Study on the Preparation of Nanocellulose Powder and Its Formation Mechanism

Nanocellulose is a kind of cellulose based nano material with fantastic properties and numerous potential applications. However, due to the fact that nanocellulose exhibits colloidal properties when its concentration is high, the drying of nanocellulose has always been an urgent problem to be solved. To address this problem, the precipitation and drying mechanism of cellulose nanocrystal (CNC), one of the most common types of nanocellulose, was studied in this paper. The CNC was precipitated from the aqueous suspension by salting out to avoid the CNC colloidal state when concentrated in water. The obtained CNC precipitation with a small amount of water and the actively added electrolyte was dehydrated and puried by solvent displacement with volatile organic. Then CNC powder can be taken shape by drying the mixture of CNC and organic solvent with different drying methods. The mechanism of CNC precipitation from aqueous suspension and CNC powder formation from volatile organic solvent was studied. After comprehensive consideration, a method for preparing re-dispersible CNC powder was established, and the properties of this powder were studied, that provided a solution for the industrial preparation and application of nanocellulose.


Introduction
Cellulose, the oldest natural polymer and the most abundant renewable resource on earth, is mainly derived from woods, bamboos, cottons, etc. (Moon et al. 2011). Nanocellulose, as a source of cellulosebased nanomaterial, exhibits better performance than traditional cellulose, such as nano-scale effect (Dufresne 2013), high surface area (Phanthong et al. 2018), outstanding mechanical strength (Nystrom et al. 2018) and high surface activity (Yang et al. 2017), etc. Recent years, with the further development of research, nanocellulose is now widely applied in biomedicine (DeLoid et al. 2018;Salimi et al. 2019), nanocomposite materials (Liu et al. 2010;Vollick et al. 2017), aerospace (Kargarzadeh et al. 2015), and environmental areas, etc. (Abouzeid et al. 2019;Kong et al. 2018;Korhonen et al. 2011). Cellulose Nanocrystal (CNC) is one of the most common types of nanocellulose which is mainly prepared by acid hydrolysis method. Normally, CNC is stored in the form of aqueous suspension due to the large number of hydrophilic groups on its surface (Sinque eld et al. 2020). However, CNC aqueous suspension tends to transform into gel state when its concentration exceeds a certain range which limits the further increase of its concentration (Peng et al. 2011). Therefore, CNC is mainly stored, transported and used in the form of low-concentration aqueous suspension (0.1-10w%) that greatly limits its industrial application (Wang et al. 2019b). Thus, the effective drying of CNC suspension is an urgent problem to be solved.
To achieve convenient storage and transportation of CNC, not only CNC must be effectively dried, but the dried CNC can be redispersed in water. Ways to improve CNC powder re-dispersion property mainly include chemical modi cation on CNC surface and changing the drying method of CNC suspension (Huang et al. 2020). Chemical modi cation is considered as an effective method to improve CNC dispersibility (Shang et al. 2019). Velásquez-Cock, J., et al. used maltodextrin (MDX), which was a capping agent, to avoid irreversible agglomeration during CNF (cellulose nano ber) powder formation process (Velásquez-Cock et al. 2018). Liu et al. found that, when the grafting amount of PMMA on CNC was higher than 70%, the CNC dispersible powder can be successfully prepared (Liu et al. 2020). Zhang et al. also reported a new method to prepare dispersible CNF powder by modifying CNF with phenyl trimethoxy silane (PTS), which showed excellent dispersibility in water (Zhang et al. 2019). All of the above modi cation methods can obtain dispersible nanocellulose powder. However, the CNC's surface functional groups or chemical structures were changed during these chemical treatments. Importantly, the employment of toxic reagents would be harmful to the environment.
Another way to prepare nanocellulose powder is to change the drying method of its aqueous suspension.
The drying methods of nanocellulose usually include oven drying, freeze-drying, spray drying, and supercritical CO 2 drying (George and Montemagno 2017;Posada et al. 2020;Tsioptsias et al. 2008;Zheng and Fu 2019). Oven drying is a simple and economical way to obtain dried nanocellulose.
