3. Results and discussion
3.1. Fabrication and characterization of L-CNM
As schematically shown in Scheme 1, the L-CNM with a sheet-like
structure was one-step in situ constructed by employing a facile elec trospinning method. The spinning dope was prepared by mixing LaCl3
with a cellulose solution, while a NaOH aqueous solution was used as
the coagulating bath. The sheet-like La(OH)3 was decorated on CNM
through hydrogen bonding interaction and physical wrapping. Owing
to the synergy of the abundant hydroxyl groups and the special elec tron-deficient orbital structure of La(OH)3, the L-CNM exhibited su perhydrophilicity and high affinity toward CR having two amine groups
and electron-rich sulfonic groups, which could be of great advantages in
oil/water separation and selective removal of dyes.
The micro morphologies of pure CNM and L-CNM were observed
with SEM. As showed in Fig. 1a, the neat electrospun cellulose nano fibers had an average diameter of ~360 nm (see the fiber diameter
distribution in Fig. S1) and the interpenetrated nanofibers formed nu merous macropores, which ensured a high water flux of the membrane.
Fig. 1b displayed the SEM image of L-CNM surface, where quite a lot of
nanosheets with a uniform square shape were clearly visible. These
nanosheets with a thickness of ~400 nm and a width of 1–9 μm were
evenly dispersed in the cellulose nanofiber scaffold without any large
aggregates. Meanwhile, most of the pores were still preserved. More over, the average fiber diameter of L-CNM decreased to ~280 nm (Fig.
S1). This was possibly attributed to the increased conductivity of the
spinning dope after the addition of LaCl3 [37]. From the high-resolution
SEM image in Fig. 1c, the nanosheets were wrapped by the cellulose
nanofibers and intimately attached to them with vague phase bound aries, which may benefit the immobilization of those nanosheets in the
membranes. Besides, the surface roughness (Ra) of CNM and L-CNM
were measured using CLSM. As shown in Fig. S2, the Ra of pristine CNM
was 0.065 µm. After being modified with La(OH)3, the Ra apparently
increased to 0.306 µm. And this improved roughness could benefit the
wetting behavior of L-CNM according to the Wenzel model [38].
The surface chemical composition of the membrane was estimated
by energy-dispersive spectrometry (EDS). The EDS results of pure CNM
and L-CNM were presented in Fig. 1g and h, respectively. The char acteristic peaks assigned to La element were observable in the EDS
spectrum of L-CNM, but hardly could be detected for pure CNM.
Moreover, the EDS mapping images revealed the C, O and La elements
were distributed homogeneously in the whole material matrix
(Fig. 1d–f), indicating the uniform dispersion of as-synthesized La(OH)3
nanosheets in the membrane.
The chemical composition of the membrane was further in vestigated by ATR-FTIR and XRD. Fig. 2a gave the ATR-FTIR spectra of
CNM and L-CNM. The bands at 1437 cm-1 and 860 cm-1 in both samples
were assigned to symmetric CH2 bending and CH2 stretching vibration,
respectively [39]. A new peak appeared at 674 cm-1 in the L-CNM’s
spectrum, which was attributed to the bend vibration of La-O-H [40].
Fig. 1. Structure characterization of the membranes: SEM images of (a) pristine CNM and (b,c) L-CNM at different magnifications. (d–f) EDS mapping images of L CNM. EDS analysis of (g) original CNM and (h) L-CNM surface.
From the XRD diffraction pattern of CNM (Fig. 2b), the cellulose II
diffraction peaks at around 20.2° and 12.6° were observed owing to
reflection from the (1 1 0) and (1–10) crystal planes of cellulose II [41].
The diffraction peaks at 27.3° and 43.7° for L-CNM corresponded to the
(1 0 1) and (2 1 0) crystal planes of La(OH)3, respectively [40,42]. These
results consistently suggested the successful synthesis of La(OH)3 with
the hexagonal phase [43].
