Titanium dioxide in its pure wide bandgap “white” form is a non-toxic, efficient, and practical photocatalyst, but predominately absorbs light in the ultraviolet range of the spectrum. The absorption range, however, can be extended into the visible by doping with oxygen vacancies or impurities, such as nitrogen, giving the material a black or brown appearance. To date, nitrogen-doped titanium dioxide has primarily been produced with approaches that require long processing times or multi-step synthesis protocols. Here, we present a fast (timescale of tens of milliseconds) all-gas-phase process, which enables the seamless tuning of the optical properties of titanium dioxide nanoparticles from white to brown. Titanium dioxide particles were synthesized through injection of tetrakis (dimethylamido)titanium (TDMAT), argon, and oxygen into a nonthermal plasma. The positions of the electrode and oxygen inlet relative to the precursor inlet are found to strongly influence particle properties. Variation of these parameters allowed for control over the produced particle optical properties from large bandgap (white) to small bandgap (brown). In addition, the particle microstructure can be tuned from amorphous to crystalline anatase phase titanium dioxide. The photocatalytic performance was tested under solar irradiation and amorphous particles exhibit the highest degree of photocatalytic decomposition of the dyes methyl orange and methylene blue.
Titanium dioxide (TiO2) is a stable, non-toxic, and efficient light-activated catalyst [1–3]. The anatase phase of TiO2 typically performs better as a photocatalyst than its rutile counterpart , but a combination of phases tends to produce the best photocatalytic activity [5–7]. Amorphous TiO2 also has significant promise as a material for photocatalytic applications [8,9]. The band edge structure of TiO2 has exceptional oxidizing power and sufficient energy to reduce hydrogen. However, to allow both photo-reduction and photo-oxidation, dopant levels within TiO2 should remain above the H2 evolution potential and below the O2 potential [2,10]. Dopants should also seek to increase or maintain carrier lifetime and transport through the crystal to facilitate reactions at the surface of the material [11,12]. To accomplish these goals, anionic dopants like nitrogen are strong candidates [2,11,13–15]. Nitrogen in particular has been shown to be a very capable dopant because its valence band orbitals overlap with those of oxygen leading to a narrowing of the band structure [2,16]. However, new techniques for the production of nitrogen (N)-doped TiO2 nanoparticles that are controllable, single-step, and scalable remain in high demand [10,11,14,17].
A number of processes that include sputtering [18,19], atomic layer deposition (ALD) , nitrogen plasma treatment [21,22], oxidation of titanium nitride [9,23], microwave plasma enhanced chemical vapor deposition (PECVD) , plasma-enhanced ALD (PEALD) [24,25], atmospheric pressure plasma deposition [26,27], micromechanical action , TiO2 treatment with ammonia [29–31], plasma synthesis , flame-synthesis [33–35], and wet synthesis methods [36–40] exist for the production of N-doped TiO2. However, long process times and multi-step procedures are common among these synthesis procedures. Flame synthesis is the industry gold standard for the gas-phase synthesis of P25 white titanium dioxide nanoparticles. This method can provide strong photocatalytic activity with reliable large scale production, albeit without significant visible light absorption or dopants . Single-step procedures for producing bottom-up doped nanoparticles with tunable properties are desirable as they allow for the production of materials that could be used to increase the overall photocatalytic performance over a pure rutile or pure anatase film [5,42]. Furthermore, direct deposition from the gas phase of these particles, without solvents and chemical waste, allows for bottom-up manufacturing of high surface area nanoparticle films with ample surface sites for photocatalysis .