However, the bers of nanocellulose will become cross-linked (Uetani et al. 2018) due to the large number of hydrogen bonds on its surface which cause the nanocellulose to dry into membranes (Beck et al. 2012). Freeze-drying is a common method for preparing nanocellulose powder (Gray 2016;Varshosaz et al. 2012). But it still cannot obtain nanocellulose powder when the concentration of nanocellulose is higher than 0.5% (Han et al. 2013). Meanwhile, higher concentration CNC in the freeze drying will lead to the formation of a large number of hydrogen bonds between bers, resulting in irreversible agglomeration of the bers (Chen et al. 2019). Meanwhile, low concentration of nanocellulose can take a long time by freeze-drying (3-5 days) (Martinez Avila et al. 2015), which is uneconomical. Up to now, nanocellulose powder preparation by freeze-drying is only suitable for laboratory.
Spray drying and supercritical CO 2 drying can also prepare nanocellulose powder (Volk et al. 2015).
However, expensive equipment, high operating costs, and low powder yield severely limit its large-scale production (Lee et al. 2013). The above existing problems restrict the promotion and application of nanocellulose which exhibit excellent performance. In order to make effective use of nanocellulose, we have done a lot of work. For example, we used ammonium bicarbonate in the puri cation process of nanocellulose after acid hydrolysis, which avoided the use of expensive and ine cient dialysis membranes. After homogenization, high concentration aqueous nanocellulose suspensions can be directly prepared (Wang et al. 2018). This process not only cut down the cost of CNC preparation, but also further reduced the preparation time drastically.
In this contribution, CNC containing impurities can be simply prepared by homogenizing the ber components which were preliminary puri ed by centrifugation after obtaining from the dissolving pulp degradation with acid hydrolysis. After CNC was precipitated from the aqueous suspension by salting out, a volatile organic solvent such as ethanol was used to replace the water in the precipitation, and the actively added electrolyte can also be dissolved out at the same time. Then, the obtained CNC and organic solvent mixture can be precipitated and dried into pure CNC powder by different evaporation methods. The speci c experiment process was that cellulose was hydrolyzed by sulfuric acid (56wt% acid concentration at 55°C for 60 min). After initial centrifugation, the hydrolysate was neutralized with sodium hydroxide. The neutral hydrolysate was homogenized 5 times under the pressure of 20,000 psi by a micro jet nano homogenizer. We chose three kinds of electrolytes, potassium acetate, copper chloride, and ferric chloride, to precipitate CNC from the homogenized suspension by centrifugation with the purpose of removing impurities, respectively. After a few centrifuges, we got a CNC suspension containing only the actively added electrolyte as impurity. This portion of the electrolyte was removed by centrifugal replacement with a volatile organic solvent, such as ethanol. Then, pure CNC in organic solvent without any electrolytes can be prepared. After the organic solvent was removed by rotary evaporation or other evaporation methods, CNC powder can be obtained, nally.
This work opens the door to a new technology. A feasible and convenient method for the production of pure nanocellulose powder is proposed. In the following discussion, we explored the in uence of the agglomeration between the actively added electrolytes (concentration, valence) and CNC. After studying the mechanism that they bind to each other, potassium acetate was considered as the optimal electrolyte in this paper. The precipitation and formation mechanism of CNC powder was proved by studying the types and strength of the force between CNC and different dispersants during evaporation. In this section, we also noticed that the steric hindrance of dispersant had a signi cant in uence on the formation of CNC powder. Large steric hindrance can effectively avoid the combination of hydrogen bonds between CNCs, thereby further providing possibility for the preparation of CNC powder. In view of the successful preparation of CNC powder in this paper, we hope that it can facilitate the large-scale preparation, transportation, application of nanocellulose, and also can provides a solution for the industrial production and utilization of nanocellulose.

