Highly-porous uniformly-sized amidoxime-functionalized cellulose beads prepared by microfluidics with N-methylmorpholine N-oxide

Uniformly-sized porous cellulose beads functionalized with amidoxime groups were prepared for the first time using a microfluidic method with N-methylmorpholine N-oxide (NMMO) monohydrate as a cellulose solvent. The molten state cellulose dope in NMMO monohydrate (cell/NMMO dope) as a disperse phase and hot mineral oil as a continuous phase were used in a T-junction microfluidic chip to produce uniformly-sized cell/NMMO droplets. Coagulation of the molten state cell/NMMO droplet at high temperature and amidoxime functionalization could prepare the highly-porous spherical amidoxime-functionalized cellulose beads with a uniform fibrous open internal structure. The prepared amidoxime-functionalized cellulose beads showed excellent metal adsorption properties with a maximum adsorption capacity of ~ 80 mg g−1 in the case of Cu2+/phthalate ions. The newly developed highly-porous cellulose beads can open many new applications with other proper functionalization at the reactive hydroxyl groups of the cellulose.

Abstract Uniformly-sized porous cellulose beads functionalized with amidoxime groups were prepared for the first time using a microfluidic method with N-methylmorpholine N-oxide (NMMO) monohydrate as a cellulose solvent. The molten state cellulose dope in NMMO monohydrate (cell/NMMO dope) as a disperse phase and hot mineral oil as a continuous phase were used in a T-junction microfluidic chip to produce uniformly-sized cell/NMMO droplets. Coagulation of the molten state cell/NMMO droplet at high temperature and amidoxime functionalization could prepare the highly-porous spherical amidoxime-functionalized cellulose beads with a uniform fibrous open internal structure. The prepared amidoxime-functionalized cellulose beads showed excellent metal adsorption properties with a maximum adsorption capacity of * 80 mg g -1 in the case of Cu 2? /phthalate ions. The newly developed highly-porous cellulose beads can open many new applications with other proper functionalization at the reactive hydroxyl groups of the cellulose.

Introduction
Cellulose, a linear polysaccharide, is the most abundant and renewable biopolymer in nature (Klemm et al. 2011) and has been widely studied because of its unique properties, such as biocompatibility, biodegradability, eco-friendliness, cost-effectiveness, and easy modification. This natural polymer cannot be melted (by heating) and solubilized (in common organic solvents) because of its strong intra-and interchain hydrogen bonding. The cellulose can be a processible liquid form by derivatized chemical modification or direct dissolution with specific solvents. Traditional viscose technology uses derivatization in which cellulose xanthogenate is formed as an intermediate derivative. However, this process produces environmentally hazardous byproducts such as carbon disulfide and hydrogen sulfide (Fink et al. 2001). The direct dissolution process is purely physical without any structural changes (Rosenau et al. 2002). Of the many solvents for direct dissolution, N-methylmorpholine N-oxide (NMMO) is a well-known commercialized solvent (Lidhure et al. 2019). Nontoxic NMMO is economically favorable because it is mostly recycled and reusable (Dawson 2012;Rosenau et al. 2001). NMMO is used in hydrated forms to dissolve cellulose. According to the phase diagram of the NMMO/water/cellulose mixture, the NMMO monohydrate can dissolve cellulose up to 13 wt% (Lidhure et al. 2019). The NMMO solution usually contains stabilizers to prevent autocatalytic degradation of the NMMO, triggered by metal ions, acylating agents, and others (Rosenau et al. 1999(Rosenau et al. , 2005. Sayyed et al. studied the effect of the crystallinity of the cellulose pulp on the swelling behavior and critically reviewed the swelling process during dissolution of the cellulose pulp (Sayyed et al. 2019a;Sayyed et al. 2019b). The melting temperature of the NMMO monohydrate is * 76°C; therefore, the NMMO monohydrate should be heated to * 80°C to make the cellulose/ NMMO (cell/NMMO) dope. Upon cooling, the cell/ NMMO dope becomes solid at temperatures between 20 and 40°C (Biganska et al. 2002).
Cellulose and cellulose derivatives have been used as adsorbents in the form of hydrogels, films, beads, microfibers, and microcrystals (Hua et al. 2019). The cellulose beads can be prepared by sequentially forming spherical cell/NMMO dope, solidification via temperature decrease, and coagulation with water . The cellulose beads can be used for the stationary phase in chromatography, protein purification, and drug delivery (Alam et al. 2019). The cellulose beads prepared from the cell/NMMO dope have much porosity because a large amount of NMMO is extracted from the cell/NMMO dope during coagulation. Their high porosity enhances the interfacial area available for interactions with the target molecules when using the cellulose beads as adsorbents. However, the cellulose beads prepared from lower cellulose concentrations have weak mechanical properties (although they provide high porosity); therefore, crosslinking is necessary to enhance mechanical properties. The temperature and composition of the coagulation medium also influence the morphology, internal surface area, and pore size distribution. Hence, to take full advantage of cellulose beads, the porosity and surface area should be controlled and optimized with coagulation conditions because they determine the adsorption capacity of the cellulose beads. The size of the cellulose bead depends on the manufacturing method such that conventional dropping and dispersion techniques produce cellulose droplets with a size above and below * 350 lm, respectively, with a certain size distribution Rosenberg et al. 2007). Recently, the microfluidic method has become common for producing uniformly-sized particles. However, in our limited knowledge, highly controlled manufacturing methods for producing cellulose beads using microfluidics have not yet been developed using NMMO monohydrate as a solvent of cellulose.