Flame synthesis [33–35,44] and plasma synthesis  methods are approaches that lend themselves to fast, single-step production of N-doped TiO2 nanoparticles. A two-stage flame spray pyrolysis was reported by Huo et al. to, first, produce TiO2 nanocrystals and, second, dope them with nitrogen by injecting a nitrogen precursor (water + ammonia) above the flame . This method yielded TiO2 powders with surface substitutional nitrogen doping leading to enhanced absorption in the visible and an apparent yellow color. Bi et al. utilized an approach similar to this two-stage synthesis approach but report that nitrogen surface doping is mainly interstitial . Smirniotis and coworkers presented an elegant single-step approach by injecting a stock solution containing both titanium and nitrogen precursors into a flame reactor [33,34]. The resulting powder contained predominately interstitially doped nitrogen, also leading to the enhancement of visible light absorption, yielding strongly improved photocatalytic decomposition of 4-chlorophenol. Buzby et al. reported N-doped TiO2 nanocrystals produced with an inductively coupled plasma, using titanium tetraisopropoxide and anhydrous ammonia as precursors. The authors suggest that their nanoparticles are likely substitutionally doped and report enhanced photocatalytic degradation of 2-chlorophenol. However, the paper does not report specific information about absorption spectra.
Gas phase approaches for TiO2 nanoparticle production to date have not shown control over the levels of substitutional and interstitial nitrogen-doping. However, deep level substitutional nitrogen dopants are of particular interest to significantly increase optical absorption deep into the visible. Moreover, current approaches lack the ability to tune the material phase and they mostly produce polydisperse particle sizes. Hence, the ability to tune particle properties such as dopant location between substitutional and interstitial nitrogen and phase between amorphous, rutile/anatase, and anatase are of great interest as such nanocrystal films could provide significant advantages over single phase or single dopant films [5,42].
Here, we demonstrate that N-doped TiO2 nanoparticles with controlled compositions can be synthesized and deposited as nanoparticle films from the gas phase with a single-step nonthermal plasma process. Particle properties can be tuned by varying the oxygen injection location relative to the onset of particle nucleation within the reactor. The particles are synthesized using tetrakis (dimethylamido)titanium, argon, and oxygen as precursors. These nanoparticles have varying degrees of nitrogen doping that range from entirely interstitial to predominantly substitutional. Nanoparticle films consisting of amorphous and crystalline anatase phase can be achieved with this method. The influences of reactor geometry and plasma power on the synthesized powder are analyzed for multiple reactor configurations. The nanoparticles’ photocatalytic behavior is tested under solar irradiation with methyl orange and methylene blue without the use of metal ion dopants or hydrogen peroxide that is often used in other studies and compared against P25 white TiO2 [45,46].
Plasma Synthesis Conditions and Reactor Details.
N-doped TiO2 nanoparticles are synthesized via a nonthermal plasma process. The reactor is shown in Fig. 1. Radio frequency power from a power supply (RFPP RF5S 500 W 13.56 MHz RF power supply generator) is coupled via a matching network (Vectronics HFT-1500) to a copper ring electrode. The capacitively coupled electric field partly ionizes the gas inside a quartz tube. The quartz tube has an outer diameter of 25 mm and an inner diameter of 21.8 mm. There are three Ar streams, one to control the upstream pressure (Arup, 35 standard cubic centimeter per minute (sccm)), a second one (ArTDMAT, 35 sccm) to assist in the vapor transport of TDMAT (Tetrakis (dimethylamido) titanium(IV), C8H24N4Ti) (Sigma Aldrich 99.999%), and a third one that is injected with the O2 (Ardown, 100 sccm). TDMAT is heated in a sublimator using Brisk Heat heating tape (BWH051040L) and controller (SDC120JC-A). The heated portions of the reactor are highlighted in red in Fig. 1 and correspond to two regions, one to heat the TDMAT to generate a sufficiently large vapor pressure (HeaterTDMAT, 50 °C) and a second to prevent condensation of the TDMAT prior to entering into the reactor (Heaterup, 70 °C). For these experiments, the O2 inlet distance, blue, in Fig. 1 refers to the separation between the TDMAT/Arup inlet upstream entrance location and the midstream injection port of the Ardown/O2 gasses. The electrode position, orange arrow in Fig. 1, refers to the separation between the TDMAT/Arup entrance and the electrode. The positions of the electrode, oxygen inlet, and power are varied throughout this study. Reactions are carried out at rough vacuum pressure (3 Torr). The nanoparticles that are produced in the gas phase are flown downstream through a slit orifice that is used to maintain a pressure difference that accelerates the particles onto a translated substrate to produce nanoparticle films for characterization.