Materials and reagents
Hardwood bleached dissolving pulp was obtained from Shan Dong Sun Paper Industry Joint Stock Co., Ltd. Potassium acetate, Copper chloride, Ferric chloride, and 1.2-Propylene glycol were purchased from Shanghai Maclean Biochemical Co., Ltd. Methanol, Ethanol, N-propanol, N-butanol were produced by Tianjin Fuyu Fine Chemical Co., Ltd. All these chemicals of reagent grade were used as received without further treatment. Deionized water was used in all experiments.

Preparation of CNC powders
Hardwood bleached dissolving pulp (20g, oven dry) was added into 160ml sulfuric acid (56, 59, 61, 64wt%) in a three-necked round-bottom ask. The hydrolysis was stirred by using an electrical agitator for 60 min at 55°C. The gained suspensions were washed two times with deionized water by centrifugation (4000rpm, 10min), then NaOH was added until the pH was neutral and continued centrifugation. After centrifugation, these suspensions were homogenized by high-pressure homogenization (pressure of 20,000 psi) for 5 times. After that, we added potassium acetate to the homogenized CNC to remove impurities by centrifugation. Then, ethanol was added to the precipitate to remove the potassium acetate (electrolyte content can be measured by conductivity). Finally, the CNC powder was obtained by oven drying or rotary evaporating the CNC-organic precipitation.

Stability analyzer analysis
The stability of the puri ed CNC aqueous suspension was analyzed by LUMisizer (Germany) at a centrifugal speed of 3000 rpm and a temperature of 25°C.
2.3.3. AFG particle charge titrator analysis AFG particle charge titrator (Germany) is used to determine the surface charge of different states of CNC (before and after puri cation). The cation titrant is Polyelectrolyte sol.

Scanning electron microscopy (SEM)
Scanning electron microscopy (SEM) was done with Hitachi Regulus8220 (Japan). All samples need to be sprayed with gold before testing.

Thermogravimetric analysis (TGA)
Thermal properties of CNC powder were studied by thermogravimetric analysis (TGA) (Q50, USA) under a nitrogen atmosphere with an alumina sample pan. The sample was heated at 10°C/min from room temperature to 700°C.
2.3.6. X-ray diffraction (XRD) analysis X-ray diffraction patterns of the cellulose materials were determined using an X-ray diffractometer (Japanese Science SE) equipment with Cu Kα radiation in the 2θ range of 5 − 80°. The crystallinity index of cellulose materials was calculated according to the empirical method developed by Segal et al. (L. Segal et al. 1959).
I002 is the intensity of the lattice diffraction peak, Iam is the peak intensity of the amorphous region. A diffraction angle of around 2θ = 22.5° was the peak for plane (002), and the lowest intensity at a diffraction angle of around 2θ = 18.0° was measured as the amorphous part.

Atomic force microscopy (AFM)
Atomic force microscopy was performed by Multimode 8 (Germany) for specimen deposited on freshly cleaved mica, CNC concentration is 0.001%.

