(2020). Controlling the Processing of Co-Precipitated Magnetic Bacterial Cellulose/Iron Oxide Nanocomposites. Materials and Design.

• The starting reactants are not important for the characteristics and properties of the Fe 3 O 4 NPs without BC. • When NPs are formed inside BC networks, the starting materials in ﬂ uence the nanocomposites characteristics and properties. • Morphology, size, crystal structure, loading amount, and magnetic phase of Fe 3 O 4 NPs are affected by the BC net- works. • The ability toalterandcontrolthe prop- erties of the NPs enables one to direct towards speci ﬁ c target applications. reactants (Fe(II) and Fe(III) salts) in the fabrication and control of the properties of BC/iron oxide nanocomposites. It was found that the choices of starting reactants are not important for synthesizing NPs outside of the BC networks. However, the starting reactants do affect the formation of NPs when they are synthesized in the BC network.Signi ﬁ cantdifferences inthe morphologies, sizes,crystal structures, and magneticphasesof NPsoccurs when in this environment. The nanopores ofBC networksin some instances force the aggregation of theNPs, either within thepores,or on thesurfacesofthe ﬁ brils.Nanocompositessynthesizedfrom Fe(II)sulfate andFe(III) chloride were found to exhibit the highest magnetization. These nanocomposites have potential for ﬂ exiblesen-sors, actuators, or electromagnetic shielding. Nanocomposites from Fe(II) acetate and Fe(III) chloride, though exhibiting lower magnetization, preserve a porous structure. Thus, they have potential as adsorbents or for wound healing applications. The


Introduction
Bacterial cellulose (BC) is a fascinating and renewable natural nanomaterial, produced from the cultivation of gram-negative bacteria e.g. Gluconacetobacter xylinum. It has received a large amount of interest owing to its unique structural features and favorable properties, for example, remarkable mechanical properties, porosity, water absorbency, moldability, biodegradability, and excellent biological affinity [1,2]. Intensive research and exploration on BC nanomaterials have been conducted. This research has mainly focused on the biosynthetic process to achieve low-cost preparation [3][4][5], but also towards potential applications in a wide range of fields, such as paper making [6], food and food packaging [7], and medicine [8,9]. Nevertheless, BC lacks certain properties (antibacterial, electrically conductive, ferromagnetic), which limits its applications in various fields, such as an antimicrobial material, for electrical devices, batteries, sensors, or electromagnetic shielding. Therefore, the synthesis of BC nanocomposites, combining it with other materials, has been conducted to address these limitations [10][11][12][13].
It is therefore known that various forms of iron precursors can be used for the synthesis of BC/Fe 3 O 4 nanocomposites. Nevertheless, no systematic study has been reported on the effect of the starting reactants on the formation, morphology, size and distribution, and magnetic properties of magnetic BC composites, particularly where the NPs are formed in situ within the BC network. The objective of the present work is to bridge that gap, and investigate the effect of using different starting reactants in the control of the fabrication of BC/iron oxide nanocomposites for specific properties. This knowledge is significant for understanding the mechanism for the formation of iron oxide NP inside the BC network structure. It is also practically useful to select the optimized choices of the starting chemicals for the designed properties. Three Fe(II) salts and two Fe(III) salts were chosen as the starting reactants, resulting in 6 pairs of starting chemicals. The structure and morphology, as well as magnetic properties of each pair, are systematically studied and discussed in terms of the properties they exhibit. The understanding of the structure and properties of the magnetic BC nanocomposites via the controlled synthesis using the right pairs of starting reactants would suggest implications for the design for target applications.

Chemicals
The chemicals used in this work were iron (II) chloride

Biosynthesis of bacterial cellulose
Bacterial cellulose (BC) pellicles were prepared by cultivation of the bacterial strain Komagataeibacter nataicola (strain TISTR 975), supplied from the Microbiological Resources Centre, Thailand Institute of Scientific and Technological Research (TISTR). The culture was grown in a medium consisting of 100 g of anhydrous D-glucose (Ajax Finechem) and 10 g of yeast extract powder (HiMedia) in 1 L of de-ionized (DI) water. This combination was cultivated for 24 h at 30°C in a shaker incubator (150 rpm). Then, 1 ml of the cell suspension was introduced into a container containing 50 ml of a fresh liquid culture medium. The cultivation continued at 30°C under static conditions for 14 days. After that, the BC pellicle was harvested. It was purified by boiling in distilled water, following by soaking in NaOH solution for 24 h. Finally, it was rinsed with DI water several times until a pH of 7 was reached. The purified BC hydrogel was transformed into a BC aerogel by freezedrying which was kept at room temperature before further use.

