1. Introduction
In recent years, silica nanofibers with a mesoporous structure have attracted considerable attention due to their high surface area, high pore volume and large pore size [
1,
2,
3]. These physical properties are essential for many applications. For example, porous nanofibers comprised of different inorganic oxides proved to be effective adsorbents in the removal of toxic pollutants in wastewater presumably by facilitating the diffusion of target molecules to active sites exposed throughout the porous structure [
4,
5,
6,
7,
8,
9].
There are several techniques for the production of silica nanofibers, such as electrospinning (ES) [
1,
2] centrifugal jet spinning (CJS) [
10], and solution blow spinning (SBS) [
11,
12]. In general, most spinning techniques used to synthesize silica nanofibers involve a sol-gel procedure that is intimately dependent on the precursor solution/sol properties and the polymeric aids used. Lately, SBS has been highlighted as a simple, affordable, versatile, high efficiency method of nanofiber production [
13,
14]. It is also preferred for its ability to produce nanofibers from a wide variety of polymers and solvents. SBS uses pressurized air to make fibers with a controllable and flexible morphology including highly porous nonwoven fiber mats formed by cross-layering nanofibers using a rotating collector. Nonwoven fiber mats can be useful for several applications due to the high accessibility of the pore network of the individual nanofibers and the ease of collecting and handling of fiber mats [
14,
15].
Typically, the mesoporous structure of the fibers can be obtained by using a template or surfactant as structure directing agent. The most common surfactants include pluronic P123 [
1,
7,
16] and cetyltrimethylammonium bromide (CTAB) [
2,
17,
18]. They are efficient in producing mesopore structures in several oxide nanofibers. However, their use makes the process complex and more expensive, as it is still necessary to add a spinning aid (organic polymer) to achieve a spinnable solution.
Spinning aids are essential in achieving the molecule chain entanglement required to improve the viscosity of the precursor solution and maintain a continuous jet that will reach the collector. Among the polymeric spinning aids, polyvinylpyrrolidone (PVP) is by far the most widely used for silica nanofibers production owing to its excellent chemical compatibility and stability [
3,
19,
20]. Moreover, PVP has been used in the production of mesoporous polymer films [
21] and carbon nanofibers [
22] due to its amphiphilic character. Thus, due to its bifunctional properties, PVP can make the preparation and production of mesoporous silica nanofibers less complex and costly when used as a spinning aid polymer.
Studies indicate that polymer characteristics influence the physical properties of polymeric nanofibers produced by SBS [
13,
23]. However, there is no study on the influence of polymeric additives on the physical properties and/or dye adsorption capacity of blow spun ceramic nanofibers. Thus, the novelty of this investigation is to analyze the influence of polymers characteristics on the physical properties and dye adsorption capacity of silica fibers obtained by SBS.
To date, studies have investigated the effect of varying the amount of polymer in polymer solutions on the morphology and properties of spun polymeric fibers [
24]. Molecular weight is known to affect the properties of polymer solutions as well as the properties of the nanofibers they produce. However, there is a lack of information on the effect of the molecular weight of PVP on the properties and morphology of oxide ceramic nanofibers. There are two common grades of PVP products for making silica nanofibers by electrospinning (M
w = 360,000 and 1,300,000). One objective of this study was to further improve the physical properties of silica nanofibers using a polymer (PVP) rather than surfactants. A further objective of this study was to compare the properties of porous silica nanofiber produced by SBS from the two common grades of PVP and without the use of surfactants. The fiber properties compared included fiber morphology, physicochemical properties, specific surface area, pore structure, and dye (methylene blue) adsorption behavior.
3. Results
Silica nanofibers (SNFs) were successfully prepared via the SBS technique with the aid of PVP as a porogenic agent and spinning aid.
Figure 1a–d shows the morphologies of the as-spun and calcined nanofibers at different PVP molecular weight (360,000 and 1,300,000). SEM images show a structure comprised of randomly oriented nanofibers with a cylindrical and thin, elongated shape. The mean diameter of SNFs/PVP-LMw and SNFs/PVP-HMw was 432 ± 263 nm (
Figure 1a inset) and 549 ± 356 nm (
Figure 1c inset), respectively.