Study on the combination of CNC with different electrolytes
Generally, long time-consuming is an unavoidable problem in the process of preparing CNC by sulfuric acid hydrolysis. The main reason is that during centrifugal puri cation process, CNC is di cult to precipitate due to its light weight (Wang et al. 2019a) and high surface charge density (Abitbol et al. 2018), which make it di cult to remove impurities from CNC aqueous suspension. Normally, scientists use dialysis to purify CNC after sulfuric acid hydrolysis (Dong et al. 2016;Kim et al. 2016). The advantage of dialysis is that it can obtain pure CNC. However, the dialysis process is expensive and timeconsuming (Tang et al. 2013;Zhang et al. 2020), which greatly hinders the e cient preparation of CNC.
Some researchers found that adding a positively charged electrolyte in CNC dispersion can neutralize the negative charge on its surface (Fall et al. 2011;Fukuzumi et al. 2014). It would reduce the repulsive force between CNCs, and destroy the stability of CNC suspension system based on DLVO theory (Boluk et al. 2012;Phan-Xuan et al. 2016). Therefore, here, we propose a new method to prepare CNC powder. The process is adding an electrolyte to CNC aqueous suspension during the centrifugal puri cation. After centrifugation, the CNC-electrolyte precipitate is dispersed in a speci c organic solvent to remove the actively added electrolyte to gain the CNC-organic solvent mixture. Then, CNC powder can be obtained by oven drying or rotary evaporating this mixture. The preparation process is shown in Fig. 1(a). By comparison, we nd that the new method in Fig. 1(a) only needs several hours. However, the preparation of traditional CNC suspension needs 2-7 days (Cranston and Gray 2006). What's more, the new method can further prepare re-dispersible CNC powder, direct.
After preliminary attempts and analysis, we found that successful preparation of CNC powder requires an effective pairing of electrolyte and solvent. The basic conditions are as follows: (1) the addition of electrolytes can cause CNC suspension to precipitate during centrifugation; (2) the electrolyte has a good solubility in the matched solvent; (3) the solvent needs to be miscible with water which cause the electrolyte to be removed from water to solvent; (4) the solvent should be easy to volatilize. Based on the above considerations, we chose ethanol as solvent and CH 3 COOK, CuCl 2 , and FeCl 3 as electrolytes in this paper. The following steps were used to study their attribute on the preparation of CNC powder. First, we investigated the agglomeration ability of CNC with CH 3 COOK, CuCl 2 , and FeCl 3 , respectively. As shown in Fig. S1, we added different concentrations of electrolytes (CH 3 COOK, CuCl 2 , and FeCl 3 ) to CNC aqueous suspension (0.5%), then mixed well. Finally, we could obtain CNC-electrolyte precipitates after centrifuging the mixtures above, respectively.
In this test, we found that when adding different valence electrolytes in CNC suspension, the precipitate formed by high-valent electrolytes such as CuCl 2 and FeCl 3 was di cult to re-disperse in water which made it hardly to remove impurity ions in subsequent puri cation process. The probable reason for this phenomenon is that different valence ions have different agglomeration capabilities with CNC. Highvalent ions (Cu 2+ , Fe 3+ ) may have a strong agglomeration ability with CNC. Interestingly, in follow-up study, we veri ed the above conjecture in a speci c experiment. By observing the tightness of CNC precipitate with different valence ions, we found that a few high-valent electrolytes can easily cause CNC precipitation.
To gain a better understanding of the above nding, we have carried out a speci c investigation. Different amounts of CuCl 2 and FeCl 3 (0.2%, 0.1%, 0.05%, 0.02%) were mixed with CNC aqueous suspension and centrifuged. Then CNC with electrolyte was centrifugal several times, and the precipitates after each centrifugation were dispersed in a large amount of deionized water. In this test, only precipitate generated by adding 0.05% CuCl 2 or 0.02% FeCl 3 would not continue to produce precipitation during next centrifugation as shown in Fig. 1(b). The above experiment veri es our suppose that various valence electrolytes have a threshold value for agglomeration ability with CNC (Lyklema 2013). When it is less than this value, CNC with electrolyte do not produce precipitation after centrifugation. On the contrary, it can increase the density of precipitation, which bring more di culties to subsequent experiments. What's more, we further studied the minimum amounts of various electrolytes which can cause CNC precipitate and the results are shown in Fig. S2.
To gain more insight into the in uence of different valence electrolytes on CNC agglomeration, we added CuCl 2 and FeCl3 to 0.5% CNC aqueous suspension, respectively. Then dispersants containing CNC and electrolyte were centrifuged. The precipitate after centrifugation was puri ed with ethanol for 4 times. In addition, we conducted conductivity tests on each dispersed CNC aqueous to detect its metal ion content.
The results are shown in Fig. 2(c). As can be seen, the color of CNC precipitate (yellow or light blue) obtained in the above experiment (conductivity no longer changes) was different from pure CNC precipitate (milky white). This difference indicated that CNC precipitate prepared with high-valent ions was hardly to purify. To explain this, we postulated that CNC may be complex with high-valent electrolytes (such as Cu 2+ , Fe 3+ ). Thus, we dispersed the precipitate in deionized water. Then, a small amount of sulfuric acid and potassium acetate were added to CNC redispersion and centrifuged. The color of new precipitate had changed to white as shown in Fig. 2(c). The reason may be that, strong acid causes the bond between CNC and electrolytes to break. As a result, electrolyte ions can be re-dissolved in the solvent and the CNC precipitate can recover to its original color.