Synthesis of BC/Fe 3 O 4 nanocomposites and Fe 3 O 4 nanoparticles
For the synthesis of the BC/iron oxide nanocomposites, firstly, the starting reactants of Fe 2+ (0.01 mol) and Fe 3+ (0.02 mol) were separately dissolved in 200 ml of DI water at 50°C with continuous stirring for 10 min, before mixing together. These processes were carried out in an ambient atmosphere. Then, the freeze-dried BC aerogels were then immersed in the iron ion solution and soaked for 90 min with continuous stirring. The color of BC pellicles changed from opaque white to orange during this process. Argon gas was then purged into the solution for 10 min. Subsequently, 400 ml of aqueous NH 3 was slowly added to the solution for the co-precipitation of iron oxide NPs. During this process, the BC pellicles gradually turned black. When the process was completed, the products were rinsed with water several times under ambient conditions to remove unwanted materials until a pH of 7 was obtained. The BC nanocomposites in the form of a hydrogel were oven-dried at 80°C overnight to obtain the magnetic BC sheets.
To investigate the effect of the starting reactants, iron (II) chloride (C), sulfate (S), and acetate (A), and iron (III) chloride (C) and nitrate (N), were used. Thus, a total of 6 pairs of chemicals were studied. The chemical reactions for the co-precipitation of each pair are as follows: The names of the BC nanocomposite samples are labeled as BC-Fe(II) Fe(III) according to the starting reactants for Fe(II) and Fe(III) ions. A total of 6 samples were fabricated, namely, BC-CC, BC-SC, BC-AC, BC-CN, BC-SN, and BC-AN, as a consequence of Eqs. (1)- (6). For example, BC-CN refers to the product from Eq. (4), where the starting reactants are Fe(II) chloride and Fe(III) nitrate.
For comparison, the synthesis of pure iron oxide NPs without BC was also carried out. The processes are the same, but only that the BC pellicles were absent from the solution. The iron oxide NPs are labeled as P-CC, P-SC, P-AC, P-CN, P-SN, P-AN.

Materials characterization
The surface morphologies of the samples were observed using a field-emission scanning electron microscope (FESEM, FEI-Helios, USA). Samples were gold coated prior to imaging with the SEM. A transmission electron microscope (TEM, FEI-Tecnai G2 20, USA) was used to investigate the size and distribution of Fe 3 O 4 NPs. Phases and crystalline structures of the samples were investigated using X-ray diffraction (XRD) with a diffractometer employing Cu-Kα radiation (PANalytical, Empyrean, USA) in the 2θ range of 10-80°. Magnetic properties measurements were carried out using a vibrating sample magnetometer (VSM) option in the VersaLab instrument (Quantum Design, USA) with the maximum applied field of 10 kOe. The functional groups of the BC/Fe 3 O 4 nanocomposites were studied using Fourier transform infrared (FTIR) spectroscopy (Bruker, TENSOR27, Germany) within the wavenumber range of 600-4000 cm −1 . Thermal properties of the samples were studied using a thermogravimetric analysis (TGA) technique (Hitachi-STA7200, Japan). In addition, X-ray absorption spectroscopy (XAS) of the Fe K-edge was conducted at BL1.1 W: Multiple X-ray Techniques at the Synchrotron Light Research Institute (SLRI), Thailand.