The pollution of water resources with harmful metals, such as lead, mercury, cadmium, copper, iron, and chromium, is a serious threat to humans (Sharma et al. 2007). Thus, detecting and removing such harmful metals in water has been critical for modern society. Various efficient but expensive methods are proposed to remove heavy metal ions from industrial effluents using ion-exchange, reverse osmosis, and electrodialysis techniques (Bolto and Pawlowski 1987;Geckeler et al. 1988;Pawlowski 2015). Chemical precipitation is promising, but the generation of precipitated bulky hydroxides is often a major disadvantage. Functional cellulose beads were prepared as biocompatible adsorbents for water treatment using chemical functionalization or blending cellulose with organic and inorganic compounds. For example, the cellulose beads carrying carboxylate groups possess high adsorption capacity for calcium, copper, silver, and lead (Hirota et al. 2009), and those prepared by blending using alginate or chitosan were applied for removing cadmium and copper (Twu et al. 2003;Zhang et al. 2005). The cellulose modified using amine-containing functional groups (aminoalkyl and 2, 2 0 -diaminodiethylamine) or an anionic moiety (phosphate, thiolate, carboxyl, and carboxymethyl) has been reported for applications to metal adsorption (Saliba et al. 2000). Strong fixation of metal ions, especially multivalent ones, to cellulose beads has been achieved by coupling with specific ligands that possess multiple functional groups for metal complexation (Diviš et al. 2009;Matúš and Kubová 2005). Several chelating ligands such as diamines, iminodiacetic acid, or ethylenediaminetetraacetic acid have been coupled to cellulose beads (Boeden et al. 1991;He et al. 1999;Kahovec et al. 1980;Yeomans-Reina et al. 2001). The cellulose beads functionalized with amidoxime groups were also used as metal adsorbents (Saliba et al. 2000). Amidoxime groups have hydroxyamino (= N-OH) and amino (-NH 2 ) groups at the same carbon atom, offering the fused features of amide, oxime, amidine, and hydroxamic acid functionality (Li et al. 2013). Adsorption occurs through sharing or exchanging electrons between the transition metal ions and amidoxime ligand (Rahman et al. 2016;Wang et al. 2015). Amidoxime groups have no affinity for the common metallic cations (Na ? , K ? , Ca 2? , and Mg 2? ) but have a strong tendency to form a chelate complex with various transition and heavy metal ions (Pb 2? , Cu 2? , Fe 3? , and Cd 2? ) in an aqueous solution. The selective chelation of metal cations of the amidoxime group has been widely applied for fabricating functional materials. For example, chelating resins, gels, fibers, microbeads, and membranes functionalized with amidoxime groups were prepared for removing heavy toxic metal ions, enriching and selective recovery of precious metals from natural water and industrial wastewater, and detecting heavy metal ions (Li et al. 2013). The amidoxime-functionalized materials also have potential applications in CO 2 capture and storage (Zulfiqar et al. 2011).
In this study, uniformly-sized, highly-porous, expandable, and amidoxime-functionalized cellulose (O-cellulose) beads were prepared with the cell/ NMMO dope using a microfluidic method. Here stable cell/NMMO droplets in mineral oil were produced with T-junction at 80°C, followed by coagulation in THF/water mixture, crosslinking with epichlorohydrin, and amidoxime functionalization. The structures of the produced cellulose depended on the state of the cell/NMMO droplet (solid or molten) and the temperature of the coagulation bath (hot or cold). The relationship between the internal structure and adsorption properties was studied using samples prepared from a critical point drying (CPD) method with liquid CO 2 . The prepared O-cellulose beads showed high affinity to multivalent metal ions that were studied by Langmuir isotherm. The relationship between the internal structure and adsorption property was established, with the cellulose beads having various porous structures prepared by different coagulation conditions. The uniformly-sized porous cellulose beads can be easily expanded to other applications because of the easy introduction of different functional groups.
Preparation of uniformly-sized cellulose droplets Scheme 1a summarizes the overall schematics of producing cellulose beads using microfluidics. The following is a detailed description of the preparation procedure. The NMMO monohydrate (100 g) was melted in a round-bottom glass tube (outer diameter = 60 mm, length = 180 mm) at 80°C in which the cellulose powders (4 g) were dissolved using a screwtype propeller (diameter = 125 mm, pitch = 35 mm) at 80°C for 2 h. The dissolved cell/NMMO dope was degassed by maintaining it at 90°C for 2 h with a rubber cap on the top to prevent water evaporation in the NMMO monohydrate. The degassed dope was transferred to a plastic syringe (inner diameter = 20.05 mm) to use as a dispersed phase (DP) in microfluidics. Mineral oil with EM90 (3 wt%) in a plastic syringe (inner diameter = 15.90 mm) was used as a continuous phase (CP). The CP and DP were connected to the T-junction microfluidic chip (PEEK, P-713, IDEX, USA), (thru-hole diameter = 1.25 mm) with two separated syringe pumps (LEGATO 100, KD Scientific, USA).
