Cellulose nanofibrils prepared by gentle drying methods reveal the limits of helium ion microscopy imaging
Full text
This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Cellulose nanofibrils prepared by gentle drying methods reveal the limits of helium ion microscopy imaging © The Royal Society of Chemistry 2019. Published version Ketola, Annika E.; Leppänen, Miika; Turpeinen, Tuomas; Papponen, Petri; Strand, Anders; Sundberg, Anna; Arstila, Kai; Retulainen, Elias Ketola, A. E., Leppänen, M., Turpeinen, T., Papponen, P., Strand, A., Sundberg, A., Arstila, K., & Retulainen, E. (2019). Cellulose nanofibrils prepared by gentle drying methods reveal the limits of helium ion microscopy imaging. RSC Advances, 9(27), 15668-15677. https://doi.org/10.1039/C9RA01447K 2019
Cellulose nanofibrils prepared by gentle drying methods reveal the limits of helium ion microscopy imaging Annika E. Ketola, a Miika Lepp¨ anen, b Tuomas Turpeinen, a Petri Papponen, b Anders Strand, c Anna Sundberg, c Kai Arstila b and Elias Retulainen * a TEMPO-oxidized cellulose nanofibrils (TCNFs) have unique properties, which can be utilised in many application fields from printed electronics to packaging. Visual characterisation of TCNFs has been commonly performed using Scanning Electron Microscopy (SEM). However, a novel imaging technique, Helium Ion Microscopy (HIM), offers benefits over SEM, including higher resolution and the possibility of imaging non-conductive samples uncoated. HIM has not been widely utilized so far, and in this study the capability of HIM for imaging of TCNFs was evaluated. Freeze drying and critical point drying (CPD) techniques were applied to preserve the open fibril structure of the gel-like TCNFs. Both drying methods worked well, but CPD performed better resulting in the specific surface area of 386 m 2 g 1 when compared to 172 m 2 g 1 and 42 m 2 g 1 of freeze dried samples frozen in propane and nitrogen, respectively. HIM imaging of TCNFs was successful but high magnification imaging was challenging because the ion beam tended to degrade the TCNFs. The effect of the imaging parameters on the degradation was studied and an ion dose as low as 0.9 ion per nm 2 was required to prevent the damage. This study points out the differences between the gentle drying methods of TCNFs and demonstrates beam damage during imaging like none previously reported with HIM. The results can be utilized in future studies of cellulose or other biological materials as there is a growing interest for both the HIM technique and bio-based materials. 1 Introduction Cellulose nanobrils (CNFs) have been under intensive investigation due to their unique properties, such as high tensile strength, large specic surface area, rheology and tendency for lm formation. 1 CNFs can be prepared from natural cellulose wood bres by mechanical and chemical treatments, 2 which makes them a biodegradable and renewable material. Various different application elds exist for CNFs ranging from paper, composites and food additives to bio-lm material for printed electronics or packaging, biomedical applications and aerogels. 3–8 Thus, proper characterisation of CNFs is important for quality control and safety assessment. 9 The nanoscale size, the branched bril structure and the high water content make the characterisation of CNFs challenging and demand sophisticated techniques. Scanning electron microscopy (SEM) is oen used for characterisation of the morphology of CNFs and bril dimensions in the mm to nm scale. 9–13 Transmission electron microscopy (TEM) is the most powerful imaging technique and can be utilized for detailed nanoscale evaluation of single bril dimensions of CNFs. 9,14–17 SEM usually requires a conductive coating before imaging and the TEM method is limited to thin samples like single brils, and because of that, the new imaging method scanning helium ion microscopy (HIM) has aroused interest during recent years. In the HIM method, a focused helium ion beam releases secondary electrons (SE) from the sample surface and creates an image in a similar way to SEM. 18–20 Because of the HIM single atom source, the probe is smaller compared to SEM, which together with the smaller excited surface volume makes higher imaging resolution possible. 18,21 Another advantage is that nonconductive samples do not need coating because the charging can be compensated for with an electron ood gun. HIM is widely utilized as a nanofabrication tool because helium ion beams can modify surfaces in the nanoscale by ion sputtering or implantation; 22–24 however, it is reported that in imaging applications the beam damage for organic samples is negligible. 