Figure 1b,d show that the fibrous structure of the samples were well preserved after the calcination process. The mean diameters of the fibers were slightly affected by the differences the molecular weight of the PVP. Thus, the mean diameter of SNFs/PVP-LMw is lower than SNFs/PVP-HMw (
t test,
p < 0.001). There is a significant shrinkage in fiber diameter (
t test,
p < 0.001) due thermal degradation/removal of the organic polymer component and a consolidation/densification process for both samples. The fiber diameters were reduced to 164 ± 88 nm for SNFs-LMw (
Figure 1b inset) and to 216 ± 106 nm for SNFs-HMw (
Figure 1d inset). The mean diameter was lower in SNFs-LMw relative to the SNF-HMw fibers (
t test
, p < 0.001). However, differences in the mean fiber diameter between SNFs/PVP-LMw and SNFs/PVP-HMw were not significantly different (Mann-Whitney test,
p = 0.0918), while differences in the range of fiber diameter s of calcinated samples are different (Mann-Whitney test,
p < 0.001).
Figure 2 shows the thermal behaviors of pure PVP-LMw (M
w = 360,000), PVP-HMw (M
w = 1,300,000), and as-spun nanofibers (SNFs/PVP-LMw and SNFs/PVP-HMw). For pure PVP, the TGA curves exhibit two stages of mass loss. An initial stage with the maximal mass loss occurred approximately at 100 °C, which can be attributed to the evaporation of physiosorbed water. In this stage, mass loss is greater for PVP-LMw (~17%) than for PVP-HMw (~14%). The lower physiosorbed moisture in PVP-LMw can be attributed to its lower molecular weight and concomitant higher surface area on which to absorb water. Moreover, the thermal degradation of PVP chains (observed in the second stage) was lower for PVP-LMw (from 310 °C to 440 °C) than PVP-HMw (from 340 °C to 465 °C). This is also a function of molecular weight and chain entanglement since the greater entanglement of the PVP-HMw molecular chains can make its decomposition more difficult. In this stage of thermal degradation, C–H, C–C, C = O, and C–N bonds are broken and degraded into carbon oxide and nitrogen oxide compounds [
27].
The size of the polymer chains also has a significant importance in the TGA curves of as-spun nanofibers. The curves revealed two stages of mass loss with the first occurring near 100 °C and was interpreted as a loss of physiosorbed water and residual free ethanol from the condensation of hydrolyzed TEOS to silica [
27]. For SNFs/PVP-LMw, the evaporation phase of free liquid started at a lower temperature than that for SNFs/PVP-HMw. This was presumably due to the lower degree of chain entanglement which allowed the solvent molecules a less tortuous path for evaporation.
The second stage of mass loss occurred near 320 °C where the PVP in the hybrid fibers exhibited lower thermal stability than pure PVP. According to Bogatyrev et al. [
28], the interaction of PVP and silica occurs only through the hydrogen bonds between the carbonyl (C = O) group in PVP and the silanol (SiO-H) group on the silica surface (
Scheme 1A). This interaction can weaken the pyrrolidone ring bonds and their degradation can occur at lower temperatures. A greater mass loss during the second stage occurred between 300 °C and 435 °C, where carbon dioxide and pyrrolidone rings were the main components released. Over this temperature range, a continual loss of mass was observed for both SNFs/PVPs, which was believed to involve the thermal degradation of PVP chains closest to the surface of the silica phase [
28,
29,
30] and the release of water in the silica framework by the condensation of silanols groups [
31]. Mass loss was found to be slightly greater for SNFs/PVP-LMw (70%) than SNFs/PVP-HMw (64%) up to 550 °C.