A syringe containing the cell/NMMO dope and the T-junction microfluidic chip is in a bath filled with ethylene glycol at 85°C (Scheme 1a). The flowrates of the DP and CP (Q d and Q c ) were controlled at 2.0 and 23 lL min -1 , respectively, unless otherwise mentioned. When the cell/NMMO dope meets at the edge of the T-junction, the cell/NMMO droplet was formed. Four cellulose beads samples were prepared, depending on the cell/NMMO droplet's state during coagulation and coagulation temperature. The cell/NMMO droplet during coagulation can be a solid state (S) or molten state (M). The S cell/NMMO droplets were prepared by cooling the M cell/NMMO droplets in the CP of hot mineral oil by mixing them with a significant amount of cold mineral oil in a beaker that was placed into an ice-water bath before. The cell/NMMO droplets were coagulated at low S and high M coagulation temperatures, representing the S and M of NMMO monohydrate, respectively. The S cell/ NMMO droplets were coagulated by removing the mineral oil from the beaker containing the S cell/ NMMO droplets, refilling the beaker with THF to mix with the remaining mineral oil, removing THF, refilling the beaker with water at 30°C (S) or 60°C (M), and maintaining them for 30 min. The M cell/ NMMO droplets were coagulated using the following method. The CP hot mineral oil (85°C) containing the M cell/NMMO droplets appearing from the microfluidic chip was directly dropped into the THF/water mixture (50/50, v/v) at 30°C (S) or 60°C (M). The THF/water mixture was used during coagulation because the CP mineral oil containing the cell/NMMO droplets could not be removed in the water bath because of the immiscibility between water and mineral oil. When the melt-state beads coming out from the microfluidic chip were directly inserted into THF, THF removed the mineral oil of the continuous phase and the melt-state beads were coagulated into big droplets. Thus, it was impossible to coagulate the cell/NMMO droplets with a sequential THF and cold water treatment. The coagulated cellulose spherical beads were filtered using filter papers (Advantec, 110 mm) and washed with water several times. The

Preparation of O-cellulose beads
The cellulose beads were crosslinked with ECH (Scheme 1b). The ECH reacting with hydroxyl groups of the cellulose introduces oxirane moieties that can further react with hydroxyl groups of the same or another cellulose chain nearby (Gough 1967). The detailed reaction conditions are as follows. Briefly, the cellulose beads (0.1 g) were dispersed in an aqueous NaOH solution (3 wt%, 20 mL) under magnetic stirring for 30 min. The ECH/MeOH (3.08 g (1.54 g each), 1:3 mol ratio of anhydrous glucose unit (AGU):ECH) was added dropwise, under magnetic stirring at 60°C for 1 h. After the reaction, the reaction medium was washed with water until a pH 7. The crosslinked cellulose (X-cellulose) beads were dried at 25°C for 2 h in a vacuum oven. The amidoxime groups were introduced into the cellulose beads (Scheme 1b). The dried X-cellulose beads were further functionalized with AN to introduce the -CN groups as follows. Briefly, the X-cellulose beads (0.1 g (0.55 mM)) were dispersed in 20 mL AN into which a 0.56 mL MeOH solution containing 5 wt% TMA chloride was added. A 0.75 mL MeOH solution containing 10 wt% sodium hydroxide at 0°C was added dropwise. The reaction was maintained at 25°C for 2 h under magnetic stirring and finally, washed with water until the medium's pH reached 7. The CNfunctionalized cellulose (CN-cellulose) beads were further reacted with NH 2 OHÁHCl to introduce the amidoxime groups as follows. NH 2 OHÁHCl (8 g), Na 2 CO 3 (6 g), and CN-cellulose beads (0.1 g) were added to a 250 mL beaker, to which 100 mL water was added and sealed. The reaction was maintained at 70°C for 12 h. After the reaction, the O-cellulose beads were washed several times with water to remove the remaining salts. Water was replaced with MeOH, and the O-cellulose beads were dried in a vacuum oven at 25°C.

Critical point drying
To prepare cellulose bead samples for scanning electron microscope (SEM) and porosity/inner surface area determination, the CPD method with liquid CO 2 was used (Pinnow et al. 2008;Trygg et al. 2013) because the simple evaporation of water by vacuuming, heating, and air-drying results in the hornification of the bead surface and loss of their significant porosity and surface area . Lyophilization (freeze-drying) also causes water volume expansion upon freezing and the growth of ice crystals, resulting in a micro-and mesopore collapse in cellulose beads (Pinnow et al. 2008). For CPD with liquid CO 2 , water should be stepwise exchanged by ethanol, acetone, and liquid CO 2 , which is removed under supercritical conditions. To investigate the internal structure of the prepared cellulose beads using the CPD method, the CPD container was home-made ( Fig. S1). Detail CPD procedures can be found in Text S1.