The porosity features of CNM and L-CNM were investigated by N2
adsorption-desorption isotherms. As depicted in Fig. 3a, these isotherms
displayed the type-IV isotherm shape with a typical H2 hysteresis loop,
indicating the membranes possessed an interconnected pore structure
with both micro- and meso-pores [44]. Additionally, the high relative
pressure part (P/Po0.9) of the N2 adsorption isotherms increased
significantly, implying the existence of macro-pores in the membranes
[45]. Fig. 3b showed the distribution of incremental pore volume with
the pore width. It further demonstrated that both CNM and L-CNM had
multiscale pores. And the average pore size of L-CNM (20.2 nm) was
larger than that of CNM (10.1 nm). Cumulative pore volume versus
pore size was given in Fig. 3c. A remarkable increase in the total pore
volume of L-CNM (0.134 mL/g) by more than seven times that of CNM
(0.019 mL/g) was recorded. In addition, the BET specific surface area of
L-CNM (21.15 m2
/g) was much larger than that of CNM (5.96 m2
/g).
The above results confirmed that the total pore volume, pore width and
specific surface area of L-CNM increased sharply as compared with
CNM, providing more water transfer channels for a high water flux.
3.2. Possible formation mechanism of La(OH)3 nanosheets
To investigate the formation mechanism of the La(OH)3 nanosheets,
the time-dependent growth of La(OH)3 in the membrane was mon itored. When the LaCl3-containing electrospun cellulose nanofibers
were immersed in aqueous NaOH for a short time (10 min), a large number of small and irregularly shaped La(OH)3 nanoparticles were
observed at the surface of cellulose nanofibers (Fig. 4a), as a result of
the reaction between La3+ and OH−. When the immersion time ex tended to 90 min, the La(OH)3 nanoparticles agglomerated to form thin
nanosheets (Fig. 4b). The nanosheet formation could be presumably
explained by the Ostwald ripening mechanism [46]. In brief, due to the
intrinsic anisotropic property of La(OH)3, the La(OH)3 nanoparticles
prefer to grow into 2D nanosheets with the exposure of low-energy
facets to reduce the surface energy [46,47]. When increasing the im mersion time to 6 h, more and more regular La(OH)3 nanosheets were
formed owing to the mass diffusion and Ostwald ripening (Fig. 4c).
Moreover, these La(OH)3 nanosheets were uniformly distributed either
on the surface of cellulose nanofibers or at the spacing between ad jacent nanofibers. The porous electrospun cellulose nanofiber scaffold
offered a unique 3D space for the in situ grow of La(OH)3 nanosheets
thereof. And the hydrogen bonding interaction between the hydroxyls
of La(OH)3 nanosheets and cellulose nanofibers, as well as the physical
wrapping should be mainly responsible for the solid anchoring of La
(OH)3 in the membrane.
3.3. Wettability of L-CHM
The surface wettability was crucial for membranes in water treat ment. To evaluate the wettability of L-CNM, the dynamic wetting be haviors of a water droplet on the membrane in air and a heavy oil (1, 2-
dichloroethane as a representative oil) on the membrane underwater
were recorded. As presented in Fig. 5a, a 4 μL water droplet fleetly
spread on L-CNM within 160 ms and gave a water contact angle of near
0° (Movie S1), indicating that L-CNM was superhydrophilic. Whereas
400 ms was needed for CNM (Fig. S3a), showing a faster water
spreading rate of L-CNM. When a 4 μL 1, 2-dichloroethane was placed
onto the membrane underwater, the oil droplet still retained a spherical
shape with a high value of oil contact angle (> 150°), showing un derwater superoleophobicity. Furthermore, even if the oil droplet was
forced against the membrane until obvious deformation, it could detach
from the surface of L-CNM easily (Fig. 5b and Movie S2), indicative of
the ultralow oil adhesion. By contrast, an oil droplet was found to at tach directly to CNM (Fig. S3b). We believe the super-wetting behavior of L-CNM should be attributed to the synergy effect of the abundant Fig. 5. (a) Photographs of a water droplet
spreading on L-CNM in air. (b) Photographs of
dynamic underwater oil-repelling on L-CNM.
(c) Anti-oil-fouling properties of L-CNM and
original CNM. Crude oil was used in this
demo. Digital image in the red frame pre sented the oil fouling of CNM. (For inter pretation of the references to colour in this
figure legend, the reader is referred to the web
version of this article.)
hydroxyl groups from La(OH)3 nanosheets and cellulose nanofibers as
well as the enhanced roughness from La(OH)3 nanosheets (Fig. S2). The
anti-oil-fouling properties of L-CNM and CNM against crude oil were
further evaluated. Fig. 5c demonstrated the whole procedure: the
membrane was pre-wetted with water and then immersed in crude oil
totally. After lifted out of crude oil, the membrane was fully enshrouded
by the crude oil. Finally, this membrane was placed into a beaker
containing clean water. When L-CNM covered by crude oil was im mersed into water, all the crude oil was spontaneously levitated off the
membrane and no trace was left. Whereas obvious oil-fouling was seen
on CNM. These results further manifested the outstanding underwater
anti-oil-fouling property of L-CNM even for the crude oil with a high
viscosity. To sum up, the L-CNM exhibited remarkable selective wett ability and outstanding anti-oil-fouling properties, which was favorable
for oily wastewater treatment [7].