FT-IR spectra analysis
The bond between electrolyte and CNC we mentioned above may be a coordination bond formed by metal cation in electrolyte and hydroxyl group on CNC (Wu and Brodbelt 1994). Then, to verify the above suppose, we added different valence electrolytes into CNC aqueous suspension. After that, the CNC with electrolyte was stirred evenly, and dried. The dried samples were analyzed by Fourier transform Infrared (FT-IR) spectrum. The infrared spectra of all samples are shown in Fig. 2(a). All samples have strong absorption peaks at 3440, 2890, 1640, and 1430 cm -1 . Among them, the absorption bond at about 3440 cm -1 is related to the stretching vibration of O-H (Phanthong et al. 2017a), the peak at around 2900 cm − 1 is corresponding to the C − H stretching vibration (Phanthong et al. 2017b). The absorption band at about 1430 cm -1 results from the symmetric CH2 bending vibration (Squinca et al. 2020). Since electrolytes only affect the strength of hydroxyl group and do not react with other functional groups, the CH 2 peak is set as internal standard peak. Through the ratio of peak area OH to CH 2 to determine intensity change trend of the OH peak. The results are shown in Fig. 2(b). As content of Cu 2+ and Fe 3+ gradually increased, the ratio of the peak area of OH to CH 2 will drop signi cantly, which shows that intensity of the hydroxyl peak is continuously decreasing. However, the addition of K + do not cause this change which shows that CH 3 COOK is di cult to produce complex with CNC during mixing process.

Particle charge titrator analysis
Meanwhile, we have conducted surface charge tests on CNC at different stages. We added 0.3% CH 3 COOK, 0.03% CuCl 2 and 0.01% FeCl 3 to 0.5% CNC aqueous suspension, respectively. Before and after puri cation, surface charge of CNC-electrolyte has been measured and shown in Fig. 2(c). From the gures, we can notice that surface charge of CNC added with high-valent electrolyte dropped sharply compared with monovalent electrolyte. And only the surface charge of CNC-K after puri cation is close to its original value. Surface charge of CNC puri ed by adding CuCl 2 and FeCl 3 is much lower than that of original CNC. This also illustrate that CNC's hydroxyl group may complex with Cu 2+ and Fe 3+ .

Stability analyzer analysis
In addition to the methods above, the state of CNC complexes with high-valent electrolytes were also con rmed by Stability Analyzer. Here, we added some electrolytes (CH 3 COOK, CuCl 2 , FeCl 3 ) to CNC suspension, and then puri ed it with ethanol until the conductivity no longer changed. Finally. the precipitate was dispersed in deionized water and tested. From Fig. 2(d), it can be seen that transmittance of three CNC suspensions is similar, about 80%. This is because most electrolytes have been removed in the centrifugal puri cation. In Fig. 2(e), we can see that the integral transmittance of CNC-K was higher than others. This is most likely due to the complexation of Cu 2+ or Fe 3+ with CNC, which reduces the transmission. Thereby the light transmittance of CNC-Cu and CNC-Fe is slightly lower than that of CNC-K.
By analyzing the three experiments above, we found that high-valent electrolytes could form complex with CNC that in uence the stability of CNC dispersion. Thus, high-valent electrolytes such as FeCl 3 and CuCl 2 were not suitable to be used in this paper, although only a small amount of thus electrolytes can cause CNC precipitation. Consequently, the CH 3 COOK was selected to agglomerate CNC in subsequent experiments.