Measurements
Attenuated total reflection (ATR)-Fourier transform infrared (FTIR) (ATR-FTIR) spectra were recorded in the range of 500-4000 cm -1 at an average of 64 scans using an FTIR spectrometer (Frontier, PerkinElmer, USA). The dried samples were ground into a fine powder using a pestle and mortar. Field-emission (FE)-SEM (FE-SEM; SU8220, Hitachi, Japan) images of the cellulose beads were obtained from the fractured surfaces of the platinum-coated epoxy-molded samples. The samples for SEM were dried using the CPD method with liquid CO 2 . The epoxy molding was performed by positioning the beads in the cap (Cavity Embedding Mold, Ted Pella, USA), filling the cap with an epoxy mixture of EPON (2 mL), DDSA (1.25 mL), NMA (1.25 mL), and DMP (0.075 mL), curing at 60°C for 24 h, and breaking by half with a vise. The cellulose beads were observed using an optical microscope (ANA-006, Leitz, Germany) equipped with a digital camera (STC-TC83USB-AS, SenTech, Japan) with a transmission mode. The Cu 2? / phthalate solution concentration was determined using ultraviolet (UV)-visible (vis) (UV-vis) spectroscopy (UV-2401PC, Shimadzu, Japan). The chemical composition of the cellulose beads was studied using an elemental analyzer (EA, Flash 2000, ThermoFisher, USA). The concentrations of the metal ions in water were investigated using an inductively-coupled plasma spectrometer (ICP, Optima 7300DV, Perk-inElmer, USA). The surface area of the cellulose beads (S-S, M-M) was determined using a surface area and pore size analyzer (Quadrasorb Evo, Quantachrome, Austria) using the Brunauer-Emmett-Teller (BET) method. The samples for BET analysis were dried using the CPD method with liquid CO 2 and pre-treated at 100°C for 12 h in a vacuum oven.

Adsorption of metal ions
A Cu 2? /phthalate complex aqueous solution was used to demonstrate the O-cellulose beads' performance on the Cu 2? ion detection using UV-Vis spectroscopy. The vacuum-dried O-cellulose beads were used for metal adsorption applications because large amounts of the cellulose bead could not be produced using the CPD method. However, the structure (in water) of the O-cellulose beads prepared by vacuum drying was almost the same as those prepared by CPD as shown in Figure S4. It was prepared by dissolving Cu(NO 3 ) 2 (controlled from 50 to 1000 ppm) and KHP (4.08 g, 0.2 M, excess) in water (100 mL). Fig. S2(a) shows the UV-Vis spectra of the Cu(NO 3 ) 2 (1000 ppm), KHP (0.2 M), and Cu 2? /phthalate (1000 ppm/0.2 M) complex aqueous solutions. The Cu(NO 3 ) 2 aqueous solution's UV-Vis spectrum shows a small peak at 780 nm. However, the Cu 2? /phthalate complex aqueous solution's UV-vis spectrum shows a strong absorption peak at 730 nm with a hypochromic shift, consistent with reported results (Saliba et al. 2000). The Cu 2? /phthalate complex aqueous solution's calibration curve (730 nm) was obtained at different concentrations with a linear regression curve (r 2 = 0.998) ( Fig. S2(b)). Each adsorption experiment was conducted thrice to investigate the adsorption behavior of metal ions onto the cellulose bead adsorbent. The concentrations of the aqueous metalion solutions before and after adsorption were measured using a UV-vis spectrophotometer (UV-1800, Shimadzu Corporation, Japan) at a wavelength range of 400-800 nm. The adsorption capacity (q t ) for the Cu 2? /phthalate complex was evaluated using Eq. (1).
where C 0 (mg L -1 ) and C t (mg L -1 ) are the dye concentrations at the initial time and time t, respectively, V (L) is the volume of the dye solution, and m (g) is the weight of the dried adsorbent.

Desorption of metal ions
For the desorption experiment, the metal-ion-loaded O-cellulose beads were immersed in 8 mL of 2 M HCl for 30 min under magnetic stirring at 25°C, and the O-cellulose beads were regenerated into the initial form after sequentially rinsing with water and pH 6 buffer solution several times. The adsorption test was performed again by magnetic stirring regenerated O-cellulose beads (4 mg) in Cu 2? /phthalate solution (4 mL, 1000 ppm) at pH 7 for 2 h. The adsorption/ desorption cycle was repeated five times to investigate the recyclability of the O-cellulose bead.

Results and discussion
Production of cellulose beads using microfluidics To find the optimum Q d and Q c in microfluidics, the cellulose (4 wt%)/NMMO dope was evaluated at different Q d and Q c in the T-junction microfluidic chip. The production of the cell/NMMO droplet in the T-junction microfluidic chip in the ethylene glycol bath could not be monitored in situ through an optical microscope; therefore, the cellulose (M-M) bead's shape after coagulation was examined as a model study. Figure 1a shows the matrix diagram of processability at different Q d and Q c . The matrix graph can be divided into three regions representing the nonprocessible condition (Region I) and conditions for producing asymmetric-prolate (Region II) and symmetric-spherical (Region III) cellulose beads. The inset figures show representative optical microscopy images of the cellulose beads. When Q c [ * 10.0 lL min -1 and Q d [ * 1.0 lL min -1 (Region III), the symmetric-spherical cellulose beads were produced. In the case of 6.0 lL min -1 \ Q c \ 10.0 lL min -1 (Region II), the asymmetric elongated prolate cellulose beads were produced. When Q c \ 6.0 lL min -1 (Region I), the droplets could not be produced. At low capillary numbers (low Q c ) in the T-junction chip, droplets are formed in a squeezing mode, forming a pressure gradient across the droplet as it is formed (Seemann et al. 2011). Droplets generated in the squeezing mode travel through the channel as a plug confined by the channel walls. As the Q c increases (i.e., the capillary number increases), droplet generation transits from the squeezing to the dripping mode. It is unclear whether Region III (highest Q c ) is in the dripping mode or not (Zhu and Wang 2017). For clarification, the cellulose/NMMO droplets produced at Q d = 4 lL min -1 and Q c = 30 lL min -1 (high Q c in Region III) were evaluated at different coagulation temperatures. Figure 1b, c show the optical microscopy images of the cellulose beads coagulated from the M and S cell/NMMO droplets at different temperatures, respectively. Both M and S cell/NMMO droplets are the same at the T-junction in the microfluidic chip. In the case of the M cell/NMMO droplets (Fig. 1b), ellipsoidal cellulose beads were produced at T coa = 10°C, 20°C, 30°C, and 40°C, although spherical cellulose beads were at T coa-[ 50°C. The cellulose beads prepared at T coa-[ 50°C represent cellulose (M-M) beads. The ellipsoidal shape of the cellulose beads produced at low T coa indicates that the cell/NMMO droplets at the T-junction were produced in the squeezing mode.