20,25,26 Thus far the HIM-imaging method has been used for cellulose-based materials usually with low to intermediate resolutions. Li et al. (2016) 27 used HIM for successful imaging of a VTT Technical Research Centre of Finland Ltd, P. O. Box 1603, FI-40101 Jyv¨ askyl¨ a, Finland. E-mail: elias.retulainen@vtt. b University of Jyv¨ askyl¨ a, Nanoscience Centre, Department of Physics and Department of Biological and Environmental Science, FI-40014 Jyv¨ askyl¨ a, Finland c ˚ Abo Akademi University, Johan Gadolin Process Chemistry Centre, Porthansgatan 3, FI-20500 ˚ Abo/Turku, Finland Cite this: RSC Adv.,2019,9, 15668 Received 26th February 2019 Accepted 13th May 2019 DOI: 10.1039/c9ra01447k rsc.li/rsc-advances 15668 |RSC Adv.,2019,9, 15668–15677 This journal is © The Royal Society of Chemistry 2019 RSC Advances PAPER Open Access Article. Published on 20 May 2019. Downloaded on 5/21/2019 10:07:53 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online View Journal | View Issue
paper-like composites of activated carbon and mechanically brillated CNFs whereas Torvinen et al. (2017) 4 studied kaolin— CNF composites of different CNF-types. Virtanen et al. (2018) 28 studied mechanically brillated CNF-aerogels, and HIM had a good depth of view for the freeze-dried CNF-aerogel when compared to SEM. The only study with high-resolution imaging of cellulose-based materials is by Postek et al. (2011) 21 where cellulose nanocrystals on mica were imaged. Because of that, the high-resolution imaging capabilities of HIM for cellulosic materials are still unclear. Vacuum based imaging methods require dry samples and in order to obtain images that represent the nanocellulose structure in the wet-state, a gentle drying procedure is necessary. Improper drying will result in the collapse of the pores and coalescence of brils. 29 This can be prevented, for example, by using cryoxing, in which the sample is rapidly frozen in a cryoliquid, followed by drying of the frozen sample under a vacuum, where the ice sublimates into a gas without collapsing the structure. Liquid nitrogen (LN 2 )is commonly used for freezing of various samples but is known to be affected by the Leidenfrost-effect where the boiling liquid forms an insulating gas layer between the sample and the coolant. This delays the freezing process, giving time for unwanted structural changes caused by ice crystal formation. With liquid propane (LPGS), used near its freezing point, this effect is reduced. 15 Another method for the preservation of the wet structure of the material is the critical point drying (CPD), where the collapse of the sample structure is prevented by passing the liquid–gas interphase in the drying by replacing the solvent with the supercritical uid which is then turned into gas. 30 Gentle drying of CNFs from water using cryoxing or CPD results an aerogel-like materials with large surface area. 11,31,32 Porous aerogels are oen used as an insulators or adsorbents, but unlike brittle silica or carbon, native CNFaerogels have been shown to possess mechanical toughness, exibility and soness. 11,33–35 CNF-aerogel studies have involved CNFs of different types, drying techniques and applications, including conductive enzymatic-CNF aerogels dried with cryo-LPGS, 33 insulating and transparent TEMPO-oxidized liquid crystalline CNF aerogels dried with CPD, 34 magnetic bacterial-CNF aerogels dried with freeze drying, 35 soenzymatic-CNF and TCNF aerogels dried with CPD 11 and hydrophobic mechanically brillated CNF aerogels dried with freeze drying. 31 The effectofdrying techniques on CNFs structure 31,36 and properties 37,38 has been demonstrated showing unambiguously how the CPD drying is able to preserve the open bril structure of CNFs better than freeze drying. TCNF aerogels have been also shown to have higher specic surface area (SSA) 39 than bacterial-CNF when dried with the same technique. 11 Ahigh SSA indicates that the original open bril structure has been well preserved during and SSA as high as 480 m 2 g 1 has been reported for TEMPO-oxidized CNFs aer CPD. 40 Also freeze drying can achieve relatively high (100–300 m 2 g 1 ) SSA values for CNFs depending on the applied procedure. 