not. In this situation, we need to increase the number of imaging sites to compensate a large variation of number of measurable CNFs. Also number, size, or volume of aggregated CNFs can be measured or counted. This state of aggregation of CNFs might be another “characteristic” of the CNFs. This kind of investigation is only possible when we observe and take pictures most area of the TEM grid. In brief, we need many large FOVs for the determination of the length distribution by covering for both well dispersed and aggregated CNFs on a TEM grid.

4. Conclusions

The most suitable support film resulted as carbon film for large FOV imaging. Carbon support film showed more durable to high electron beam energy exposure for large area imaging, while silicon oxide support film was broken or cracked by electron beam more readily. Carbon support film dispersed CNF better individually than silicon monoxide film. Silicon monoxide showed many aggregated forms of CNF due to the higher fiber retention ability related to hydrophilicity. Carbon film has less background noise than silicon monoxide. The different background noise between carbon and silicon monoxide film was possibly due to the composite structure of the films and luminescent effect of oxide parts.

Characterization of CNF using automated TEM resulted the average length and width were 375 (±323) nm and 7 (±2) nm. The distribution of length and width were in range of 53–1459 nm and 4–13 nm. Lw of CNF was 651 nm which is about two times of Ln. Application of Lw is necessary to determine the correlations of between average length and properties of CNF. This concept is based on the wide range of length distribution, the probability of non-counting and the higher effects on CNF properties of long fibers.

The automated TEM produced a large FOV image with high magnification covering 2500 μm2 of supporting film. The advanced imaging method provided the largest area of sample image with high magnification as compared to other microscope methods. Large FOV imaging by automated TEM is useful to provide massive information of CNF characteristics as many as statistically acceptable samples to determine average length of CNF.


This research was supported by Chungbuk National University Korea National University Development Project (2020).

Literature Cited
1. Jonoobi, M., Oladi, R., Davoudpour, Y., Oksman. K., Dufresne, A., Hamzeh, Y., and Davoodi, R., Different preparation methods and properties of nanostructured cellulose from various natural resources and residues: a review, Cellulose 22:935-969 (2015).
2. Tanaka, R., Saito, T., Ishii, D., and Isogai, A., Determination of nanocellulose fibril length by shear viscosity measurement, Cellulose 21:1581-1589 (2014).
3. Lavoine, N., Desloges, I., Dufresne, A., and Bras, J., Microfibrillated cellulose – its barrier properties and applications in cellulosic materials: a review, Carbohydrate Polymers 90:735-764 (2012).
4. Ioelovich, M., Cellulose as a nanostructured polymer: a short review, BioResources 3:1403-1418 (2008).
5. Klemm, D., Kramer, F., Moritz, S., Lindström, T., Ankerfors, M., Gray, D., and Dorris, A., Nanocelluloses: A new family of nature-based materials, Angewandte Chemie International Edition 50:5438-5466 (2011).
6. Morais, J. P. S., Rosa, M. F., de Souza Filho, M. D. S. M., Nascimento, L. D., do Nascimento, D. M., and Cassales, A. R., Extraction and characterization of nanocellulose structures from raw cotton linter, Carbohydrate Polymers 91:229-235 (2013).
7. Li, Q. and Renneckar, S., Molecularly thin nanoparticles from cellulose: isolation of sub-microfibrillar structures, Cellulose 16:1025-1032 (2009).
8. Usov, I., Nystroml, G., Adamcik, J., Handschin, S., Schutz, C., Fall, A., Bergstrom, L., and Mezzenga, R., Understanding nanocellulose chirality and structure-properties relationship at the single fibril level, Nature Communications 6:7564 (2015).
9. Song, W.-Y., Shin, S.-J., and Juhn, S., Comparison of characteristics of cellulose nanofibrils from corn stalk and bamboo fiber, Journal of Korea TAPPI 49:76-83 (2017).
10. Kwon, H.-A., Shin, S.-J., and Kwon, O., Optimal conditions of TEM grid for quantitative morphological investigation of nanocelluloses by transmission electron microscopy, Journal of Korea TAPPI 49:25-31 (2017).
11. Papadopulos, G., Spinelli, M., Valente, S., Foroni, L., Orrico, C., Alviano, F., and Pasquinelli, G., Common tasks in microscopic and ultrastructural image analysis using ImageJ, Ultrasturural Pathology 31(6):401-407 (2009).