The shear stress generated at the T-junction of the microfluidic chip during droplet production can be easily released during coagulation in the case of the cellulose (M-M) bead because of high-temperature coagulation of the M cell/NMMO droplet. Thus, spherical cellulose could be observed for cellulose (M-M) beads, although they were formed in the squeezing mode. The squeeze mode is evident from the S cell/NMMO droplets (Fig. 1c). Ellipsoidal cellulose beads were produced at all studied T coa s because the ellipsoidal shape of the cell/NMMO droplets at the T-junction was fixed during cooling the cell/NMMO droplets to make the S cell/NMMO  (c) droplets. Thus, the cell/NMMO droplets at the T-junction are in the squeezing mode. We could not increase Q c to reach the dripping mode because of the limitation of the syringe pump because of the high viscosity of cell/NMMO dope.
The size distribution of cellulose beads produced under different coagulation conditions was examined.  occurs through water diffusion into the cell/NMMO droplet; therefore, the coagulation temperature is critical to control the internal porous structure of the cellulose bead because the produced coagulated structure is governed by the rate of water diffusion, which also strongly depends on the coagulation temperature. The internal structure was studied using a dried sample prepared using the CPD method.

Structure of the cellulose bead
Morphologies of the prepared cellulose beads were analyzed with samples dried using the CPD method. beads. During coagulation, water dilutes the NMMO content in the cell/NMMO droplet, and the diluted NMMO has low solvation power of cellulose, leading the phase separation. For cellulose (S-S) (Fig. 3a(i)) and (M-M) (Fig. 3a(iv)) beads, the internal structures are homogeneous, although the cellulose (M-M) bead is coarser and more porous than the cellulose (S-S) bead. The reason could be that the dissolved cellulose chains in the M cell/NMMO droplet are coagulated (phase-separated) with high mobility to form large phase-separated fibrous bundles in the case of the cellulose (M-M) bead. The surface areas of cellulose (S-S) and (M-M) beads measured using a BET method are 196 and 218 m 2 g -1 , respectively, indicating that cellulose (M-M) beads have a more porous structure, consistent with morphology data from SEM. Thus, spherical homogeneous cellulose (M-M) beads with open and porous structures were studied for further tests. For cellulose (M--S) beads ( Fig. 3a(iii)), M cell/NMMO droplets were directly dropped into the THF/water mixture (at 30°C) during coagulation. The outer surface of the cell/NMMO droplet in the molten state contacts the cold THF/water mixture in the coagulation beaker; therefore, the outer part of the cell/NMMO droplet becomes solid but the inner part is still in the molten state because the temperature cannot be cooled enough to become a solid state during initial coagulation. Thus, the structure in the outer part of the cellulose (M-S) bead is different from that in the inner part. The outer and inner parts were coagulated in solid and molten states, respectively, so that the structure of the core is coarser than that of the shell, as discussed with cellulose (M-M) and (S-S) beads ( Fig. 3a(i), (iv)). For cellulose (S-M) beads (Fig. 3a(ii)), the S cell/NMMO droplet was dropped into hot water (at 60°C) during coagulation; therefore, the S cell/ NMMO droplet was changed to the M cell/NMMO droplet by heating through hot water. However, the outer part was coagulated (phase-separated) quickly in the solid state before reaching the melting temperature for the S cell/NMMO droplet, although the thickness of the shell is small (Fig. 3a(ii)). Thus, similar to the cellulose (M-S) beads, the coagulation between the outer and inner parts is different to make the core/shell structure; therefore, the outer and inner parts are coagulated in solid and molten states, respectively.