Results and discussion
The results on the synthesis of the iron oxide NPs prepared by coprecipitation from different starting reactants, but in the absence of the BC networks, were firstly presented. Fig. 1 presents SEM images showing the morphology of the iron oxide NPs. No obvious differences can be found among the samples. Clusters of irregular-shaped NPs were observed with sizes in the range 30-38 nm ( Table 1). The TEM images in Fig. 2 show a distribution of sizes of Fe 3 O 4 NPs. The boundaries of each particle are more visible, but, again, there are no distinct features between the samples synthesized using different starting reactants. The particle sizes of each sample varied in the range of 9-11 nm (labeled in Fig. 2 as well as in Table 1), which in comparison to the SEM images suggests a smaller ultimate NP size. This smaller size may be because of clustering of the NPs in the SEM images making it difficult to resolve individual particles.
Typical XRD patterns are presented in Fig. 3. These patterns enabled the phase and crystal structures of the NPs to be determined. The XRD pattern of each sample was compared to the standard Fe 3 O 4 reference (ICDD: 00-019-0629). No secondary phases were observed. This indicates that the co-precipitation method, whether using any pair of starting reactants, can form a magnetite phase, with a negligible change in the lattice parameters (8.314-8.350 Å). The crystallite size of each NPs, calculated by using the Scherrer equation, is in the range of 10-13 nm (Table 1), which is within the same range as values obtained from TEM analysis. It should be noted that the XRD pattern for the magnetite (Fe 3 O 4 ) phase is similar to the maghemite (Fe 2 O 3 ) phase (ICDD: 00-039-1346) (also shown in Fig. 3). Therefore, the synthesized NPs in this work could possibly be either the Fe 2 O 3 phase or the Fe 3 O 4 phase, or possibly a combination of both phases. The fact that it is very difficult to distinguish between these two phases has been previously noted in the literature [39,40].
The magnetic property measurements are shown in Fig. 4. The data for all samples are similar. The magnetization (M) versus magnetic field (H) curves follow a typical S-shape for the ferrimagnetism of magnetic oxides. An absence of a hysteresis loop, and the near-zero coercivity (H c ), imply superparamagnetism, which is thought to be due to the small sizes of the NPs. The magnetic susceptibility (slopes of the M-H curves) of all samples are almost identical, and the saturation magnetization (M s ) varies only slightly from 55 emu/g for P-SC to 61 emu/g for P-CC (Table 1). Since the morphology, size, and crystal structures of the NPs from each pair of reactants are similar, it is not unexpected that the magnetic properties of each sample are virtually unchanged. From the above results, it can be concluded that using different starting reactants in preparing iron oxide NPs by co-precipitation does not significantly affect the morphology, size, phase and crystal structure, or magnetic properties of the synthesized particles. Whether it is P-CC or P-AN, or other samples, they show no distinct characteristics. Any kind of ferrous or ferric salts can be employed in the co-precipitation process of iron oxide NPs. These findings are however contradictory to previous reports which found differences in the size and M s values for iron oxide NPs using various Fe reactants [41]. The discrepancy between the present work and others [41] could be attributed to several factors related to the details of the co-precipitation synthesis. Firstly, the alkaline agent in the present work (NH 4 OH) is different from NaOH used for reference [41]. The kinetics for the chemical reactions between NH 4 OH and NaOH may be different and, thus, the rate for converting iron ions into iron oxides is likely to vary. Dissimilar alkaline reagents would lead to resultant iron oxide NPs with differing morphologies and sizes. Moreover, the concentration of precursors, the time and temperature of co-precipitation all have an effect on the morphology and size of the magnetic iron oxides. Nabiyoui et al. found that increasing the co-precipitation time resulted in a larger particle size [42]. They also found that by lowering the magnetic ion concentration, larger NPs were observed, while heating during the process increased both size and M s of the NPs. Therefore, the results in the present work cannot be directly compared with other studies since the processing parameters were not systematically controlled for a direct comparison.
