Conformal Nitrogen‐Doped TiO₂ Photocatalytic Coatings for Sunlight‐Activated Membranes

Photocatalytic degradation of organic contaminants is enticing for addressing challenging, nontraditional water sources. A novel nitrogen-doping method is utilized to grow conformal titania coatings with visible-light activity on porous membranes. The resulting membranes exhibit effective degradation of model organic species in simulated sunlight while at the same time requiring substantially lower transmembrane pressure than undoped membranes. Porous membrane materials offer a wide range of applications including water treatment, energy production and conversion, environmental remediation, pharmaceutics, and biotechnology.1 Reflecting their societal impact, annual global demand for membranes is expected to reach $26.3 billion by 2019.2 As such, vast efforts are directed toward investigating next-generation membrane materials that offer improved functionality and performance.3 Here we show a facile, reliable vapor-phase synthesis method for the fabrication of conformal thin films of nitrogen-doped TiO2 on membranes via atomic layer deposition (ALD). This novel means to introduce nitrogen doping into the TiO2 has the effect of shifting the optical absorption onset from the UV range into the visible. Such doping enables the use of lower energy light, including sunlight, to activate the photocatalytic function of the membrane. Our approach offers compatibility with virtually any porous inorganic membrane and allows for uniform, conformal control of materials exclusively on the surface of the substrates—including internal surfaces of porous substrates—and atomic-level thickness control and pore-size tuning. Ceramic membranes exhibit burgeoning activity because they offer relative thermal, chemical, and mechanical robustness.1 Typical techniques to fabricate ceramic membranes include chemical vapor deposition, solid state sintering, and sol–gel processing.4 Titanium dioxide (TiO2) is one of the most widely studied materials in this class due to its photocatalytic ability as an additional function to the membrane system.5 The conventional deposition technique to modify and functionalize membranes with TiO2 is solution-based sol–gel synthesis via dip coating, which typically results in inhomogeneous deposition, pore clogging, and inconsistent membrane production. ALD, which has been predominately used in the field of microelectronics, offers uniformity and conformality with precise tunability of deposition layer thickness.6 The intrinsic optical band gap of TiO2 is 3.2 eV (for anatase crystalline phase), thereby necessitating UV light to drive its photocatalytic function.7 This increases costs and limits practical use, especially in developing world scenarios. Visible-light-active photocatalytic activity of TiO2 can be achieved through metal (e.g., Fe, Cr, Co, Mn, V, or Mo) or nonmetal (e.g., N, F, S, B, or various carbon materials) doping, and codoping, with nitrogen being one of the most promising dopants due to its comparable atomic size with oxygen, small ionization energy, and high stability.8 Despite recent progress in various routes for fabrication of N–TiO2 in a thin-film form (e.g., chemical vapor deposition, sol–gel methods, pulsed laser deposition, ALD, and magnetron sputtering),9 effective utilization of N–TiO2 in membrane technology remains elusive. Thus far, solution-based N–TiO2 nanoparticles have been applied for membrane materials.10 A better, controllable deposition technique is needed to tune pore structure, interfacial chemistry, and physical properties of the membranes without sacrificing permeability. Here we demonstrate for the first time direct, controlled, conformal deposition of N–TiO2 onto membrane substrates via ALD. We used two qualitatively different membrane pore morphologies (i.e., tortuous pore network and cylindrical pores, see the details in the “Experimental Section”). Figure 1 shows a schematic illustration of the method where self-limited surface reactions between TiCl4 and NH4OH result in a uniform conformal deposition of N–TiO2 on a porous matrix. One cycle of ALD consists of four steps: (1) TiCl4 vapor is pulsed and forms a monolayer of chemisorbed TiClx species on the surfaces of the porous membrane; (2) unreacted TiCl4 and HCl byproducts are purged out by nitrogen; (3) the next precursor, NH4OH, is subsequently pulsed yielding a mixture of gaseous H2O and NH3 which react with the chemisorbed TiClx species yielding a monolayer of N–TiO2; and (4) unreacted precursor and HCl byproducts are again purged out by nitrogen. This cycle is repeated to achieve the N–TiO2 target thickness. Figure 2a shows optical absorption spectra of N–TiO2 and TiO2 membranes. A TiO2 membrane fabricated using TiCl4 and water as precursors is used as a control system. Difference in the absorption edges of TiO2 and