The in uence of different dispersants on CNC drying
On the basis of the analysis above, we found the optimum electrolyte and corresponding solvent for preparing CNC. Furthermore, when we tried to dry the new CNC with ethanol by rotary evaporation, we unexpectedly discovered that the nanocellulose suspension did not exhibit a gel state during the concentration process. After continuing to dry, the nanocellulose powder is obtained. In order to study the formation mechanism, we conducted the following series of experiments Upon drying process, CNC can get diverse drying morphology in different dispersants, which play a key role in CNC drying process. Generally, CNC is dried in water to form a lm. However, in the above experiment, the CNC powder can be obtained by drying with ethanol. We consider that the biggest difference between ethanol and water (as a dispersant) is that only one hydrogen bond can be formed between ethanol and CNC, while water can form two. Different number of hydrogen bonds formed between dispersant and CNC may change the dry form of CNC. In order to test the in uence of dispersant on CNC dry form, we tested several other dispersions such as methanol, tetrahydrofuran, acetone and 1,2-Propylene glycol. First, we added electrolyte to make CNC precipitate, then used the above organic dispersants to remove electrolyte salts by centrifugation, respectively. Finally, the dried CNC could be obtained by drying these CNC-organic solvent mixtures. After that, we observed the morphology of the dried CNC. The results are shown in Fig. 3(a). We can notice that CNC powder can be obtained with dispersions (such as methanol, ethanol, tetrahydrofuran, and acetone) which can form a single hydrogen bond with CNC. On the contrary, the dispersants (water,1.2-Propylene glycol) that form double hydrogen bonds with CNC formed CNC lm after drying.
After the above analysis, we postulate that, dispersant such as water or 1.2-Propylene glycol may generate multiple hydrogen bonds with adjacent CNCs. As drying proceeds, the CNC on both sides of dispersant will be aggregated by hydrogen bonding, resulting in the "zipper effect" (Zhou et al. 2019). Therefore, a large number of hydrogen bonds cross-links between CNCs are generated. On the contrary, dispersants (methanol, ethanol, tetrahydrofuran and acetone) that can form a single hydrogen bond with CNC will not cause "zipper effect". What's more, Fig. 3(b) shows the formation of different hydrogen bonds between above dispersant and CNC. It illustrates that CNC can be pulled to any position in dispersants that form hydrogen bonds only, such as ethanol, unlike the "zipper effect" formed by CNC drying in water. Meanwhile, the different movements between bers during CNC drying will also in uence their dry forms.
In previous experiment, we also found an unexpected result, the shape of CNC powder dried by ethanol is larger than that dried by methanol, under the same conditions, as shown in Fig. 3(a). The approximate reason is that steric hindrance of dispersant has an effect on the nal dry morphology of CNC. According to the perspective of steric hindrance, ethanol is larger than methanol, so CNC powder dried in methanol is ner. To gain more insight, we conducted further experiments to prove the above postulate. We used oven and rotary evaporation to dry CNC-methanol, CNC-ethanol, CNC-n-propanol, and CNC-n-butanol, respectively. The results are shown in Fig. 3(c), CNC-methanol, CNC-ethanol, and CNC-n-propanol are all bulk solids after oven drying, while the form of CNC after drying is powdery. A similar situation also occurs in rotary evaporation. Among them, CNC-methanol form a powder with big pieces, while CNC-ethanol, CNC-n-propanol, and CNC-n-butanol all form powder with small pieces after drying. The comparison shows that the shape of the dried CNC will smaller as the steric hindrance of the dispersant increases, until it forms powder. This result is entirely consistent with our suppose. It also shows that different drying methods will cause different consequences.
To further investigate the formation mechanism of CNC powder, the powders and lms prepared above were analyzed by scanning electron microscopy, and all results are shown in Fig. 4. Figure 4a and 4b show the microstructure of dried CNC which is produced by water and 1.2-Propylene glycol, showing that surface of the sample is smooth. But in Fig. 4b we can see obvious cracks between bers, which may be caused by the large steric hindrance. Figure 4c-4f show the microstructure of dried CNC prepared with methanol, ethanol, n-propanol and n-butanol by oven drying. The results show that the surface of samples is rough and bers are separated which are different from the previous two images. This is mainly due to the different number of hydrogen bonds formed between dispersant and CNC. It also shows that the CNC surface which is dried by methanol is relatively at, and connections between bers are tight. In contrast, the surface of other samples is uneven and shows many fractures due to the change of steric hindrance. This also explains why the dry form of CNC changed from bulk to powder. Figure 4g-4j show the microstructure of dried CNCs prepared with methanol, ethanol, n-propanol and nbutanol by rotary evaporation. These images also con rm this explanation. Although the dry solids are powders, the cross-linking of the bers within the powder can be greatly reduced with the increase of steric hindrance of dispersants.
Based on the above exploration, we consider that the successful preparation of CNC powder which is dispersible needs to meet the following three conditions. 1). Hydrogen bond: The number of hydrogen bonds formed between dispersant and CNC can in uence the morphology of dried CNC. After the previous analysis, we suppose that the existence of double hydrogen bonds will cause "zipper effect" between CNC during drying process. Then, CNC forms an inseparable hydrogen bonds with cross-linked network. In contrast, dispersion that form a single hydrogen bond will randomly connect CNC, and destroy this cross-linked network.
2). Surface tension: Water has a high surface tension (surface tension of water: 72.76 mN·m-1) (Kalová and Mareš 2015) which may cause CNC to be subjected to a strong tensile force when it is dried (Hanif et al. 2017). Compared with water, alcohol has a lower surface tension .60 mN·m-1) (Basařová et al. 2019;Hu et al. 2008;Soucˇkova et al. 2008;Yue and Liu 2016), thus alcohol can only produce lower tensile force during dry process. Consequently, dispersants with low surface tension can be more effectively to dry CNC into powder morphology.
3). Steric hindrance: Steric hindrance also plays an important role in the formation of CNC powder. Dispersants with high steric hindrance are di cult to cause tight cross-linking state of CNC during the drying process. It can be proved by the above SEM images that as the steric hindrance of the dispersant increases, the distance between CNC will also increase, which can make it di cult to produce dense hydrogen bond cross-links between CNC. Hence, the successful preparation of CNC powder also requires dispersants with large steric hindrance.
In view of the fact that CNC powder has been successfully prepared by the new method, in order to further explore its property, we compared the performance of prepared CNC powders (CNC-Ethanol, CNC-Nbutanol) and CNC lm (CNC-Water). Among them, CNC-Water is a CNC thin lm prepared by drying CNC in water. CNC-Ethanol and CNC-N-butanol are CNC powders prepared by solvent replacement and rotary evaporation.