X-ray photoelectron spectroscopy (XPS) was conducted on a PHI Versa Probe III XPS and UPS (UV Photoelectron spectroscopy) system using an Al Kα source. A 55 eV band pass energy was used for 10 scans to collect high-resolution scans. The spot size for sample collections was 100 μm. Sample collection was performed under neutralizing ion and electron irradiation (dual beam charge neutralization). Peaks were fitted using PHI’s “Multipak” software. Between different samples, some peaks exhibit slightly offset values which could be the result of differences in peak fitting due to the relatively small signal-to-noise ratio present in the nitrogen XPS spectra. Peaks were adjusted according to adventitious carbon signature located at 284.8 eV. To determine atomic percentages, XPS survey scans were taken at a band pass energy of 280 eV for five scans. The atomic percentages were then calculated using “Multipak.”
Raman spectroscopy analysis was completed on a Witec Alpha 300R confocal Raman microscope with a 532 nm laser. Spectra were collected with a 1800 grooves/mm grating. Six accumulations of 20 s were performed for the samples with 0.5 in. and 1.5 in. oxygen inlet position; five accumulations of 10 s were used for the samples with 1.0 in. oxygen inlet position.
Ultraviolet-visible (UV-Vis) absorbance measurements were carried out on a Cary 700 UV-Vis spectrometer. UV-Vis spectra were collected from particles deposited on glass and normalized to 1; glass was used as a background .
For standard particle size characterization, a Tecnai T12 transmission electron microscope (TEM) was used in bright field imaging mode at an accelerating voltage of 120 kV to determine particle sizes. About 300 particles were sized for each process condition in order to fit the log-normal size distributions and histograms shown in the text. In order to plot the histograms and lognormal distributions, originpro was used. Particle sizes were binned in 1 nm increments starting from a 0–1 nm bin. The particles were manually counted and sized with imagej using roughly 230 TEM images. Although particles were sonicated in ethanol for several minutes, they still showed significant agglomeration. Hence, particles were sized primarily at the edges of the agglomerates in an attempt to obtain the clearest size distribution possible. However, these sizes may not be entirely representative of the full distribution. Geometric means, μg, and geometric standard deviations, σg, are also provided in the figures.
For high-resolution TEM and elemental analysis, a Thermo Scientific Talos F200X (scanning) transmission electron microscope (STEM) equipped with a Super-X energy-dispersive X-ray (EDX) detector operating at an accelerating voltage of 200 kV was used to acquire high-resolution TEM (HRTEM) images, high-angle annular dark field (HAADF) STEM images, as well as STEM-EDX maps. Spatially resolved STEM-EDX maps were collected with 180 × 180 pixels over a 130 × 130 nm2 area, with a dwell time of 100 μs/pixel, acquisition time of 10 minutes, and drift correction after every frame. The K-edges of Ti, O, and N were background-subtracted and integrated, producing elemental maps.
Methyl orange (MO, C14H14N3NaO3S) and methylene blue (MB, C16H18ClN3S) were used for photocatalytic degradation studies. MO powder () was purchased from Santa Cruz Biotechnology, Inc. MB solution of 1.5% (1.5 g/100 ml) was purchased from Sigma-Aldrich. For the MO photodegradation studies, 100 ml of MO dye solution was mixed with 0.01 g of TiO2 nanoparticles. For the MB photodegradation studies, 100 ml of of MB dye solution was mixed with 0.01 g of TiO2 nanoparticles. An equivalent concentration of commercial TiO2 nanopowder (Degussa P25) was measured for reference. Because MO degrades much slower than MB under similar conditions [47,48], a lower concentration of MO dye solution was used for those samples than with the MB samples. The mixture was sonicated for 15 min to disperse the nanoparticles before being stirred in the dark at 400 rpm for 120 min to achieve an adsorption-desorption equilibrium . Photocatalytic degradation was conducted under natural light clear sky irradiation at the University of Minnesota, Minneapolis, MN (latitude: 44 deg 58′ 39″ North, longitude: 93 deg 15′ 52″ West; time: 11:00 am – 3:00 pm, temperature: 79 °F) on Sunday, Jul. 5, 2020. The light intensity according to the ASHRAE clear sky model varied between 80 and 92 mW/cm2 .