3.4. Oil/water separation performance
The L-CNM had excellent capability in separation of oil/water
mixtures. As demonstrated in Fig. 6a, the hexane/water mixture (v:v/
1:1) was poured onto the membrane, which was pre-wetted by water.
Only driven by gravity, water could rapidly permeate through the
membrane owing to the superhydrophilicity, whereas hexane was re pelled above the membrane because of the underwater super oleophobicity. No visible oil residue was found in the filtered water,
suggesting the high separation efficiency of the membrane. In addition,
the water flux was as high as 5897.7 L m-2 h-1 for L-CNM, almost two
times that of CNM at a similar thickness. Such a high water flux could
be attributed to the faster water wetting rate and more water transfer
channels available in L-CNM as compared with CNM (Fig. S3 and
Fig. 3). In order to imitate the real separation applications, different
types of oil were tested, including cyclohexane, petroleum ether, liquid
paraffin, pump oil and crude oil. From Fig. 6b, all these oil/water
mixtures could be effectively separated with high fluxes (from 6318.9
to 8586.1 L m-2 h-1). Moreover, the separation efficiencies were well
above 98.8% (hexane 98.8%, cyclohexane 98.9%, toluene 99.7%, pump
oil 98.9%, petroleum ether 99.3% and crude oil 99.1%). In addition, the
recycling performance of the membrane was also evaluated using the
hexane/water mixture as a representative. Note that the membrane was
just simply rinsed with water prior to next separation. Surprisingly,
after 60 separation cycles, the membrane still preserved a high se paration efficiency above 98.2%, along with a high flux over 5443 L m-2
h-1 (Fig. 6c), indicating its excellent recyclability. In addition, the L CNM could still maintain remarkable underwater superoleophobicity
with ultralow oil adhesion after 60 separation cycles (Fig. S4). From
Table 1, the flux, separation efficiency and recyclability of L-CNM were
comparable to or even outperformed many early reported electrospun
membranes [48–56]. Besides, the bearable oil height for L-CNM was
very high, up to 100.5 cm as demonstrated in Fig. 6d. The intrusion
pressure was calculated to be 6.95 kPa. It means the oil (hexane) cannot
pass through L-CNM under this pressure. This intrusion pressure was
comparable to that of SSA-PAAS-g-PVDF composite membrane as re ported by Zhang et al. [57], demonstrating a good stability of the se parating system.
In order to further evaluate the emulsion separation performance of
L-CNM, Tween 80-stabilized toluene-in-water emulsion was prepared
by mixing toluene with a Tween 80 water solution in a ratio of 5:100
(v/v) under vigorous stirring for 2 h. This emulsion was used as the feed
solution to permeate through L-CNM by gravity. After separation, the
white milky emulsion became transparent and colorless. Under a po larizing microscope, the whole view was composed of numerous micro scale droplets in the feed solution, whereas no oil droplet was observed
in the filtrate (Fig. 7), suggesting the successful separation of the
emulsion. The emulsion separation efficiency was calculated to be
98.5% according to the absorbance changes on toluene’s characteristic
peak at 261 nm in the UV–vis spectra before and after separation [58].
Moreover, the water flux was determined to 436.4 L m-2 h-1 during the
emulsion separation process, which was comparable to or outperformed
other electrospun membranes [49,53].
The effects of La(OH)3 loading on the oil/water separation and the
dye adsorption capacity of L-CNM were also investigated and discussed
in detail in the Supporting Information (Fig. S5 and S6).