Thermal stability of CNC powder
The Thermogravimetric (TG) test results of all samples are shown in Fig. 5(a) and 5(b). All samples have a little weight loss (about 1-3%) at low temperature (< 100°C) is corresponding to the evaporation of water. Among them, CNC-Water curve has a signi cant reduction at 300°C. CNC-Ethanol and CNC-Nbutanol show signi cant weightlessness at 220°C and 200°C, respectively, which is lower than CNC-Water. This is because that powdery CNC are loosely connected, exposing more surfaces, which may lead to lower thermal stability. At the same time, through the analysis of previous experiments, the bers in CNC powder dried with n-butanol are relatively loosely arranged. Therefore, its decomposition temperature is lowest, about 20°C lower than CNC powder dried with ethanol. From the TG diagram, we can conclude that there is a connection between the degree of looseness of CNC powder and the steric hindrance of dispersant which supports our previous conclusions.
3.4. The crystallinity of CNC powder Figure 5(c) and 5(d) show the XRD measurements and crystallinity of the samples. It can be seen from Fig. 5(c) that diffraction peaks appear at 16°, 22°, and 35°, and there is no double peak at the largest peak, so the prepared CNC is a typical type I cellulose crystal structure (Zhuo et al. 2017), which shows that different CNC drying forms would not affect their crystal structures. Figure 5(d) shows the crystallinity of the samples. We found that the crystallinity of these samples does not changed signi cantly. As a result, in this paper, different drying methods have no effect on the crystallinity of the CNC after drying (powder or lm).