At given time intervals, 5 mL aliquots were withdrawn and measured using a Cary 700 UV-Vis spectrometer. The percentage of degradation was recorded as C/C°, where C° refers to the initial concentrations and C to the remaining concentrations of the MO and MB solutions after irradiation, respectively. C and C° were estimated by the peaks of the absorbance spectra at 464 nm for MO and 665 nm for MB, respectively . Catalyst-free photolysis was conducted under the same conditions.
Results and Discussion
Influence of Reactor Geometry.
Control over particle properties such as the degree of substitutional or interstitial nitrogen doping, chemical composition, or phase depends heavily on the position of the plasma exciting electrode relative to the position of the oxygen injection port. Figure 2 illustrates the impact of the electrode position at given oxygen inlet position of 0.5″ and plasma power of 50 W. The subfigures show (a) normalized optical absorbance data, (b, c) surface characterization with XPS, (d) Raman spectroscopy results, as well as (e−g) particle size distributions determined via TEM, with inset photos showing the visual appearance of the respective samples. (For consistency, the same layout is also chosen in Figs. 3 and 4).
Figure 2(a) highlights the strong dependence of the optical absorbance on the electrode position for a discharge power of 50 W. It should be noted that all UV-vis absorption spectra presented in Figs. 2–4 are normalized to 1 to account for small difference in film thickness and density between samples. As depicted in the insets of Figs. 2(e)–2(g), TiO2 nanoparticles exhibit strong absorbance when the electrode is at the same position as the oxygen inlet, 0.5″ downstream of the tetrakis (dimethylamido) titanium (IV) (TDMAT, C8H24N4Ti) inlet; the collected powder shows a brown color, Fig. 2(e). The visible light absorbance decreases when the electrode is shifted downstream of the oxygen inlet, yielding a light yellow powder at a position 1″ downstream of the TDMAT inlet and a white color 1.5″ downstream of the TDMAT inlet, Figs. 2(f)–2(g). This difference in particle properties is likely related to the location of the onset particle nucleation relative to the oxygen inlet and to the different degrees of mixing of TDMAT and oxygen before the gas mixture enters the plasma.
When the oxygen inlet is positioned 0.5″ downstream of the TDMAT inlet and the electrode is positioned further downstream at 1″ from the TDMAT inlet, the TDMAT and oxygen mix prior to entering the plasma. When both the electrode and oxygen inlet are positioned at 0.5″, particles may begin nucleating before entering the more oxygen rich plasma region. As the electrode is shifted further downstream, oxygen and TDMAT mix more thoroughly before entering the plasma.
X-ray photoelectron spectroscopy was used in order to determine the bonding conditions of titanium and nitrogen, as shown in Figs. 2(b) and 2(c). The peak locations depend on the different degrees of oxidation and on the presence and bonding of nitrogen within the sample [52–54]. Peak locations for Ti–N are found around 396 eV and correspond to substitutional nitrogen doping [2,39,53,55–58]. Peaks located at 397 eV correspond to another substitutional nitrogen bond, albeit in a slightly more oxidized state [55,59,60]. The peak located around 400 eV is somewhat contested in literature and could correspond to either Ti-O-N [56,61] and/or to interstitial nitrogen bonded to oxygen [57,62]. Binding energies of 402 eV correspond to interstitially bound nitrogen to oxygen [56,57,62,63] or chemisorbed N2 . The small high energy peak located at 403.5 eV corresponds to either C–N–O or N–O bonds [64,65]. Titanium 2p3/2 peaks corresponding to TiO2 are located around 458 eV and its corresponding doublet 2p1/2 peak, with a fixed peak to peak separation of 5.7 eV, is located at higher energies around 463.7 eV [9,66,61,67]. TiOxNy is located around 456 eV .