10–12,32,41 The objective of this study was to nd the optimal preparation methods for wet TEMPO-oxidized CNFs (TCNFs) in order to preserve its ne bril structure and evaluate the suitability of HIM for imaging the porous CNF-material with high resolution. TCNFs were selected for the work over other CNF-types as they have well-characterized ne and homogeneous structure, large specic surface area and high charge; 14 thus, they can be considered to be the most delicate structures to reveal the convenience of the methods. Also, TCNFs alone have not been imaged with HIM and the effect of different drying techniques on TCNF aerogel structures has not been evaluated before. TCNFs were dried with four different drying procedures including cryoxing with LN 2 or LPGS followed by freeze dryingandsolventexchangefollowedbyCPDincludingtwo different solvent exchange procedures. The rst CPD procedure involved sample xation with glutaraldehyde (GA) and osmium tetroxide (OsO 4 ) and dehydration with ethanol (EtOH) before CPD. The second procedure involved only dehydration using EtOH and acetone (AE) prior to CPD. Nitrogen (N 2 )- sorption and BET-analysis were used to determine the SSA of the TCNFs samples in order to quantify the differences between the drying methods. The results are expected to give useful information for the future studies of delicate bio-based structures with HIM. 2 Experimental 2.1. Materials 2.1.1. Chemicals. Propane (class 2, UN 1965: 95% propane and 5% butane) and liqueed nitrogen (LN 2 ) was purchased from AGA Gas Ab, Liding¨ o, Sweden. Acetone (AE, $99.9%) was obtained from Sigma-Aldrich, Darmstadt, Germany. Sodium cacodylate (NaCac, R1104, Agar Scientic, Stansted, UK) was acquired as a powder and a 0.4 M buffer stock-solution was prepared using ultrapure-water. Glutaraldehyde (GA, 25%- solution for electron microscopy) was obtained from Merck (Darmstadt, Germany) and used as a 2%-solution in 0.1 M NaCac-buffer solution (pH 7.4). Osmium tetroxide (OsO 4 , Electron Microscopy Sciences, Hateld, USA) was a 4%-solution, which was diluted to a 2%-stock solution with ultrapure-water. Ethanol (EtOH, absolute AA, Etax) was obtained from Altia Oyj, Rajam¨ aki, Finland. 2.1.2. TEMPO-oxidized CNFs. TCNFs were prepared from never-dried birch krapulp by TEMPO-mediated oxidation and uidization. TEMPO-mediated oxidation was performed according to a previously described procedure, 2,14 in which bres were rst suspended in a water solution of TEMPO and sodium bromide. Then, NaClO solution was added to the suspension (5 mmol to 15 mmol per gram of bres) and the pH was adjusted to 10 at room temperature with NaOH. The reaction was considered complete when the pH remained stable. Aer oxidation, the bres were washed thoroughly with deionized water followed by treatment with a microuidizer M711530 (2 passes). The carboxylic content of TCNFs, determined by conductometric titration, was approximately 1.0 mmol g 1 of dry CNFs. Detailed characterisation of these particular TCNFs can be found elsewhere. 42 This journal is © The Royal Society of Chemistry 2019 RSC Adv.,2019,9, 15668–15677 | 15669 Paper RSC Advances Open Access Article. Published on 20 May 2019. Downloaded on 5/21/2019 10:07:53 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online
2.2. Methods 2.2.1. Gentle drying of TCNFs using cryoxing. The TCNFs were cryoxed in LN 2 (cryo-LN 2 ) and in LPGS (cryo-LPGS). LPGS was prepared by liquefying propane gas using LN 2 and cooled until its freezing temperature (189 C) was nearly reached. A drop of TCNF-gel (1.08% [w/w]) was placed on a TEM grid (300 mesh) and immediately plunged in LN 2 (approximately 196 C) or in LPGS. The cryoxed samples were placed on a LN 2 cooled metal plate and dried in a freeze drier at 50 C under vacuum (Christ LOC-1m) over night. The dried samples were kept in a desiccator until HIM imaging or BET-analysis was carried out. 2.2.2. Gentle drying of TCNFs using solvent exchange and critical point drying (CPD). Two different methods for solvent exchange prior to CPD of TCNFs were used. Solvent exchange with GA, OsO 4 and EtOH was performed by rst attaching a drop of TCNF-gel (1.08% [w/w]) to a glass coverslip with epoxybased glue. Samples were placed in a 24-microtiter plate containing a xative (2% GA in a 0.1 M NaCac buffer, pH 7.4) and incubated for 4 h. The samples were then washed with 0.1 M NaCac buffer twice and incubated with 1% OsO 4 in 0.1 M NaCac for 30 min, aer which the washing with 0.1 M NaCac was repeated three times. Aer xation with GA and OsO 4 , the samples were dehydrated to EtOH by using a series of steps with increasing EtOH concentration: 50, 70, 90, 95 and 2 99.5%. The dwell time in each step was 30 min, and the nal step took place overnight. Solvent exchange with EtOH and AE was done rst by dehydrating drops of TCNF-gel in an EtOH stepwise, as explained earlier. Aer the last step in 99.5% EtOH the sample was placed in AE overnight. CPD (Leica CPD 300, University of Jyv¨ askyl¨ a, Finland) was the last step from EtOH or AE to the ambient conditions. The CPD programme included 16 exchange cycles of CO 2 at medium speed (speed value 5) without stirring. Slow speed was used for gas lling, heating, and venting steps. The dried samples were attached to metal stubs using carbon tape and kept in a desiccator until HIM-imaging. 