N–TiO2 in the visible range from 400 to 500 nm is evident and consistent with the calculated imaginary parts of the dielectric functions of N–TiO2 and previous experimental results.11, 12 This shift of the absorption edge is due to dominant transitions corresponding to N2pπ to Tidxy, instead of from O2pπ as in undoped TiO2. The effect of N-doping resulted in a vivid color change of the membrane from white to yellow as shown in the inset of Figure 2a. To better understand the N–TiO2 ALD process on the porous membranes, we prepared ALD films with thicknesses of ≈60 nm at different temperatures ranging from 250 to 450 °C. The corresponding absorption spectra for these materials are shown in Figure 2b. These data reveal that the absorption edge shifts into the desired visible range (400–500 nm) when the deposition temperature was 375 °C and above. Small changes in absorption intensities among samples prepared above 375 °C may be due to slight thickness variations. No significant change in absorption occurred when the deposition temperatures were between 250 and 350 °C. TiO2 deposition without any doping at the same temperature range also did not result in any absorption shift; the absorption spectrum of a pure TiO2-coated membrane fabricated at 400 °C is shown in Figure 2b for comparison. The free energy changes (∆G, in kJ mol−1 Ti) for these two processes are plotted as a function of temperature in Figure 2c. At temperatures below ≈300 °C, ∆G for the TiN reaction is positive and for TiO2 it is negative, indicating that the ALD films should be pure TiO2 at these deposition temperatures. At temperatures above ≈300 °C, ∆G for the TiN reaction becomes increasingly more negative so that TiN formation becomes more favorable, and this agrees fairly well with the onset of visible absorption in the N–TiO2 at 375 °C. X-ray diffraction (XRD) measurements performed on the membranes revealed exclusively the anatase crystalline phase with no traces of rutile or titanium nitride structures (Figure 2d). Consequently, the ALD N–TiO2-coated membranes do not require postannealing to obtain the desired anatase phase, which is the preferred crystal form for photocatalytic activity due to efficient bulk transport of excitons to the surface, longer charge carrier lifetimes, and low recombination rate of charge carriers on the surface.14 ALD coatings on anodic aluminum oxide cylindrical pore membranes were used for basic material characterization to simplify interpretation, and all membrane performance studies involved membranes having a ZrO2/TiO2 with tortuous pore network as the substrate to mimic realistic conditions. To evaluate the photocatalytic activity of N–TiO2 under visible light illumination, methyl orange (MO) was chosen as a model pollutant15 (MO is an example of an azo dye and is representative of numerous organic pollutants in water bodies). Figure 3a shows a representative absorption profile of an aqueous solution of MO (4 mg L−1) in the presence of an N–TiO2-coated membrane prepared at a temperature of 400 °C at different illumination times. The maximum absorbance at 464 nm, corresponding to an nπ* transition due to substituent perturbation (i.e., addition of (CH3)2N and SO3Na on azobenzene), was used as a proxy for MO concentration.16 The maximum band intensity decreased monotonically as a function of illumination time indicating photodegradation of MO in the presence of the N–TiO2 membrane. No noticeable photocatalytic activity was observed when the N-TiO2 film was absent, as shown in Figure 3b. To understand if there is any correlation between deposition thickness and photocatalytic activity, the number of ALD cycles was varied (Figure 3b). The corresponding N–TiO2 thicknesses for 300, 450, and 600 ALD cycles were 21.08 ± 0.19, 30.73 ± 0.07, and 39.70 ± 0.08 nm, respectively (see the “Experimental Section”). Relative concentrations of the MO pollutant (C/C0) in the presence of N–TiO2 membranes with different deposition thickness are 60.2%, 70.0%, and 76.4% for 300, 450, and 600 cycles, respectively, following 100 min of photoexposure. These results show that the population of available photocatalytic surface sites decreases as the number of ALD cycles increases even though there is more active material deposited onto the membranes. Visual porosity analysis can provide some quantification of this phenomenon as shown in Figure 3c. Surface porosity gradually decreases as the number of ALD cycles increases from 9.4% for 300 cycles, 6.2% for 450 cycles, to 1.8% for 600 cycles (see the “Experimental Section;” Figures S2–S4, Supporting Information). These results imply that, even though there is more active material deposited on the surface of the membranes to effectively absorb the incident light, flow of solution containing the pollutant is likely inhibited by restrictive pore sizes. Catalytic performance is a complex