The effect of the coagulation temperature on the thickness of the shell was studied. Figure 3b shows the D c /D T of cellulose beads (prepared from S and M cell/ NMMO droplets) as a function of the coagulation temperature (T coa ), where D c and D T are the core and total diameters (along the short axis), respectively. The D T and D c were measured using optical microscopy and SEM images (Fig. S3), respectively, because the samples in epoxy for SEM image were distorted during sample preparation. When D c /D T = 0 or 1, the cellulose beads are homogenous without a core/shell structure, although they are compact (closed) and coarse (fibrous)

Preparation of O-cellulose beads
The coagulated cellulose beads in water before drying had almost the same size as the cell/NMMO droplet. Thus, the dried cellulose beads should be shrunk significantly from the cellulose bead in the water because the solid content in the dope is only a few percent. When the dried cellulose bead is inserted into the water again for adsorption applications, the reswelling of the dried cellulose bead is critical to use as an adsorbent with a large surface area inside the cellulose bead. Fig. S4a shows the optical microscopy images of cellulose (M-M) beads without crosslinking before (in water) and after drying and reswelling of the dried cellulose beads. The completely dried cellulose did not swell to the original cellulose bead in the water, indicating that the hydrogen bonding bonds formed after drying prevent the reswelling of the dried cellulose beads. Fig. S4b shows the optical microscopy images of cellulose (M-M) beads with crosslinking. The completely dried cellulose beads swelled to the original cellulose bead in water, indicating that the crosslinking prevents hydrogen bonding formation after drying. Bigansk and Navard studied the phase separation during regeneration of cellulose from NMMO solutions (Biganska and Navard 2008). the dense and skin/core morphologies were observed from the solid (crystallized) solution and the molten solution, respectively. They explained small ordered spherical objects from the skin is due to a spinodal decomposition. Udoetok and his colleagues also reported the increased swelling after crosslinking (Udoetok et al. 2016). They explained that the greater swelling of the cross-linked polymers compared with pristine cellulose may be due to the reduced tendency of solvent to infiltrate the fiber domains of cellulose due to hydrogen bonding effects and the cross-linking of cellulose creates defect sites and micropores in the polymer framework that corresponds to an increase in swelling of cross-linked cellulose. Thus, crosslinking the cellulose bead promotes both stability and expansion ability. The expansion ability induced by crosslinking is necessary for further functionalization so that crosslinking was performed before derivatization with acrylonitrile. Table 1 shows the diameters of the X-cellulose beads prepared at different cellulose concentrations in the dope before (in water) and after drying, and after reswelling of the dried cellulose beads. The diameters of the swollen, dried, and reswollen X-cellulose beads are denoted as D 1 , D 2 , and D 3 , respectively. The diameter of the dried X-cellulose bead (D 2 ) increases as / increases because the solid content in the cell/ NMMO droplet increases as / increases. However, the ratio between reswollen and dried cellulose beads (D 3 / D 2 ) decreases as / increases. When crosslinking is performed at low /, it effectively blocked the hydrogen bonding; therefore, the small collapsed dried X-cellulose beads (because of low /) can be swollen much but less than the original size before drying. Accordingly, the D 3 /D 2 becomes larger as / decreases. However, the X-cellulose beads prepared at / = 2 and 3 wt% cannot be reswollen to the original size in water, but at / = 4 and 5 wt% reswelled to close to the original size (D 1 ); D 3 /D 1 = 0.5, 0.7, 0.9, and 1 at / = 2, 3, 4, and 5 wt%, respectively. When / increases to 5 wt%, droplet fabrication became challenging because of its high viscosity. Thus, the X-cellulose beads prepared at / = 4 wt% were used for further experiments.
The X-cellulose beads were functionalized with AN to make intermediate CN-cellulose beads that were further functionalized with NH 2 OH to make the final O-cellulose beads. FTIR spectroscopy confirmed the functionalization of X-, CN-, and O-cellulose. Figure 4a shows the FTIR spectra of the cellulose, X-cellulose, CN-cellulose, and O-cellulose beads. The FTIR spectrum of the X-cellulose bead is close to that of cellulose because of the similarity of their functional groups. The FTIR spectrum of the CN-cellulose shows a characteristic-CN stretching band at 2260 cm -1 . The FTIR spectrum of the O-cellulose shows additional peaks at * 920, 1610, and 1670 cm -1 because of the stretching vibration bands of -N-O, -N-H, and -C=N bonds in amidoxime (Saeed et al. 2008). In addition to FT-IR, The EA has been done to figure out the functionalization of the cellulose beads as shown in Table S1. The N atom cannot be detected in the cases of cellulose and X-cellulose. However, the N contents of CN-cellulose and O-cellulose beads are 4.31 and 10.05 wt%, respectively, indicative of the -CN and amidoxime Table 1 Diameters of the X-cellulose beads prepared at different / before (in water) and after drying, and after reswelling of the dried cellulose beads Dope concentration (/) (wt%) Swollen X-cellulose (D 1 ) (lm) Dried X-cellulose (D 2 ) (lm) Reswollen X-cellulose (D 3 ) (lm) Diameter ratio (D 3 /D 2 ) (volume ratio) The diameters of the swollen, dried, and reswollen X-cellulose beads are denoted as D 1 , D 2 , and D 3 , respectively. The diameter of the initial cellulose beads before crosslinking is 784 ± 24.8 lm for all /s functionalization for CN-cellulose and O-cellulose beads, respectively. Thus, the amidoxime groups were successfully introduced into the cellulose beads. The porous structures were studied using SEM with cellulose (M-M) beads and their functionalized cellulose beads (X-, CN-, and O-cellulose beads). Figure 5b, c show the SEM images of the outer and fractured surfaces, respectively, of cellulose, X-cellulose, CN-cellulose, and O-cellulose beads that were dried using the CPD method. The internal structures (from the fractured surface, Fig. 4c) of all cellulose beads are porous regardless of