2.2.3. Helium ion microscopy (HIM). HIM (Zeiss Orion Nanofab, University of Jyv¨ askyl¨ a Nanoscience Centre, Finland) was used for imaging the dried TCNFs. Acceleration voltage of 30 to 35 kV with aperture 10 mm was used resulting to an ion current of 0.1–0.3 pA. Image size 1024 1024 pixels, line averaging between 4 to 16 lines, dwell time 0.5 or 1.0 ms and working distance approximately 9 mm were used as the imaging parameters. All samples were studied without metal coating, and the electron ood gun with 750 eV energy was used to neutralize the sample charging. Fibril dimensions were estimated from HIM images by using ImageJ soware (ImageJ freeware, USA). The scale bar of the images was used to turn the soware pixels into nanometers and the brils width were collected from different spots of the image so that rough estimation of different bril widths could be done. 2.2.4. N 2 -sorption. The SSA of the TCNFs was determined using the Brunauer–Emmett–Teller (BET) method. 39 Approximately 70 mg of dried TCNFs was prepared for the analysis. TCNFs were dried using the procedures described before (Sections 2.2.1 and 2.2.2) with small modications. In cryoxing, the sample volume was increased to approximately 0.2– 0.3 g so that the structure was still able to freeze rapidly during the plunge-freezing in LN2 or in LPGS. In solvent exchange, the sample volume was increased to approximately 0.4 g (a large drop), incubation time in GA-solution was for overnight and in OsO 4 for 60 min. The dwell time in EtOH series was 60 min (the nal step took place overnight) and aer the last step in 99.5% EtOH the samples were placed in AE over two nights. Prolonged incubation times were conducted to ensure proper replacement of water in the structure. Samples were kept in a desiccator until N 2 -sorption measurement. The dried TCNFs was weighed in sample asks and degassed in a vacuum at 110 C for 30 min. Aer that the temperature was raised to 125 C for 4 h. Finally, the temperature was increased to 150 C for 15 min. The samples were then placed in a N 2 -sorption device (Micromeritics 3Flex Version 4.04, VTT Espoo) and the adsorption data was collected at 196 C by adjusting the relative nitrogen pressure from 0 to 0.99 and back. The Barrett–Joyner–Halenda (BJH) theory was used to calculate the average equivalent pore size of the TCNFs based on the N 2 -sorption isotherms. 43 The model is based on an assumption of spherically shaped pores, which is not the case in abril network system. Thus, the obtained values were mainly used as relative guidelines when comparing the samples. The average equivalent bril diameter (d) was also estimated from the SSA values using eqn (1). The density of cellulose was assumed to be 1600 kg m 3 (ref. 44 and 45) and brils were assumed to be innitely long rods with a cylindrical cross section. d¼ 4 rcBET (1) d¼bril diameter, r c ¼cellulose density and BET ¼BET specic surface area. 3 Results and discussion 3.1. HIM-imaging of TCNF-gel Both cryoxation and CPD preserved well the shape of the TCNFs. The samples could be handled without being fractured and not any clear shrinkage of the samples were observed. Cryoxed samples were white in colour and resembled dry polystyrene foam in appearance (Fig. 1a and b). CPD dried samples were light blue, transparent and resembled nestructured cottonwool (Fig. 1c and d). Blue colour could be a result of the Rayleigh scattering in the material with small length scales, previously observed also with TEMPO-oxidized liquid crystalline CNF-aerogels dried by using CPD 34 and silica-based aerogels. 46 Low magnication HIM images of dried TCNFs with a eld of view (FoV) of 400 and 100 mm presents dense and wavy surfaces (Fig. 2a–k). Waviness was most probably caused by sample handling and liquid uctuations during the drying. In order to see the differences between the actual bril structures, a closer investigation with higher magnication (FoV 10 mm) was needed. High magnication HIM images (Fig. 1c, f, i and l) 15670 |RSC Adv.,2019,9, 15668–15677 This journal is © The Royal Society of Chemistry 2019 RSC Advances Paper Open Access Article. Published on 20 May 2019. Downloaded on 5/21/2019 10:07:53 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online