On the other hand, the morphology, phases, functional groups, and magnetic properties of the BC/iron oxide nanocomposites, prepared by co-precipitation in the presence of BC, show some variations when using different starting reactants. These differences are unexpected since the reactants are the same, however the medium in which they are produced has changed to a BC network. This environment of a fibrous network may present a particular pore size that inhibits the growth of the NPs during co-precipitation. Fig. 5 shows representative SEM images of the BC/iron oxide samples. Unlike the SEM images for the NPs, the nanocomposite samples show an obvious difference between samples. For instance, the iron oxide NPs appear to be densely packed within the 3D network of the BC nanofibers for BC-CC whereas the NPs are loosely formed on the surface of the BC nanofibers for BC-AC. The formation mechanism of the iron oxide NPs in the BC template is thus dependent on the starting reactants. The porous networks of the BC also form an environment in which the NPs are synthesized. It could be that given this environment, there is an alteration of the clustering and distribution of the NPs depending on the starting reactants. The size of the NP clusters in each sample is estimated from the SEM images as shown in Table 2. It is observed that the cluster sizes of the BC-CC, BC-SC and BC-CN are larger (75-86 nm) compared to the cluster sizes for BC-AN, BC-SN, and BC-AC. (55-65 nm). It is especially noted for  BC-SN and BC-AC, that the NPs only coat the surface of the BC nanofibers but do not fill the interfibrillar pores. It is noted that the cluster sizes of the NPs in the nanocomposites are relatively larger than in the case of the pure NPs. This might be due to the open structure of the BC aerogels which allow the clustering of NPs to occur more readily in the coprecipitation process. The lateral sizes of BC fibrils have previously been reported to be in the range 10-100 nm [43][44][45], which by a consideration of their stochastic geometry yields pore sizes less than 0.1 μm [43]. Within this confined geometry, it would be expected that the NPs would cluster, or given the right size range, coat the surface of the BC fibrils. It is expected that smaller NPs will have a higher surface to volume ratio, and therefore couple to the BC fibrils through increased van der Waals interactions, or interactions between the positive charge of free Fe ions, and the partial negative charge of the -OH groups on the cellulose during co-precipitation. The mechanism for the formation iron oxide NPs in the BC nanofibril network starts from the attraction of Fe 2+ , Fe 3+ ions to the hydroxyl (-OH) group of the BC fibrils. When the alkaline agent (in this case NH 4 OH) was introduced in the system, the partial negative charge on the OH groups interact with the Fe 2+ , Fe 3+  , SO 4 2− ) in the proximity of BC fibrils might inhibit or slow the rate of the conversion due to repulsive forces from the same negative charges between the anions and OH groups. Depending on how strong the electrostatic forces between anions and the corresponding cation, the rate of the chemical reactions to form iron oxide NPs occurs differently by using various starting materials. The different rate in co-precipitation causes the difference in the growth of the NPs and clustering. For instance, for BC-CC, BC-SC, and BC-CN, the rate of their NP formation is likely to be increased compared to other samples. Thus, the NPs appear to be more densely packed in those samples, and the clustering of NPs are also higher. This interpretation is supported by the larger cluster sizes of NPs for BC-CC, BC-SC, and BC-CN (75-86 nm) compared to BC-AN, BC-SN, and BC-AC (55-65 nm). In other words, if the kinetics of the NPs formation in the porous BC networks is high for the particular pairs of starting reactants, it promotes the growth of NPs, resulting in the high density of NPs and the large cluster sizes. The phase and crystal structures of the BC/iron oxide nanocomposites have been determined by XRD analysis, as shown in Fig. 6. The XRD pattern of the pristine BC is also presented. This shows the presence of Bragg peaks located at 14.3°, 16   BC network and attach to the BC nanofibrils. This makes them more difficult to be removed by washing. This point is discussed again in the FTIR section.
