4.1. Construction of the CNTs Network on the Mullite Ceramic Substrate
As illustrated in Figure 1
a, a hierarchical network was grown on the surface of the mullite ceramic substrate. The pristine mullite ceramic substrate had an uneven surface with randomly distributed spaces between the sintered mullite particles. The average pore size of the mullite ceramic substrate was found to be 1.02 µm, as reported in our previous study [26
]. CNTs were successfully grown on the surface of the mullite ceramic substrate via the CVD method. The CNTs were intertwined with each other, forming a network-like hierarchical structure and were observed on the surface of the mullite ceramic substrate (Figure 1
b,c). The special structure of the mullite-CNT composite membrane was expected to allow for high fluxes at low operating trans-membrane pressures due to the highly porous interlocking structure of the CNTs, which possibly could achieve a highly efficient separation of bacteria.
TEM results (Figure 1
d) also clearly revealed that the resulting nanomaterials were hollow, being CNTs rather than solid carbon nanofibers, with an average diameter of 41 nm and lengths of up to several micrometers. Tip growth was the dominating mechanism of CNTs, and, as expected, catalyst particles could be detected at the tip of the nanotubes (Figure 1
e), as described in the previous report [28
]. The distance between two adjacent planes, d-spacing, was measured to be 0.33 nm, which was characteristic of the (002) plane of CNTs. The presence of metallic nickel in the CNTs layer revealed that the NiO phase was reduced during catalyst pretreatment. Nano-sized metallic nickel was quite active for the decomposition of methane [29
presents the corresponding diameter distributions of the CNTs, measured from the SEM image in Figure 1
c. The average diameter of the CNTs was ~38.7 nm, with a relatively narrow diameter distribution range of 20–50 nm.
4.5. Highly Stable Flux for Bacterial Removal
As presented in Figure 7
, the pure water flux was 38.7 L·m−2
, which was higher than the bacteria–water flux due to the absence of bacteria. The mullite-CNT composite membrane was demonstrated to effectively remove bacteria from drinking water, with a highly stable long-term flux, for the 24 h filtration process, which was driven by a trans-membrane pressure of 0.1 MPa. The initial permeate fluxes of the mullite-CNTs composite membrane were 9.5 L·m−2
and 5.4 L·m−2
for E. coli
and S. aureus
, respectively, at 0.1 MPa. Subsequently it was characterized by a very mild decrease in fluxes, of which values were 7.9 L·m−2
and 4.4 L·m−2
of E. coli
and S. aureus
, respectively, at 48 h of operation. The fluxes of E. coli
and S. aureus
were decreased by 17.3% and ~19.2%, respectively, during the filtration period. During the long filtration experiment, the fluxes of the composite membrane decreased slightly and showed excellent performance, with outstanding efficiency in the removal of bacteria, as illustrated in Figure 7
; E. coli
and S. aureus
have been eliminated totally based on the PFC method at 2 h, 24 h, and 48 h.
The results showed that the removal of bacteria from the 107
per mL initial concentration was complete, without any bacteria detected by the PFU method at the filter outlet. High levels (100%) of removal of bacteria, E. coli
and S. aureus
, were shown in Figure 8
throughout the filtration experiment. Additionally, it showed that the mullite-CNT composite membrane could be used successfully to obtain bacteria-free water at a long operation time. The results were due to the formation of a “network-like” hierarchical structure [30
] by the CNTs on the mullite ceramic substrate. The bacteria were only trapped on the CNTs network surface via surface filtration, without pore plugging, endowing the mullite-CNTs membranes with unprecedentedly low fouling propensity in order to keep high flux with operation time. To confirm this and to check the separation mechanism of the mullite-CNTs membranes for the two model bacteria, SEM images of E. coli
a,b) and S. aureus
c,d) on the mullite-CNT composite membrane after the filtration are discussed below.