show that TCNF-surfaces consisted of a very ne brillar material. The CPD-dried samples resembled each other also at higher magnication. Rough estimation of bril dimensions from the HIM images showed approximately 20 nm-wide brils, or bril bundles, in all samples. Accurate statistical analysis of bril dimensions and distributions was not possible as the sample degraded during imaging or pictures were too noisy (Fig. 3). TCNFs are found to consist of single brils of 3–4nmin width 14 but brils of those dimensions could not be distinguished here. There were also much thicker 100 nm wide bril bundles in the cryoxed samples. Similar structural differences between freeze dried and CPD dried commercial-CNF 36 and enzymatic-CNF 15 have been observed before with SEM. The formation of clear TCNF-lms was more severe in cryoLN 2 samples than cryo-LPGS samples. The heterogeneous structure and the lm formation indicated that there was ice crystal formation during cryoxing of the sample that pushed the ne material to the edges of the ice crystals. 10,16,40,47 The CPD-dried samples showed a more homogenous microstructure. In addition to the small size, TCNFs have high negative charge and high specicsurfacearea, 14 meaning that they bind a lot of water in their structure. Removal of the water in such a way that the ne bril structure remains open is challenging and was not fully achieved with the used cryoxmethods. Detection of single brils turned out to be challenging because of the sample degradation during the imaging. Especially, the CPD-dried TCNFs were sensitive at higher magnications. Imaging, as in Fig. 2c,f,iandl,wasconducted by rst using the 2 mm FoV to focus a beam and align the ood gun and then the actual image was taken with 10 mmFoV. Because of the focusing step, a clear hole was formed to the centeroftheimageascanbeseeninFig.2iandl.SmallerFoV had a higher ion dose per area and to conrm that the effect was dose based, the imaging was done to the fresh areas with Fig. 1 Dried TCNFs attached on black carbon tape ready for HIM imaging. (a) cryo-LN2, (b) cryo-LPGS (white epoxy glue shows underneath the slightly transparent sample), (c) CPD (EtOH, AE) (d) CPD (GA, OsO 4 , EtOH). Fig. 2 HIM-images of dried TCNFs with different magnifications. (a), (b), (c) cryo-LN 2 , (d), (e), (f) cryo-LPGS, (g), (h), (i) CPD (EtOH, AE) and (j), (k), (l) CPD (GA, OsO 4 , EtOH). Top row FoV 400 mm, middle row FoV 100 mm and bottom row FoV 10 mm. Fig. 1i and l show clearly the holes caused by the ion beam. This journal is © The Royal Society of Chemistry 2019 RSC Adv.,2019,9, 15668–15677 | 15671 Paper RSC Advances Open Access Article. Published on 20 May 2019. Downloaded on 5/21/2019 10:07:53 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online
different doses by adjusting the amount of averaging. The total ion dose was calculated with eqn (2): D¼ It eA(2) D¼areal dose, I¼ion current, t¼pixel dwell time and A¼pixel area. With dose of 4.9 10 13 ions per cm 2 some brillar network could be resolved, but the noise level was quite high, which caused the graininess of the image (Fig. 3a). When the dose was doubled (Fig. 3b), some deformations of the bril structure was detected already. Interestingly, this dose was under the theoretical limit of the sub-nanometer imaging because there were 0.9 ions per nm 2 . When tenfold dose was used (Fig. 3c) the structure of the bril network was collapsed leaving some individual brils driover the imaging area. Degradation of the bio-based materials during the HIMimaging has not been described in the literature previously. For example, Joens et al. (2013) 26 demonstrated how biological organisms can be imaged with high magnications without damage and in the study by the Lepp¨ anen et al. (2017) 22 the dose 1.1 10 16 ions per cm 2 was used to image the dried agar gel network of 10 nm brils with no clear damage. Comparison of the ion doses among the literature is challenging because usually not all the imaging parameters are listed. Fox et al. (2013) 48 studied a graphene akes with HIM and Raman spectroscopy and found that dose of 10 17 ions per cm 2 was required for proper edge contrast to obtain sub-nanometer resolution and 5 10 14 ions per cm 2 already caused signicant damage to graphene lattice. Livengood et al. (2009) 49 studied the defect formation in the silicon and copper by HIM and found that with over 5 10 15 ions per cm 2 , subsurface lattice dislocations are found from the TEM cross-sections. These materials are quite different compared to the polymeric CNFs and the direct comparison is not possible. CNFs consist of cellulose chain bundles with alternating amorphous and crystalline regions having intermolecular hydrogen bonding between the chains on the crystalline part. 50,51 The amorphous regions of the network can be considered the weakest points of the structure. Most probably, the ionization of the cellulose by the ion beam, especially in the amorphous regions collapses the structure. Actually, single ion can cause several ionizations because it is known that secondary electron yield of helium ranges from 3 to 10 depending on the material 18 Crystalline CNCs, in which the amorphous regions are no longer present, have been imaged with HIM without similar degradation. 