interplay between light absorption, water/solute flux through the membrane, and population of surface active sites. Changing the N–TiO2 thickness affects each of these parameters simultaneously—including a secondary effect on the number of charge carriers reaching the water/titania interface (as opposed to suffering from recombination) as the films increase in thickness, which hinders deciphering their individual roles. In order to demonstrate the multifunctional capabilities of these functionalized membranes, we examined their filtration performance in addition to their catalytic properties. Figure 4 shows pure water flux of the fabricated membranes as a function of applied transmembrane pressure ranging from 6 to 30 psi (a range selected to remain safely below the rated pressure limit of the commercial substrates). Since the ALD coatings reduce the pore size, one would expect the flux to decrease following the deposition. Surprisingly, the N-coated TiO2 membranes presented substantially higher flux at all tested pressures as compared to a control membrane without the coating. The average pore size of the control membrane is 140 nm, whereas effective pore sizes of N–TiO2-coated membranes are ≈100, 80, and 60 nm for the 300, 450, and 600 ALD cycle membranes, respectively, based on the assumption that the N–TiO2 coating is uniform and conformal in the pores. Specifically, the average permeance of the N–TiO2 membrane by 300 cycles (6.32 × 103 L m−2 h−1 at 1 bar) has almost an order of magnitude increase as compared to the control system (9.78 × 102 L m−2 h−1 at 1 bar). Permeances of N–TiO2 by 450 and 600 cycles are 1.59 × 103 and 2.89 × 102 L m−2 h−1 at 1 bar, respectively. Membrane reliability tests using the control membranes showed that the flux difference from manufacturing variations is significantly smaller than the measured flux differences among processed membranes (see Figure S5, Supporting Information). Also, the control membrane annealed at 420 °C did not show any significant change in the flux performance (Figure S6, Supporting Information). This enhancement of the permeance despite smaller pore sizes may be largely due to augmentation of hydrophilicity. To experimentally probe this hypothesis, TiO2 deposition without N-doping via ALD was carried out based on the same experimental procedures, and the permeance between TiO2 and N–TiO2, both having ≈20 nm deposition thicknesses, was compared (see the “Experimental Section;” Figure S7, Supporting Information). Although the deposition thickness is the same, the permeance of the N-doped TiO2 membrane is dramatically higher than the TiO2 membrane (≈500 L m−2 h−1 for TiO2 vs 6000 L m−2 h−1 for N–TiO2 at 1 bar). Moreover, surface wettability studies show that the N-doped TiO2 membrane exhibits enhanced hydrophilicity as compared to the TiO2 membrane (water contact angles for N-doped TiO2 and TiO2 membranes with the same deposition thickness of 20 nm are 22.9° ± 2.52° and 107.4° ± 0.82°, respectively) (see Video S1 Supporting Information), consistent with previous reports by using various fabrication routes.12, 17 Based on these macroscopic studies, we conjecture that nitrogen doping plays a vital role in chemically enhancing hydrophilicity, which facilitates water transport. Future studies will explore the relative role of surface properties including composition, charge, and morphology on the membrane performance. In summary, this study demonstrates a novel means to fabricate nitrogen-doped TiO2 membranes via ALD without any need for postannealing. Conformal and thickness-controllable doping results in a lowered band gap enabling photocatalysis using visible light. N–TiO2 membranes show significantly enhanced permeance and photocatalytic activity for degrading organic pollutants under solar irradiation. The work presented here provides the basis for creating tailored multifunctional inorganic membranes capable of destroying pollutants and self-cleaning under solar irradiation that can be utilized in various applications, most notably in water treatment and purification. Materials Preparation: For the fabrication of TiO2 and N-doped TiO2 (N–TiO2) membranes, Al2O3 with cylindrical pores having 200 nm pore size (Anodisc, Whatman) and ZrO2/TiO2 with tortuous pore network having 140 nm pore size (Sterlitech Corp.) with 47 mm membrane diameters were used as model substrates. Either TiO2 or N–TiO2 was deposited by atomic layer deposition (ALD) via alternate exposure to TiCl4/water and TiCl4/NH4OH, respectively. The NH4OH (Sigma Aldrich) consisted of a 30% aqueous NH3 solution. At room temperature (21 °C), this solution has a vapor pressure of 17 psi, and composition that is 95% NH3 and 5% H2O. These conditions are well suited for N–TiO2 ALD given the much higher reactivity of H2O compared to NH3 using TiCl4. Compared to previous studies that alternated between TiCl4/NH3 