The crosslinked CNTs network formed unique shape and pore systems, confirmed by the SEM pictures as shown in Figure 9
a,c. Figure 9
a,c also demonstrated that the mullite-CNT composite membrane was porous. The pore size of the mullite-CNTs composite membrane was very small compared to the size of bacteria, so bacteria cannot permeate through the membrane; only water molecules were allowed to permeate through the membrane pores. Higher porosity indicated that a higher number of pores and channels were available for water transportation. SEM images also indicated that the E. coli
and S. aureus
cells were completely retained by the CNTs layer, by a sieving mechanism, due to size exclusion. The CNTs membranes for bacteria removal based on size exclusion or sieving, often required high pressure for operation, which were prone to pore plugging and performance deterioration upon filtration of environmental samples in the studies discussed before. A major advantage of using the mullite-CNT composite membrane, over conventional membranes, lies in the high flux with excellent removal efficiency over a long period of time.
At higher magnification, Figure 9
b,d, it was found that E. coli
and S. aureus
cells were captured by the tangled CNTs networks. The bacteria were removed along with a surface-filtration, not via in-depth pore blocking, which would lead to a rapid decrease flux depth-filtration mechanism, that is, captured by the CNTs on the surface of the mullite-CNT composite membrane. Accordingly, the high flux was retained throughout the filtration time due to the formation of the porous membrane surface with a special randomly-oriented CNTs network structure, featuring very high three-dimensional open porosity, allowing water to rapidly transport in any direction. The high flux was also predicted by the “flow enhancement” of the slip flow due to the small diameter of the CNTs [31
]. The mullite-CNTs composite membrane was more effective in capturing the bacteria on the surface, but not in the pore channels of the composite membrane. This suggests that the filtration step may have little effect on the water permeation flux without blocking the pores, thus overcoming the inherent limitation of the trade-off effect between high removal efficiency and high flux.
A fluorescence-based viability kit (Live/Dead), as shown in Figure 10
, was used to determine the percentage of E. coli
(a) and S. aureus
activated on the mullite-CNT composite membrane and the mullite ceramic substrate, used as control samples for the comparison. The two bacteria, E. coli
and S. aureus,
demonstrated greater resilience to exposure to the CNTs membrane in this study, which was not consistent with the results reported dealing with CNTs toxicity tests on bacteria physiology [32
]. The two hour exposure time led to inactivation of only less than 1% for both the two bacteria; the inactivation of bacteria on the CNTs membrane did not increase significantly over the longer incubation time of 6 h exposure to the mullite-CNT composite membrane. Studies have shown that MWCNTs were less effective than SWCNTs in cell inactivation, the top layer in this study was MWCNTs, so the effect may be not as evident as reported by the literature [32
]. It has been hypothesized that direct contact with the CNTs was necessary to attain inactivation. The carbon nanotubes were fixed on the surface of the mullite ceramic substrate, which provided a small contact area with the bacteria, and the contact time between the membrane surface and the filtrate was relatively short. Due to the short contact time and small contact area, between the membrane surface and the bacteria, the MWCNTs layer and the metallic nickel showed less antimicrobial activity towards E. coli
and S. aureus
, and the influence of the prepared mullite-CNT composite membrane on the inactivation of the trapped bacteria was negligible.
Release of CNTs into the environment is a major concern. To assess any possible environmental risk of the developed ceramic-CNT membrane in water treatment due to the existence of some metal ions, such as Al and Ni, in the membrane starting materials, a gas-cleaning run was conducted to ensure that no CNTs were unbound from pore channels under the shear stress of water. The selected permeated water samples were examined by SEM and no CNTs were found. Additionally, the concentration of these metal ions in the filtrate was measured using ICP-MS (inductively coupled plasma mass spectrometry, Nex ION 300D, PerkinElmer Inc., USA). The results shown in Table 1
clearly show very low levels of these metal ions, consistently meeting the standard of drinking water criterion issued by the World Health Organization (Guidelines for Drinking-Water Quality Fourth Edition, WHO, 2011).
Therefore, the release of CNTs from the membrane might not be of a concern for the impact on the environment and ecosystems. The mullite-CNT composite membrane was able to stand a pressure difference of 0.1 MPa, suggesting that a pressure difference of 0.1 MPa does not weaken the integrity of the membrane. Thus, it was supposed that the strong adhesion force between the CNTs layer and the porous mullite ceramic substrate was important for long-term operation in water treatment application.