21,28 However, the quality and magnication of the images and sample preparation methods cannot be directly compared to this work. Cellulose-based materials have been found to be a highly sensitive also to the electron beams and the imaging of the single cellulose nanobril or cellulose nanocrystal is challenging with TEM. 52 More specically, a critical dose where the diffraction from the crystalline part of nanocellulose has been halved as the sign of destruction has been found to be about 6 10 15 electrons per cm 2 . This is about 10-times more than the helium ion dose causing the collapse of the brillar network in our experiment. An interesting application of the beam damage is demonstrated in Fig. 4, which shows the time series over the sample area with 0.3 pA ion current with total imaging time of couple of minutes. The overall structure in the rst gure was quite plain with no individual brils observed. When the same area was imaged multiple times, a hole started to form in the right-upper corner as a loosely bound lm-like bril layer on top of the sample milled away revealing the underlying bril network. This, so-called “unwrapping”property of the ion beam could be generally utilized to detect different materials based on the milling rate. The image area shis a little bit upward and right during the imaging, which was also a common nding. The bril network underneath was most likely collapsing and resulted in an overall change of shape. In principle, helium ions can penetrate tens of mm deep with this acceleration voltage, which means that also the structure underneath the surface can be damaged. Traditionally OsO 4 has been used to reduce sample charging during imaging with SEM. 53,54 In the current study the xation with glutaraldehyde and OsO 4 (Fig. 1l) did not protect the TCNFs against degradation. Atomic layer deposition or chemical vapour deposition of titanium oxide 16,31 or sputtering of a Pt or Au/Pd layer on the sample surface could provide a protective layer for brils and enable imaging with high magnications; however, metal sputtering can distort Fig. 3 The effect of the ion dose to the beam damage on the TCNFs (a) ion dose 4 10 13 ions per cm 2 (b) 9 10 13 ions per cm 2 (c) 4 10 14 ions per cm 2 (eqn (2)). 15672 |RSC Adv.,2019,9, 15668–15677 This journal is © The Royal Society of Chemistry 2019 RSC Advances Paper Open Access Article. Published on 20 May 2019. Downloaded on 5/21/2019 10:07:53 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online
the bril dimensions 26,55,56 and possibly damage the ner bril structure. 3.2. Specic surface area (SSA) of TCNFs Quantitative evaluation of the differences between the gentle drying methods of TCNFs was done by using BET-analysis, which determines the specicsurfacearea(SSA)ofamaterial that is accessible to nitrogen. In the case of TCNFs, the higher the SSA, the more open the structure and the less coalescence of brils has taken place during drying. All of the samples had a type IV N 2 -sorption isotherm curves with a type H3 hysteresis loop (Fig. 5), which means monolayer-multilayer adsorption of nitrogen on mesoporous structure with pore widths between 2–50 nm. 57,58 This hysteresis type typically indicates that the structure was formed of aggregates of platy particles with slit-shaped pores (Fig. 6). 11,32,57–60 This slit-shape possibly makes pores prone to collapse during drying, as surfaces want to minimize their energies by binding to each other and the closer they are, the stronger are the surface energies leading to collapse. 61 Samples dried with cryo-LN 2 did not show any hysteresis loop in the N 2 -sorption isotherms. When the adsorbed quantities were plotted as relative values (Fig. 5b; the adsorbed quantity divided by the highest detected adsorption), a small hysteresis loop could be detected. There Fig. 4 Time series of the same area of TCNFs dried with CPD (GA, OsO 4 , EtOH) showing the effect of beam damage (or ion milling) on the sample surface within seconds when using the 0.3 pA ion current. Total imaging time 80 seconds (8 s per image). Imaged using a 16-line average with FoV 1 mm. Scale bar is 100 nm for all the images. Fig. 5 (a) N 2 -sorption isotherms of TCNFs dried with gentle drying methods (b) relative N 2 -sorption isotherms of cryo-N 2 and CPD (EtOH, Ac) dried TCNFs. This journal is © The Royal Society of Chemistry 2019 RSC Adv.,2019,9, 15668–15677 | 15673 Paper RSC Advances Open Access Article. Published on 20 May 2019. Downloaded on 5/21/2019 10:07:53 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online