This difference suggests that high M
w chains favor a higher packing and association with the silica network compared to low M
w PVP where more intimate carbonyl-silanol interactions may occur (
Scheme 1A). It may be that the depolymerization of PVP at high temperatures may be slowed by the number of polymer units bound with silanol groups on the silica surface [
28]. Hence, the residual mass is composed of the silica phase with small amounts of carbon. In contrast, low molecular weight PVP chains may have a greater intimate contact with ≡ SiOH groups to reach the oligomeric forms. Then, the condensation rate of the silanol groups was more pronounced, which released a greater amount of water and increased the size of the silica particles (
Scheme 1B).
Table 1 shows the carbon (C), hydrogen (H), and nitrogen (N) content of the samples which verified the presence of polymer residues in the silica nanofibers after calcination. The results confirm that both samples have similar N content. However, the amount of C was three times greater for SNFs-HMw than for SNFs-LMw.
FT-IR analysis of the SNFs-LMw and SNFs-HMw are shown in
Figure 3a. The spectra of both samples show characteristic IR bands of amorphous silica. The broad band between 3100 and 3700 cm
−1, centered at 3460 cm
−1 is attributed to O-H stretching frequency related to physiosorbed water and silanol groups on silica surface [
32]. The symmetric and antisymmetric Si-O-Si (siloxane) stretching are observed at 1080 and 470 cm
−1, respectively. Stretching vibration at 970 cm
−1 and bending frequency at 799 cm
−1 correspond to the surface silanol groups Si-O(H) [
33,
34,
35,
36]. However, the spectra for SNFs-HMw clearly showed peaks at 2913 and 2845 cm
−1 which was attributed to CH
2 absorption of PVP which confirmed that the existence of residual carbon content was more significant in this sample. The XRD analysis (
Figure 3b) indicated the presence of amorphous silica for both calcinated samples, with only a small broad band in the 2-theta range of 15–30° [
19,
20].
N
2 adsorption-desorption isotherms for SNFs-LMw shown in
Figure 4a are identified as type IV, indicating a mesoporous material according to IUPAC classification. In the low relative pressure region (
p/
p0 < ~0.1) the uptake of nitrogen can be observed, which indicates the presence of micropores. Mesoporous structures are also observed when the relative pressure (
p/
p0) is more than ~0.5 which is associated with capillary condensation. This is where the nitrogen uptake increases and starts the typical H4 hysteresis loop [
37] and indicates the presence of narrow slit pores [
38,
39]. On the other hand, plotted isotherms of SNFs-HMw can be classified as composites of type I + IV isotherms with a discreet hysteresis loop type H4 (See
Figure 4b). Isotherms type I are typical of microporous materials [
40], which have high surface areas.
The physical parameters of the samples are summarized in
Table 2. BET analysis indicated the largest surface area (921.7 m
2 g
−1) for SNFs-HMw while in SNFs-LMw samples had a lower surface area in comparison (635.6 m
2 g
−1), which was indicative of mostly micropores. This result was confirmed by the
t-plot results in which the micropore volume and micropore area were much high for SNFs-HMw than for SNFs-LMw. In contrast, BJH desorption cumulative volume of pore width between 1.7 nm and 300 nm indicates a high pore volume (0.74 cm
3 g
−1) and desorption average pore width of 3.67 nm for SNFs-LMw. For SNFs-HMw those values are 0.12 cm
3 g
−1 and 1.99 nm, respectively (
Table 2).
The results of physical characterization indicate that both samples contained micro/mesopores. However, SNFs-LMw was predominantly comprised of mesoporous pores and SNFs-HMw was predominantly microporous in nature. This can be explained by the different sizes of the polymer layers, which impacted the PVP-silanol and silanol-silanol interactions. As mentioned above, the lower M
w PVP favored a greater intimate contact between the silanol groups. This contributed to the formation of nuclei that favored the growth of silica particles by the hydrolysis/condensation step, resulting in large particles. Therefore, large particles favored the formation of larger pores (
Scheme 1B).