The crystallite size of the NPs in BC, calculated by using Scherrer equation, is in the range of 10-16 nm ( Table 2). These sizes are very close to the crystallite sizes for the pure NPs without BC. This indicates that although the NPs are found to cluster, with much larger overall sizes in the BC networks, the sizes of the iron oxide crystals are still relatively unchanged. It is noted however, that due to peak broadening, because of the small size of the NPs, that these figures are subject to some error when determined using the Scherrer equation. As noted, since the XRD patterns of Fe 3 O 4 and Fe 2 O 3 are very similar, the nanocomposites could be either BC/Fe 3 O 4 or BC/Fe 2 O 3 , or the combination of both phases. The presence of mixed phases could potentially influence the magnetization of the nanocomposites, as discussed later. Fig. 7 shows the FTIR spectra of the BC as well as the BC/iron oxide nanocomposites. The FTIR spectra of the Fe 3 O 4 and Fe 2 O 3 references are also included for comparison. Under the measuring range, Fe 3 O 4 and Fe 2 O 3 do not exhibit any distinct features, since the FTIR band for Fe\ \O stretching is between 400 and 600 cm −1 [48,49]. Thus, the presence of the iron oxide NPs should not obstruct the spectra of the BC nanocomposites. The characteristic absorption bands for the pristine BC are observed; namely, the hydroxyl (-OH) group at~3350 cm −1 , C\ \H stretching at~2900 cm −1 , and C-O-C bond stretching at 1060 cm −1 [24,46]. The BC/iron oxide nanocomposite samples show unique changes in the positions of bands representing functional groups. The bands representing -OH groups of all samples were observed to shift to a lower wavenumber position in the range 3020-3350 cm −1 . In co-precipitation, the Fe 2+ and Fe 3+ ions are thought to anchor to the -OH groups of the cellulose, before being converted to iron oxide NPs; this is believed to be the cause of a shift in the position of the band representing -OH groups [19]. The characteristics of the band representing C-O-C bonds in the pristine BC were also modified. Depending on the functional groups of the starting reactants, the characteristic bands for the nanocomposites can be found in the wavelength range 1000-1500 cm −1 . These bands can be attributed to the byproducts of the co-precipitation process from the chemical reactions (1)- (6). For example, the characteristic band of ammonium chloride (NH 4 Cl) at~1390 cm −1 [50] is found in the spectrum of BC-CC, which corresponded well with the XRD secondary phase observed for this sample. Similarly, the bands located at~1425 cm −1 and~1050 cm −1 for BC-SC are assigned to the N\ \H bending and the sulfate group of ammonium sulfate (NH 4 ) 2 SO 4 respectively. The band located at 1300 cm −1 for BC-AC is associated with ammonium acetate (NH 4 C 2 H 3 O 2 ) [51]. The NO 3 asymmetric stretching band located at 1315 cm −1 [52] from ammonium nitrate can be found in the BC-SN sample. The observation of these functional groups indicates that even after thoroughly washing the BC nanocomposites in water several times, these salt by-products were not totally eliminated from the samples. Additionally, the presence of these functional groups could be  another reason for the shift of the -OH groups of the BC since they can form strong hydrogen bonds with these groups. The existence of the salt byproducts in the BC nanocomposites could play an important role in the formation of iron oxide NPs in the BC network. As explained earlier, the difference in the morphology and sizes of NPs in BC is due to the NP formation mechanism which involves the presence of various anions according to each chemical reaction. These anions may impede the reaction between the partial charge on the OH groups and the iron ions. Alternatively, they may react with NH 4 + ions and form salt byproducts coated on the surface of BC fibrils, as evidenced from FTIR. These byproducts may also hinder the formation of iron oxide NPs in the BC network. Since the byproducts of each reaction are different, depending on the starting reactants, their effects on the formation of NPs are not the same. Such effects would result in the difference in the morphology, cluster sizes, and distribution of NPs in the BC nanocomposite, as shown in Fig. 5. This in turn may also influence the magnetic properties of the nanocomposites. The magnetic M-H curves of the nanocomposites are presented in Fig. 8. The characteristic curves of the superparamagnetism of the iron oxide NPs still persist. However, there are variations in the magnetizations of the BC/Fe 3 O 4 nanocomposites prepared from different reactants, as tabulated in Table 2. The M s values for each sample are ranked in the following order BC-SC (~57 emu/g) > (BC-AN, BC-CC) > (BC-CN, BC-SN) > BC-AC (~26 emu/g). This suggests that the Fe (II) sulfate and Fe(III) chloride salts are the best candidates for forming BC/Fe 3 O 4 nanocomposites if the magnetization is the most important requirement. This particular nanocomposite could be potentially utilized in several industrial applications e.g. flexible sensors, actuators, or electromagnetic shielding. In contrast, although the BC-AC nanocomposite exhibits a lower M s , the SEM image (Fig. 5) clearly shows a more open structure, with high porosity and surface area of the nanofibers. These features make this nanocomposite an excellent candidate for adsorption or wound healing applications. Thus, by adjusting the starting   reactants for the co-precipitation of Fe 3 O 4 in BC structure, the characteristics and properties of the BC nanocomposites can be controlled towards an application.