was a steep increase in the isotherms aertherelativepressure of 0.82, and because capillary condensation in mesopores occurs at the higher pressure values, the result indicates that most of the pores were larger in size. This was also seen in the pore size distribution (Fig. 7.) of the samples where most of the pores were in the mesopore and micropore range (15–100 nm). Over 50 nm pores or pores under 2 nm cannot be detected accurately by BET, and for this reason, it is probably not the most suitable method for brillated materials like TCNFs, but the results were still useful for comparing differences between thesamplesinthecurrentstudy. Table 1 shows the mean values of SSA, equivalent pore size, pore volume, and bril diameter values. The Barrett–Joyner– Halenda (BJH) method was used to calculate the pore size distributions of the samples. The CPD-dried samples adsorbed more nitrogen than cryoxed samples and had higher SSA, equivalent pore size and pore volume. CPD (EtOH, Ac) samples had the largest SSA (386 m 2 g 1 ), equivalent pore size (14.3 mm) and pore volume (1.4 cm 3 g 1 ). The equivalent pore size was similar to the rest of the samples (around 11 mm) but the pore volume was signicantly lower for cryoxed samples than for CPD samples. Less than 2% of all pores in all samples were of micro scale, and the rest of the pores were mesopores or larger (Table 1 and Fig. 7). This could mean that even though the volume of the pores in cryoxed samples was lower than in CPD samples, the pore structure was similar. According to the results, CPD drying was able to prevent the coalescence of the bril structure better than cryoxing; however, most of the micropores in the structure were not preserved or could not be detected. As high SSA values as 500–600 m 2 g 1 have been reported for TEMPO-oxidized liquid crystalline CNF-aerogels dried by using CPD and EtOH–solvent exchange. 34 The liquid crystalline arrangement achieved by acid-treatment could have promoted the structure stability, which was also observed as high toughness of the dried material. CPD has been reported to be a promising drying method also for TCNF-nanopapers, yielding SSA values of 480 m 2 g 1 measured by N 2 -sorption 40 and TCNFs dehydrated with EtOH and AE before CPD showed higher nitrogen adsorption and SSA than samples xed with GA and OsO 4 before dehydration in EtOH. The last AE step could be more benecial for CPD as the CO 2 -gas is more miscible with AE than with EtOH. 62 In electron microscopy, GA in combination with OsO 4 is a commonly used protein xative of both plant and animal samples. 53,63 On the other hand, it has been observed to be more efficient in preservation of internal plant structures than surfaces, 63 and did not provide any signicant support for cellulose-bril structures against the ion beam damage in the current study. Again, cryo-LPGS samples showed higher nitrogen adsorption and SSA (172 m 2 g 1 ) than cryo-LN 2 samples (SSA 42 m 2 g 1 ). LN 2 is known to suffer from the Leidenfrost-effect, and the low SSA was probably a result of the coalescence of brils by the ice crystal formation. The corresponding results of drying efficiency of cellulose microbrils (CMF) with CPD, LN 2 and LPGS have been previously reported. 15 Sehaqui et al. (2011) 11 reported a SSA of 150–280 m 2 g 1 for TCNFs dried from water using solvent exchange to EtOH and tert-butanol and cryoxing with LN 2 . Exchanging the water inside the material to a solvent with low surface tension has been shown to increase the SSA also for regenerated cellulose (160–190 m 2 g 1 ). 10 Thus, it could be possible to increase the SSA of cryo-LPGS samples by rst conducting a solvent-exchange of the samples, like it is done with samples prepared for CPD. On the other hand, if the target is to image TCNFs as they appear in water, where certain charges and interactions between brils occur, the solvent-exchange from water to a non-polar solvent could also change these interactions and the surface structure. Thus, it is not obvious that the higher SSA value of the solvent-exchanged samples really describes the sample structure in aqueous conditions. The calculated equivalent bril diameters of TCNFs were approximately 7 nm, 15 nm and 60 nm for CPD, cryo-LPGS and cryo-LN 2 dried samples, respectively. Large bril diameters of Fig. 6 A sketch of possible configuration of a slit-shaped pore between two cellulose nanofibrils. Fig. 7 Estimated pore size distribution obtained using BET method of the TCNFs dried with different gentle drying methods. Average pore width (nm) on the x-axis at logarithmic scale and pore volume (cm 3 g 1 ) on the y-axis. 15674 |RSC Adv.,2019,9, 15668–15677 This journal is © The Royal Society of Chemistry 2019 RSC Advances Paper Open Access Article. Published on 20 May 2019. Downloaded on 5/21/2019 10:07:53 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online