Based on isotherm data (the data represent the mean of three experiments and the error bars represent the standard deviations of triplicate tests (
Figure 5a,b), the adsorption yield onto both adsorbents increased as initial MB concentration increased. This can be attributed to the high number of dye molecules colliding with the adsorbent, which provided a driving force that overcame the mass transfer resistance between them. On the other hand, the adsorption capacity decreased as the temperature increased, indicating an exothermic process. The influence of the physical properties of SNFs-LMw and SNFs-HMw on the adsorption capacity of MB was clearly demonstrated. Despite the higher specific surface area of SNFs-HMw, SNFs-LMw had higher adsorption capacity. As shown in
Figure 5a,b, the maximum adsorption capacity of SNFs-LMw and SNFs-HMw at 298 K is 278.86 mg g
−1 and 123.29 mg g
−1, respectively. These differences can be attributed to the properties of the pores. The predominant mesoporous structure of SNFs-LMw may have facilitated the transport of the MB molecule into the porous fiber structure where more sites for adsorption could be accessed. In contrast, the molecular dimensions of MB molecule (1.70 nm × 0.76 nm × 0.33 nm) [
41] is close to the average pore diameter of SNFs-HMw (1.99 nm), which may have made MB access to binding sites more difficult (resulting in less uptake).
Table 3 shows the results obtained from the adsorption parameters derived according to the linear forms of the two models (Equations (3) and (4)). The statistical parameters were also explored to identify the fit quality of each model. The best fit model was chosen based on the highest correlation coefficients (
R2) and the lowest
SD and
RMSE values. Comparing the results, the adsorption of MB on both adsorbents was best fit by the Langmuir isotherm.
Figure 6a,b show the experimental and Langmuir linear fit (best fit model) obtained when Ce/Qe was plotted against Ce at 298 K, 308 K and 318 K for SNFs-LMw and SNFs-HMw, respectively.
Qmax(calc) and
KL were evaluated from the slope and intercept. The
Qmax(calc) of SNFs-LMw and SNFs-HMw is 275.5 mg g
−1 and 121.9 mg g
−1, which are very close to the experimental values (
Qe). These results means that a monolayer coating of MB molecules was formed on the energetically homogeneous surface of the nanofibers. Therefore, the differences in the physical properties of the samples did not change the adsorption behavior of the dyes.
Zeta potential values of the adsorbents were determined varying the pH from 3 to 10 in order to verify the surface charge. Moreover, silanolate groups (SiO
−) were expected to be the prevalent silica form in a large pH range due to the low isoelectric point of the silica surface (IEP ≈ 2–3) [
41]. As shown in
Figure 7a, for both adsorbents the surface is negatively charged for the range of all pH conditions tested. SNFs-HMw becomes more negative as the pH increases, and no significant change is observed for SNFs-LMw between pH 4 and 10. Thus, SNFs-LMw exhibited higher density of SiO
− groups even at low pH conditions. Although SNFs-HMw has a significant carbon content that can modify the surface charge, there was no charge difference between HMw and LMw fibers at pH greater than 8. The high content of hydroxyl groups on the surface of the silica fibers surface develops negative charges in aqueous solutions, notably at high OH
− concentrations. This means that the degree of deprotonation of hydroxyl groups increases at a high pH [
33].
Under high pH conditions, the silanol groups are deprotonated and become negatively charged. Hence, ≡Si-O
− can attract and bind the positively charged MB
+ molecules. However, SNFs-HMw shows a low adsorption capacity in comparison to SNFs-LMw under all pH conditions (
Figure 7b). Despite data showing that SNFs-HMw has a higher BET surface area and a zeta potential at pH 10 similar to the SNFs-LMw, the adsorption capacity of MB onto SNFs-LMw is three times higher than that of SNFs-HMw at this pH. This means that the amount of silanolate groups and total surface area are not the main factors that determine the adsorption capacity. Since SNFs-LMw have a much higher mesopore volume than SNFs-HMw, it is possible that mesopores are an important factor in governing the adsorption of MB dye. Indeed, MB binding may be strongly influenced by the combination of the physical characteristics of the SNFs and the high density of the SiOH groups.