One of the reasons for the change in M s could be attributed to the content of the magnetic phases in the nanocomposite. In general, the higher the content of the magnetic phase, the larger the magnetization.
To determine the content of the magnetic phases, the BC and the nanocomposite samples were subjected to TGA analysis. As shown in Fig. 9, BC, without the presence of NPs, has a prominent weight loss in the temperature range 250-350°C. This loss is typically associated with the degradation of BC, including decomposition, dehydration, and depolymerization of the glycosidic units [24,53]. Above 600°C, the residual weight of the sample is stable. For the BC/iron oxide nanocomposite samples, weight losses initiate at around 200°C. The weight loss profiles are different, depending on the samples. For example, BC-AC and BC-CN exhibited significant weight losses of 55-60% at a temperature of 250°C, whereas the weights of other samples gradually decrease up to temperatures in the range 600-700°C. The difference in the weight loss profiles is believed to be because of the retained functional groups in each nanocomposite, as shown in Fig. 7. To determine the content of the magnetic Fe 3 O 4 phase, the weights of the pristine BC at 600°C were subtracted from the weight of the nanocomposites at the same temperature. At this temperature, the retained functional groups are decomposed, and the remainder is the iron oxide residue. The weight percentages of the iron oxide NPs are presented in Table 2. From this analysis, it is noted that the largest value of M s for BC-SC correlates with the highest iron oxide NPs content. Conversely, BC-AC has the lowest The mean size of the magnetic NPs also has an influence on the M s value. It was found that the smaller NPs exhibit lower M s due to a surface disordering effect [54]. The surface of the NPs is composed of some canted or disordered spins that prevent the core spins from aligning along the magnetic field direction [55]. Since the smaller NPs consist of a higher fraction of atoms at the surface, the M s values of these NPs is proportionally lower. In the present work, from the XRD analysis, iron oxide NPs with small crystallite sizes (10-13 nm) are observed for BC-AC and BC-SN, which are relatively smaller than for other samples. These observations correlate well with the magnetic measurements, with the lowest M s values being recorded for these two samples. Another factor that could contribute to magnetism is the crystal structure. Only the crystalline phase of magnetite or maghemite can contribute to ferrimagnetism. A secondary phase or amorphous structure cannot enhance the magnetization. From Fig. 6, the XRD pattern of BC-CC shows a prominent peak for the secondary phase (located at 32.4°), which is likely to be non-magnetic, thereby suppressing the magnetization of the nanocomposite. This is probably the reason that the BC-CC sample exhibits a not so large M s despite a relatively substantial amount of the iron oxide NPs in the BC structure. Furthermore, the secondary phase (at 2θ = 44.5°). is also observed for the BC-AC sample. Its presence reduces the magnetization, and thus contributes to the lowest M s value of this sample.