Dispersibility of CNC powder
Since re-dispersibility is a very important property of CNC powder, we redispersed the dry CNC powder produced with different electrolytes (CH 3 COOK, CuCl 2 , FeCl 3 ) in deionized water to obtained the CNC redispersions. Then, we irradiated the redispersion with a laser pointer (λ = 650nm), and examined the redispersibility. It can be seen from Fig. 6, three kinds of CNC redispersions are accompanied by the obvious Tyndall phenomenon which prove their colloidal-like state and nano-size. However, the powders in 6(b) and 6(c) show different colors which indicate that there is a small amount of metal ions (blue represents Cu 2+ , yellow represents Fe 3+ ) in the powders. Meanwhile, we discover that a little precipitate was at bottom of the CNC redispersion as shown in Fig. 6(c), which also demonstrate the presence of complexes. For better comparison, the above CNC re-dispersions were observed by Atomic Force Microscope (AFM). The result is shown in Fig. 6. AFM observation shows that bers in CNC-K are obviously separated. On the contrary, bers in CNC-Cu or CNC-Fe are partially agglomerated. This result is consistent with what we said and discussed earlier that CH 3 COOK and ethanol are most suitable for this preparation method of CNC powder with re-dispersibility.
3.6. The yield and morphology of CNC prepared in new method We further detect the practical application performance of the CNC prepared in this article. Here, we compared the yield and ber morphology of the new method with the traditional method. We adopted the new method to prepare CNC with different concentrations (56wt%, 59wt%, 61wt% and 64wt%) of sulfuric acid, respectively. Then, in puri cation process, we used CH 3 COOK and ethanol as electrolyte and corresponding dispersant. The results are shown in Fig. 7 that the sulfuric acid concentration has a marked impact on the yield of CNC suspension. When the concentration of sulfuric acid increased from 56wt% to 64wt%, the yield of CNC dropped signi cantly. Among them, the highest yield reached 91.2% which produced by 56wt% sulfuric acid. The reason for this situation is that high sulfuric acid concentration will cause further acidolysis which degrade cellulose into glucose (Benini et al. 2018). And these results are similar to the published literature (Mahmud et al. 2019;Shahabi-Ghahafarrokhi et al. 2015).
The morphology of the samples observed by AFM were showed in Fig. 7. From AFM images, we can clearly see that the length of CNC will decrease while the acid concentration increases. The average length of CNC made from 56wt% sulfuric acid is 128nm, and the average length of CNC made from 59wt% sulfuric acid is 97nm, the average length of the CNC made from 61wt% sulfuric acid is 76nm, and the CNC made from 64wt% sulfuric acid has the shortest length, with an average length of 61nm. The length of prepared CNC is consistent with the results of previous research (Baek et al. 2013;Boluk et al. 2011;Jia and Liu 2019).

Conclusions
This paper proposed a novel method for preparing dispersible CNC powder. In the preparation process of nanocellulose, we innovatively adopted a new puri cation method by adding potassium acetate to the sulfuric acid hydrolysate of the dissolving pulp to remove impurity ions. Then, ethanol was used to get rid of the actively added potassium acetate to obtain concentrated pure CNC. The CNC powder was successfully produced by simple drying, nally. Then we further explored its formation mechanism by studying the number of hydrogen bonds formed between the dispersant and CNC, the surface tension and steric hindrance of the dispersant. The results showed that the above three factors would affect the nal morphology of CNC after drying.
Compared with CNC lm, the crystallinity of CNC powder did not change, but the thermal stability of CNC powder was decreased (about 80°C). The prepared CNC powder also has good dispersibility in water.
Meanwhile, the yield and morphology of the prepared CNC powder are comparable with those of the traditional method.
The formation of CNC powder is analyzed from the microstructure level. Since the gaps between microscopic bers in CNC powder are signi cantly larger than that in the CNC lm, this allows the CNC powder to be dispersed in water. The production of nanocellulose powder will have a signi cant in uence on transportation of traditional nanocellulose aqueous suspension, which can greatly reduce transportation costs.