Figure 8 shows FTIR analysis of SNFs-LMw (
Figure 8a) and SNFs-HMw (
Figure 8b) after adsorption of MB (blue line) in comparison with the adsorbents before removal (black line). For SNFs-LMw-MB and SNFs-HMw-MB, the peak at 1642 cm
−1 corresponds to Si-O(H) shifted to 1610 cm
−1. In addition, the -O(H) vibration peak centered at 3460 cm
−1 also shifted and corresponds to 3432 cm
−1 and 3422 cm
−1 for SNFs-LMw and SNFs-HMw, respectively. These results may be attributed to the interaction between MB molecules and the adsorbents, through hydrogen bonds [
42]. Furthermore, the new small peaks attributed to the stretching vibration of -C-N, -C-H and C = S in the MB molecule are shown at 1334 cm
−1, 1395 cm
−1, and 1492 cm
−1, respectively [
43,
44].
The adsorption of MB on both SNFs was evaluated at different time intervals in the range of 15 to 480 min (
Figure 9). As can be seen, approximately 65% of the MB adsorbed occurred in the first 15 min for SNFs-LMw. The rapid uptake could be due to an initial adsorption occurring on the external surface and on the larger mesopores of the adsorbent. For SNFs-HMw which have only a small fraction of mesopores, only 37% of MB was adsorbed at the first 15 min. Consequently, adsorption was relatively slow as the dye molecules diffused into the smaller pores. The adsorption equilibrium was achieved after 360 min for SNFs-LMw, which reached ~96% of MB removal. Adsorption equilibrium for SNFs-HMw occurred after 240 min with only ~65% of MB removal.
The kinetic parameters were determined by adjusting the kinetic data of the MB adsorption in the models of pseudo-first-order, pseudo-second-order and intra-particle diffusion. The results are summarized in
Table 4. Intra-particle-diffusion occurred during the adsorption process due to porosity and high surface area. However, the kinetic constant of
Kdif, and the relatively low
R2 value confirmed that Intra-particle-diffusion is not the sole rate-limiting step for the adsorption process [
18]. For pseudo-first-order and pseudo-second-order models, the values of the experimental adsorption capacity (
Qe(exp)) were closer to the calculated adsorption capacity (
Qe(cal)) by pseudo-second-order models for both adsorbents. The results confirm second-order adsorption kinetics best fit the adsorption behavior. This is in agreement with the high
R2 values and the lower values of SD and RMSE and indicate that the chemisorption process is the main rate controlling step for MB dye adsorption for both SNFs.
Table 5 shows energy (Δ
G⁰), entropy (Δ
S⁰) and enthalpy change (Δ
H⁰) estimated for both MB adsorption onto SNFs-LMw and SNFs-HMw at 298 K, 308 K and 318 K. Results show the influence of physical properties of the fibers on the thermodynamic parameters. Negative values and positive values of Δ
H⁰ indicating that the adsorption process is exothermic and endothermic for SNFs-LMw and for SNFs-HMw, respectively. This is in agreement with the results in adsorption isotherms for SNFs-LMw, in which the adsorption capacity decreased with increasing temperature. For SNFs-HMw, the adsorption capacity with increasing temperature from 308 K to 318 K may not be considerable. The positive values of ΔS
⁰ reflect the increase in randomness at the solid/liquid interface during adsorption for both SNFs. The more negative values of ΔG
⁰ for SNFs-LMw indicate an adsorption more spontaneous and energetically favorable than that for SNFs-HMw.
Table 6 shows the comparison with the results of previous studies on MB adsorption capacities available in the literature [
8,
16,
45,
46,
47,
48,
49]. Most of them are mesoporous materials with high surface area. This is a demonstration of the efficiency of mesoporous materials for application as dye adsorbents. Therefore, the SNFs-LMw adsorption capacity is higher than those observed in literature. This indicates that the improved adsorption capacity is attributed to the superior physical properties of the sample.