Finally, the form of magnetic phases of NPs could lead to a variation in the magnetization. As mentioned earlier, the XRD patterns of magnetite (Fe 3 O 4 ) and maghemite (Fe 2 O 3 ) are very similar, and it is challenging to distinguish between these two phases from the XRD results. The BC nanocomposites likely contain both phases in different proportions. However, these two phases exhibit different M s values. The bulk M s value for Fe 3 O 4 (92 emu/g) [56] is significantly higher than that of Fe 2 O 3 (76 emu/g) [57], which means that the nanocomposites with a higher fraction of Fe 3 O 4 are likely to possess higher magnetization. To determine the fraction of Fe 3 O 4 and Fe 2 O 3 , the XAS analysis for Fe Kedge of the BC nanocomposite samples were carried out. The normalized XAS data were processed and analyzed after background subtraction in the pre-edge and post-edge region using the ATHENA software, which is included in the IFEFFIT package [58]. The XAS spectra of the BC nanocomposites, as well as the Fe 3 O 4 and Fe 2 O 3 references, are shown in Fig. 10. Since the absorption edge and peaks, and the characteristic features of Fe 3 O 4 and Fe 2 O 3 are different, the linear combination fit method in the ATHENA software was used to quantify the fraction of each phase, as represented in Table 2. As expected, all the nanocomposite samples comprise both magnetic oxide phases. In every sample, the  proportion of the Fe 2 O 3 phase is relatively high with respect to the total iron oxide content. The presence of the high proportion of Fe 2 O 3 could be due to the fabrication process under ambient conditions. The Fe 2+ ions were possibly oxidized to the Fe 3+ ions during the coprecipitation process. Furthermore, the Fe 3 O 4 NPs inside the BC structure could be partially oxidized after rinsing and drying the BC nanocomposite hydrogels under ambient conditions. Comparison among all samples, the fraction of Fe 3 O 4 in BC-CC is noticeably lower than the other samples. It implies that using Fe(II) and Fe(III) chloride salts as starting reactants is likely to promote the formation of the Fe 2 O 3 phase. The possible reason is that, according to the chemical reactions, only the Cl − anion was created when using Fe(II) and Fe(III) chloride salts (Eq. (1)), whereas two types of anion were generated for the other reactions (Eq. (2)- (6)). This might be associated with the ease of oxidation of Fe 2+ , having just one type of anion, and thus contributing to the conversion of Fe 3 O 4 to Fe 2 O 3 in the case of BC-CC. It is also a part of the reason for the reduced M s of BC-CC due to the presence of a large proportion of the Fe 2 O 3 phase. On the other hand, BC-SC comprises a relatively large fraction of Fe 3 O 4 (0.509) which correlates with a high M s value for this sample.

Conclusions
A systematic study on the synthesis of BC/iron oxide nanocomposites by a co-precipitation method has been presented. The effect of starting reactants on the physical properties of magnetic nanoparticles formed inside bacterial cellulose networks was reported. For these reactants, chloride, sulfate, and acetate salts were used as the source of Fe 2+ ions, whereas chloride and nitrate salts were used as the source of Fe 3+ ions. For the fabrication of iron oxide NPs, different starting reactants do not seem to significantly affect the morphology, size, phase, and magnetic properties of the synthesized NPs when formed outside of the bacterial cellulose environment. However, this does have an impact on the physical characteristics of the BC/iron oxide nanocomposites when formed in situ. Using different starting reactants leads to differences in the sizes of the NPs formed inside the BC structure, and various functional groups of the retained by-products. The main phases of Fe 3 O 4 and Fe 2 O 3 were found in all nanocomposite samples, but the presence of secondary phases was also noted in some samples. Moreover, the loading amounts of the incorporated NPs were found to vary as the starting reactants were changed. Additionally, using different Fe(II) and Fe(III) salts as the starting reactants also yields a different ratio of the magnetite (Fe 3 O 4 ) and maghemite (Fe 2 O 3 ) phases. These ratios have significant influence on the measured magnetic properties. In summary, the BC/iron oxide nanocomposite synthesized from Fe(II) sulfate and Fe(III) chloride, BC-SC, yielded the most abundant amounts of NPs, pure phases of these materials, large particle sizes, and the highest Fe 3 O 4 /Fe 2 O 3 ratio. All these factors are thought to result in the highest magnetization value for the BC-SC nanocomposites compared to other samples. On the other hand, the BC-AC sample exhibited the lowest magnetization because of the low magnetic oxide content, small particle size, the existence of secondary phases, and a low proportion of the Fe 3 O 4 phase. The present study thus shows that the choices of the starting reactants have a direct implication on the design of the magnetic BC systems. If the magnetization is the most critical factor for an application, such as flexible sensors, actuators, or electromagnetic shielding, it is suggested that Fe(II) sulfate and Fe(III) chloride should be used as the starting reactants. On the other hand, though exhibiting lower magnetization, the BC-AC nanocomposite preserves a porous structure and a high surface area of the nanofibers, which makes it more suitable for applications such as adsorbents or wound healing.

Data availability statement
The raw/processed data required to reproduce these findings cannot be shared at this time as the data also forms part of an ongoing study.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.