cryoxed TCNFs can be explained by the bril agglomeration during drying. Multilayer adsorption of nitrogen can increase the detected surface area, which decreases the calculated bril diameters, and the actual diameters were most likely larger. The BET results (Table 1) supported the observations from the HIM images of TCNFs (Fig. 2). CPD resulted more homogeneous and ner bril structure, with less intensive lm formation compared to cryoxing. As mentioned before, the smallest brils observed from the HIM images were approximately 20 nm in width (Fig. 2c, f, i and l), which was signicantly larger than the width estimates from the BET-analysis. This could be due to beam damage during the imaging, making the smallest brils disappear or due to the multilayer adsorption of nitrogen in the BET-analysis that affected the SSA calculations. In addition, the HIM-images show only the surfaces of the samples and do not represent the whole structure. Quantitative differences between different drying methods was not obtained by microscopy, and for that N 2 - sorption and BET-analysis were needed. Based on these results (Fig. 2 and Table 1), the solvent exchange in EtOH/AE combined with CPD drying is the most preferred method, and also involves less hazardous chemicals and liquid exchange steps than treatment with GA/OsO 4 /EtOH. Solvent exchange with AE combined with CPD is found to result in a high SSA also with the bres; 64 however, is worth to keep in mind that solvent exchange could modify the interactions between the brils that occur in water. TCNFs have added carboxyl groups on the cellulose chain, which increases their hydrogen bonding ability in water. When water is exchanged to the less polar media the hydrogen bonding is hindered and can be responsible for more open bril structure in the dried material. Cryoxing in LPGS surpasses LN 2 in SSA values, but cryoxing in LN 2 is more simple and faster than cryoxing in LPGS. In order to select a suitable drying method for TCNFs one needs to consider if a highly preserved structure is necessary, and how much time and effort is practical to use. 4 Conclusions The suitability of HIM for imaging the porous TCNF-aerogels with high resolution was evaluated and different aerogel preparation methods using gentle drying were compared. Highresolution HIM-imaging of TCNFs was compromised by the dose-related damage as not described before with ion beams. Further research is needed about ion beam induced damage on the organic materials to have reliable imaging methods in the future. Comparison of the different gentle drying methods showed that all methods preserved the wet structure of TCNFs at some degree. CPD was considered to be the best method for drying delicate samples with SSA of 386 m 2 g 1 . Cryo-LPGS provided moderate result with SSA of 172 m 2 g 1 , but SSA for cryo-LN 2 was only 42 m 2 g 1 , and should be carefully considered if detailed surface structures of wet cellulose bril materials are studied. Sample handing procedure in the preparation phase seemed also to affect the large-scale structures of the sample, but clear systematic differences between the samples was seen only in nanoscale. Conflicts of interest There are no conicts to declare. Acknowledgements This research was part of the Academy of Finland funded project ExtBioNet (Tailored bre–bre interactions for boosted extensibility of bio-based bre networks, decision No. 285627). The authors thank Mr Panu Lahtinen (Biomass processing, VTT, Espoo) for providing the brillated cellulose materials and Ms Mirja Muhola (Thermochemical conversions, VTT, Espoo) for the BET-analysis of the CNF samples. References 1 N. Lavoine, I. Desloges, A. Dufresne and J. Bras, Carbohydr. Polym., 2012, 90, 735–764. 2 T. Saito and A. Isogai, Biomacromolecules, 2004, 5, 1983– 1989. 3 S. Ahola, M. ¨ Osterberg and J. Laine, Cellulose, 2008, 15, 303– 314. 4 K. Torvinen, F. Pettersson, P. Lahtinen, K. Arstila, K. Vinay, R. ¨ Osterbacka, M. Toivakka and J. J. Saarinen, Flexible Printed Electron., 2007, 2, 024004. 5 F. W. Brodin, Ø. W. Gregersen and K. Syverud, Nord. Pulp Pap. Res. J., 2014, 29, 156–166. Table 1 Mean values of the SSA, equivalent pore size and pore volume of TCNFs dried with different gentle drying methods BET surface area Pore size b Fibril diameter Pore volume a Micropore volume <2 nm m 2 g 1 nm nm cm 3 g 1 cm 3 g 1 Cryo-N 2 41.8 10.7 59.8 0.11 0.0012 Cryo-LPGS 171.8 11.6 14.6 0.50 0.0036 CPD (EtOH, Ac) 385.7 14.3 6.5 1.38 0.0193 CPD (GA, OsO 4 , EtOH) 375.0 11.2 6.7 1.05 0.0176 a Single point desorption total pore volume of pores less than 193.5 nm width at p/p o ¼0.990. b Desorption average pore diameter (4V/Aby BET). This journal is © The Royal Society of Chemistry 2019 RSC Adv.,2019,9, 15668–15677 | 15675 Paper RSC Advances Open Access Article. Published on 20 May 2019. Downloaded on 5/21/2019 10:07:53 AM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence. View Article Online