3.5. Selective adsorption of dyes
Considering that the organic dyes are another major contaminant in wastewater apart from oils, the dye adsorption capacities of L-CNM was
further evaluated. Cationic MB as well as anionic MO and CR were
employed in the adsorption tests, and the corresponding UV–Vis spectra
were displayed in Fig. 8a–c. Once the membrane was immersed in the
CR solution, the characteristic peak for CR at 498 nm decreased dras tically (Fig. 8c), showing the fast adsorption capacities. However, when
the membrane was immersed in either MB or MO solution, the resultant
UV–Vis spectra almost unchanged even over a prolonged time period of
240 min (Fig. 8a and b), suggesting the negligible adsorption toward
MB and MO. This intriguing phenomenon inspired the further ex ploration of the membrane in selective adsorption of different dye
pollutants. Two different mixed dye solutions, namely, MB/CR and
MO/CR (1:1 vol ratio), were prepared. When the membrane was put in
the MB/CR solution, the absorption peak at 498 nm corresponding to
CR decreased markedly, while the peak at 664 nm assigned to MB was
barely changed (Fig. 8d), indicating that CR was selectively separated
from the MB/CR mixture. And the separation efficiency was calculated
to be 91.2%. For the MO/CR mixture (Fig. 8e), only one maximum
characteristic peak appeared after mixing because the maximum ab sorption peaks of MO and CR overlapped with each other, like the dye
mixture of neutral red (NR) and reactive blue (RB) as reported by Cai
et al. [59]. Specially, the shoulder peaks of CR at 238 nm and 340 nm
weakened evidently while the shoulder peak of MO at 270 nm just
decreased slightly after adsorption (Fig. 8e and Fig. S7). Moreover, the
color changed from orange red for the MO/CR solution to pure orange
for MO (Fig. 8f and Movie S3), suggesting that CR was selectively ad sorbed by the membrane while MO was still left in the solution. These
results indicated that L-CNM was capable of not only separating dye
mixture of opposite charges, but also efficiently separating dye mixture
with the same charge.
Fig. 6. (a) Gravity-driven oil/water separation experiment on L-CNM. Oil (hexane as the representative oil) was dyed with Oil Red O to aid visualization. (b)
Variations in water flux and separation efficiency of L-CNM for different types of oil/water mixtures. (c) Evolution of the permeation flux and the corresponding
efficiency of L-CNM for hexane/water mixture over 60 cycles (hexane as the oil example). (d) Measurement of oil (hexane as the representative oil) intrusion pressure
of L-CNM. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
3.6. Adsorption mechanisms
To examine the nature of interaction between CR and the adsorbent,
the Freundlich [60] and Langmuir isotherm [61] models were used to
simulate the adsorption isotherm. Fig. S8 provided the Freundlich and
Langmuir isotherms of L-CNM in the solutions with different CR con centrations. The experimental data could be better fitted by the
Freundlich model (R2 = 0.998) as compared with the Langmuir model
in term of the correlation coefficient (R2 = 0.872). This result indicated
the multilayer adsorption of CR on the heterogeneous surface of L-CNM
[60]. The dye adsorption capacity of L-CNM was calculated to be
624 mg/g, much higher than that of CNM (260 mg/g). Moreover, the
CR removal efficiency as a function of time at 30 mg/L initial con centration was shown in Fig. S9. In the first 1 min, almost 65% of total
CR was removed by L-CNM, while 120 min was needed for CNM to
achieve the same result. These results manifested that L-CNM had a
faster adsorption rate and a higher adsorption capacity toward CR,
which should be ascribed to the tremendous active sites provided by La
(OH)3 nanosheets.
ATR-FTIR spectra of L-CNM before and after CR adsorption were
given in Fig. S10. After adsorption, the new peaks at 1022 cm-1 (at tributed to the stretching vibration of S]O) and 1163, 1378, 1469 cm-1
(assigned to the framework vibration of benzene rings) appeared, which
confirmed the adsorption of CR onto L-CNM [62]. Note that the peak at
3421 cm-1, corresponding to the stretching vibration of OeH, shifted to
a lower wavenumber (3378 cm-1) after adsorption. In addition, the
OeH stretching peak was also broadened. This suggested that hydrogen
bonds were formed between the hydroxyl groups of L-CNM and the
amine groups of CR molecules [63]. Considering the fact that the ad sorption of anionic CR was significant while the adsorption of anionic
MO was almost negligible, it was speculated that the hydrogen bonding
interaction rather than the electrostatic attraction might play the main
role in the adsorption process. In addition, the electron-deficient La 4f
orbitals in La(OH)3 could induce strong affinity toward electron-rich
sulfonic groups in CR, leading to the high CR adsorption capacity.
Fig. 7. Digital photographs and polarizing microscopy photographs of the
Tween 80-stabilized toluene-in-water emulsion (a) before and (b) after se paration.