functionalization, although the outer surfaces exhibit a small, closed structure (Fig. 4b). The cellulose structure can easily collapse during the drying process because of the strong hydrogen bonding. The CPD is one of the best methods to preserve the swollen structure during drying. Thus, the open fibrous structure of cellulose beads can be observed. However, it is uncertain whether the observed closed structure from the outer surface is produced during CPD or initially exists because the even CPD method can cause some hornification during drying. To figure out the origin of this observation, studies are necessary but out of this study's scope. The degree of the crosslinking of X-cellulose and the conversion from -OH to -CN group of CNcellulose were calculated from the numbers of the -OH group reacted with ECH and converted to -CN group, respectively, among six -OH groups in two AGU units of cellulose. Table S1 summarizes the EA results of cellulose, X-cellulose, CN-cellulose, and O-cellulose beads, and Scheme S1 shows schematic drawing of the crosslinking and amidoxime functionalization. The C, H, and O contents of X-cellulose are 43.42, 6.58, and 50 wt%, respectively. When the crosslinking of cellulose with ECH occurs, two AGU units are crosslinked with a ECH unit, so that the molar molecular weight of the two units of the crosslinked AGU would be 358 ? 58x, where 358 is the molar molecular weight of two AGU, 58 is the increased molecular weight (CH 2 CH(-OH)CH 2 ) by crosslinking (see Scheme S1) and x is the crosslinked number among the six -OH groups. The C content in weight percentage would be (144 ? 36x)/ (358 ? 58x) 9 100%. The measured C content of X-cellulose is 43.42 wt%, so that the calculated x value is * 1. Thus, the crosslinked number among six -OH groups is * 1 and * 5 -OH groups are not reacted with ECH in average. The degree of conversion from -OH to -CN group for CN-cellulose beads was calculated. The molecular weights of CN-cellulose is 416 ? 54y, where 416 is the molar molecular weight of X-cellulose (358 ? 58 when x = 1) and 54 is the increased molecular weight by the -CH 2 CH 2 CN group. The nitrogen content in weight percentage is 14y/(416 ? 54y) 9 100 for CN-cellulose bead. The measured N content is 4.31 wt% for CN-cellulose beads, so that the calculated y value is 1.5. Thus, the degree of conversion from -OH to -CN group for CNcellulose bead is 30% (1.5/5 9 100).

Isotherm of the metal-ion on O-cellulose bead
The O-cellulose beads were used as an adsorbent for the Cu 2? /phthalate complex in water. The Cu 2? / phthalate complex enhances the intensity of Cu 2? in UV-Vis spectroscopy, as discussed before. Figure 5 shows the q t of the Cu 2? /phthalate complex on Ocellulose beads as a function of elapsed time (t elaps ) at different concentrations. Curve-fitting was performed using Eq. (2) of the pseudo-first-order kinetics (Fig. 5a) and Eq. (4) of the pseudo-second-order kinetics (Fig. 5b). The curve-fitted data of q e and k are the amount of adsorbed material at equilibrium and adsorption kinetic coefficient, respectively. The parameters of q e1 (and k 1 ) for the first-order kinetics and q e2 (and k 2 ) for second-order kinetics were calculated using modified Eqs. (3) and (5) with plots of ln q e À q t ð Þvs. t and t=q t vs. t, respectively (Fig. S5). Table S2 summarizes the curve-fitting results with parameters q e , k, and r 2 . The observed data are better curve-fitted with the second-order Eq. (4) than the first-order Eq. (2). r 2 are 0.9610 and 0.9919 for the first-and second-order equations, respectively. These with curve-fitting using a Eq. (2) of the pseudo-first-order kinetics and b Eq. (4) of the pseudo-second-order kinetics results indicate that the adsorption mechanism is because of the chelating reaction between the transition metal ions and the amidoxime group (Rahman et al. 2016;Wang et al. 2015).
ln q e À q t ð Þ¼Àk 1 t þ lnq e ð3Þ The q t of the Cu 2? /phthalate adsorption (at * 500 ppm) on pure cellulose and X-cellulose beads prepared with cellulose (4 wt%)/NMMO droplets at t elaps = 50 h were 1.13 and 0.87 mg/g, respectively, which are significantly smaller than that on the Ocellulose beads (68.5 mg/g, Fig. 5), indicating that the metal adsorption can not occur without amidoxime groups. The isotherm was performed using Langmuir (Eq. 6) and Freundlich (Eq. 8) models, where q m is the adsorption capacity (related to the number of available binding sites), C e is the aqueous phase concentration at equilibrium, K L and K F are the affinity constants for adsorption for the Langmuir (Eq. 6) and Freundlich (Eq. 8) models, respectively, and n is the index of heterogeneity. The parameters in Eqs. (6) and (8) were calculated by performing linear curve fittings with modified Langmuir and Freundlich isotherm Eqs. (7) and (9) with the Cu 2? /phthalate complex at a concentration range of 25-1000 ppm at 25°C (Fig. 6). Table S3 summarizes the calculated q m , K, and n. The r 2 values of the Langmuir and Freundlich isotherm curves are 0.99 and 0.95, respectively, indicating that the Langmuir model is wellfitted with observed data and a saturated monolayer (not multilayer) of solute molecules on the adsorbent surface was formed during adsorption. The affinity constant of K L is 0.013 LÁmg -1 , similar to the reported value (0.014 LÁmg -1 ) from poly(amidoxime) cellulose powder (Rahman et al. 2018). The calculated q m is * 81.30 mg g -1 , which is larger than the reported value (52.70 mg g -1 ) from amidoxime-functionalized poly (acrylonitrile) (PAN) nanofibers with an oxime group conversion of 25% (Saeed et al. 2008). Thus, the prepared O-cellulose beads from cellulose (M-M) beads have large adsorption capability of metal ions.
log q e ¼ log Internal structure and adsorption property relationship The O-cellulose beads made from cellulose (S-S), (S-M), and (M-S) beads were also evaluated as adsorbents for the Cu 2? /phthalate complex in water to establish the relationship between the internal porous structure and adsorption property. beads have the highest adsorption capacity because it has the highest porous structure, as discussed before, and vice versa for O-cellulose beads prepared from cellulose (S-S) beads. The O-cellulose beads prepared from cellulose (S-M) and (M-S) beads have adsorption capacity between them. Thus, we found that the internal structure of O-cellulose beads governs the adsorption capacity.