cycles for TiN ALD and TiCl4/H2O cycles for TiO2 ALD to produce a layered N-doping profile, this physical mixture of NH3 and H2O would likely produce a more homogeneous N-doping profile. These membrane materials were prepared in a custom hot-walled viscous flow ALD reactor described elsewhere.18 The growth layer with a desired thickness is controlled based on the temperature, pulse times, and number of process cycles. Briefly, the surface reactions were performed at various temperatures ranging from 250 to 450 °C using nitrogen as a carrier gas at a mass flow rate of 330 sccm and a pressure of 1 Torr. The membrane substrates were placed in the reactor with the long axes of the pores normal to the carrier gas flow. The ALD reactant pulse timing sequence (AB cycle) is given by t1–t2–t3–t4 where t1 is the exposure time for reactant A, t2 is the purge time following the A exposure, t3 is the exposure time for reactant B, and t4 is the purge time following the B exposure. ALD timing sequences (where the units are in second) for TiO2 and N–TiO2 were 0.75–5–0.75–5 and 0.75–5–0.25–5, respectively. Materials' Characterization and Analysis: Absorption measurements were carried out by using a Cary 5000 UV–Vis-NIR spectrophotometer. To determine the crystal structures of TiO2- and N–TiO2-deposited membranes, X-ray diffraction was used (Bruker X-ray System D8 ADVANCE A25). Si wafers as witness pieces were loaded along with the membranes into the ALD chamber during TiO2 and N–TiO2 deposition cycles in order to determine the thickness of the materials via spectroscopic ellipsometry (SE, J. A. Woollam, Co. alpha-SE). For the ΔG calculations, the HSC Chemistry program was used (Outotec Research Oy, Espoo, Finland). For investigation of morphology of the fabricated membranes, scanning electron microscope (SEM) images were recorded using a JEOL 7500F field emission SEM operating at 10 kV. For surface visual porosity determination, all SEM images were taken at 40 000× magnification operating at 10 kV. The surface visual porosity as a result of different ALD cycles was determined by converting SEM images into binary format, and porosity was calculated using image processing (total number of pixels representing the pores divided by the total number of pixels of the entire image) in Matlab (The MathWorks, Inc., Natick). To determine the permeance of the fabricated membranes, water flux of the membranes was measured by using a stirred filtration dead-end cell (EMD Millipore), where the flow of water (in-house reverse osmosis water) is perpendicular to the membrane surface. The water was pushed through the membrane by applied transmembrane pressures ranging from 5 to 30 psi. Nitrogen gas line was attached to the cell and was used to maintain the applied pressure gradient (note that nitrogen gas is chosen as compressed air results in pH shifts). A magnetic stirrer was used throughout the experiments to provide a shear force needed to maintain homogeneous aqueous media. The weight of permeate was measured on a digital balance connected to a LabVIEW program to control the setup as well as continuously log the flux at a given applied pressure. For photocatalytic degradation experiments, 50 mL of oxygenated aqueous methyl orange (4 mg L−1) was placed into the dead-end cell. The cell was constructed by placing a membrane at the bottom of the cell, and the top of the cell was covered and completely sealed. A xenon lamp (Perkin Elmer) was used as a light source (≈100 mW cm−2 irradiance) with an IR water filter (Newport). Throughout the experiments, the solutions were stirred at constant speed to ensure homogeneous dispersions. To ensure the accuracy of absorption measurements, methyl orange solution was stirred at least for 1 h in the dark, and absorption spectra were taken to confirm constant optical density (constant O.D. (optical density) should be read after any adsorption/desorption of MO molecules on the surface of membranes) before the illumination. For contact angle measurements, a Dataphysics Contact Angle Measurement OCA Model ES tool was used with a 30 fps camera. Deionized water drops with volumes of 0.6 µL were brought into contact with the membrane surface, and the reported contact angle was the first stable frame after the drop detached from the needle. Use of the Center for Nanoscale Materials, an Office of Science user facility, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357. This material was based upon work supported by Laboratory Directed Research and Development (LDRD) funding from Argonne National Laboratory, by the Office of Science, of the U.S. Department of Energy under Contract No. DE-AC02-06CH11357. As a to and this provides by the Such materials are and may be for are not or from than should be to the The is not for the or functionality of any by the than should be directed to the corresponding for the