Application to fillers in the column
The O-cellulose beads were applied to fillers in a column as an application to a metal filter. The uniformly-sized O-cellulose beads could be packed into the column (Fig. S6). Figure 7b shows the concentration reduction of the Cu 2? /phthalate complex aqueous solution (initial concentration; 500 ppm) as a function of the cycle number by passing the complex solution (4 mL) through the column Other metal-ion adsorptions were studied using ICP. The Pb 2? , Cu 2? , Zn 2? , Ni 2? , Cd 2? , Fe 3? , Ca 2? , Cr 3? , and Mg 2? aqueous solution was prepared with the same concentration of 1000 ppm. Their concentrations were measured 24 h after O-cellulose beads (4 mg) were inserted into the vial containing the metal-ion solution (4 mL). Fig. S7 shows the adsorption capacity of Pb 2? , Cu 2? , Zn 2? , Ni 2? , Cd 2? , Fe 3? , Ca 2? , Cr 3? , and Mg 2? ions, which are 125. 0, 75.1, 62.5, 51.1, 34.5, 33.4, 11.1, and 0 mg g -1 , respectively, indicating that O-cellulose beads can remove hazardous metal ions.   7) and (9) for (a, c) Langmuir and b, d Freundlich models for adsorption of Cu 2? /phthalate complex on O-cellulose beads. The solid lines a, c represent the calculated graphs of Eqs. (6) and (8), respectively, with parameters obtained from linear regressions of (b, d) The Pb 2? is adsorbed most and Ca 2? , Cr 3? , and Mg 2? cannot be adsorbed on O-cellulose beads, similar to other reported results (Rahman et al. 2016;Saeed et al. 2008).
The desorption and re-adsorption of the Cu 2? / phthalate on O-cellulose beads were studied. The Ocellulose beads (4 mg) in the vial were evaluated with the Cu 2? /phthalate aqueous solution. The 1000 ppm Cu 2? /phthalate aqueous solution was adsorbed on Ocellulose beads for 2 h. The Cu 2? /phthalate-loaded Ocellulose beads were treated with 2 M HCl (8 mL) under stirring of 250 rpm at 25°C for 30 min to desorb the Cu 2? /phthalate from the O-cellulose beads.
The re-adsorption test was performed using the HCltreated O-cellulose beads after complete washing with water at the same conditions as the first adsorption. The adsorption amount of the Cu 2? /phthalate was measured again. The same adsorption and desorption tests were repeated for several cycles. Figure 7c, d show the adsorption capacity and efficiency of Ocellulose as a function of the number of cycles during the adsorption/desorption of Cu 2? /phthalate, respectively. The efficiency is the relative amount of the adsorption of Cu 2? /phthalate at the i th cycle compared to that at the (i -1) th cycle defined by Eq. (10). The adsorption capacity continuously decreases as the b Reduction of the Cu 2? /phthalate complex aqueous solution (initial concentration; 500 ppm) concentration as a function of the cycle number by passing the complex solution (4 mL) through a column (diameter 5.6 mm) filled with O-cellulose beads (70 mg), with a flow rate of 2.5 mL h -1 . The Cu 2? / phthalate complex aqueous solution was circulated for cycling using a syringe pump. c Adsorption capacity (q t ) and d efficiency of O-cellulose as a function of the number of cycles during the adsorption/desorption of Cu 2? /phthalate. The efficiency is the relative adsorption of Cu 2? /phthalate at the i th cycle compared to that at the (i -1) th cycle defined by Eq. (10) number of cycles increases, although the efficiency is constant at * 88%, indicating that the recovery of the oxime groups is high enough for reuse. However, efficiency can be improved by optimizing the desorption conditions. Efficiency at the ith cycle ¼ adsorption capacity at the i th cycle adsorption capacity at the i À 1 ð Þ th cycle Â 100

Conclusions
Uniformly-sized cellulose beads were successfully prepared by coagulating the cell/NMMO droplets that were produced using a microfluidic method with molten state cell/NMMO as a cellulose solvent. The state (solid or molten) of the cell/NMMO droplet and coagulation temperature can control the morphology of the cellulose beads. Coagulation of the molten state cell/NMMO droplet at high temperature (C 60°C) and amidoxime functionalization could prepare the highly-porous spherical O-cellulose beads with a uniform fibrous open internal structure. The prepared O-cellulose beads showed excellent metal adsorption properties with a maximum adsorption capacity of 80 mg g -1 in the case of Cu 2? /phthalate ions. The novel cellulose beads could be successfully applied to fillers in the column for removing hazardous metal ions.