Additional O and C peaks were also analyzed and are discussed in conjunction with Fig. S1 available in the Supplemental Materials on the ASME Digital Collection. XPS survey scans indicate that the C content is between 15 and 27% for most samples produced, except for those produced in an oxygen deficient atmosphere, which is consistent with adventitious C contamination due to long term air exposure, Fig. S2(b) available in the Supplemental Materials on the ASME Digital Collection. It is not believed that this C content is responsible for significant absorption or color changes from white to brown of the material as white TiO2 samples also contain C on the order of 19–27%, have poor absorption in the visible, and a clear white appearance, while highly absorbing samples of nitrogen-doped TiO2 contain between 15 and 23% of carbon, have increased absorption in the visible, and have a brown color, Figs. S2(a) and S2(b) available in the Supplemental Materials on the ASME Digital Collection. Samples that are produced in an oxygen deficient atmosphere with C on the order of 50% are black and absorb in the visible, Fig. 3(a); decreasing the C content from 50% to 21% by nucleating particles further downstream of the precursor inlet did not change the absorption in the visible significantly as the substitutional nitrogen doping level did not change, but this did change the color of the powder from black to brown which is attributed to the removal of C from the nanoparticles, Fig. 3.
XPS results in Figs. 2(b) and 2(c) show that when the electrode is located at the same position as the oxygen inlet 0.5″ downstream of the TDMAT inlet and particles likely nucleate in an oxygen-poor environment, the TiO2 particles exhibit primarily substitutional nitrogen doping which causes strong absorbance in the visible. At an intermediate electrode position of 1.0″, particles exhibit both interstitial and substitutional doping and absorbance in the visible is reduced compared to the sample with an electrode position of 0.5″. The sample produced with the electrode located 1.5″ downstream of the TDMAT inlet shows an almost exclusive interstitial nitrogen doping that is associated with a strongly reduced absorbance in the visible.
Raman spectroscopy, Fig. 2(d), along with HRTEM, was used to verify the phase and microstructure of the TiO2 particles as the nanocrystals were too small to be examined with XRD. Raman transitions located at 145 (Eg), 198 (Eg), 398 (B1g), 514 (A1g + B1g), and 638 (Eg) cm−1 are assigned to the anatase phase TiO2 while 617 (A1g), 440 (Eg), 236 (two-phonon scattering), and 143 (B1g) cm−1 are associated with the rutile phase [68–70]. Accordingly, the powders for the three electrode positions are amorphous, but contain indications of anatase bonding, Fig. 2(d).
TEM images, shown in Fig. S3(b) available in the Supplemental Materials on the ASME Digital Collection, were used to determine the particle size distribution and average particle sizes, Figs. 2(e)– 2(g). Both the mean geometric particle size of approximately 6.0 nm and the geometric standard deviation of the size distribution of approximately 1.4 change little for different electrode locations.
Band gaps of the material were calculated from Tauc plots derived from their respective UV-Vis spectra . While it is commonly assumed that optical transitions are indirect for TiO2, requiring a plot of (αhν)1/2 versus hν, with α is the absorption coefficient, h is the Planck constant, and ν is the photon frequency; this assumption is not always correct for very small nanoparticles. When the particle sizes are in the sub-10 nm range using a direct band gap model often provides better fitting to find the actual band gap values [72–74]. Shown in Fig. S4 available in the Supplemental Materials on the ASME Digital Collection are the Tauc plots for a direct band gap assumption using (αhν)2 versus hν. In Fig. S4(b) available in the Supplemental Materials on the ASME Digital Collection, the band gaps are calculated as roughly 3.16 eV for particles produced with an electrode positioned at 0.5″, 3.57 eV with the electrode positioned at 1.0″, and 3.70 eV with the electrode positioned at 1.5″, indicating a significant change in bandgap with an increase in substitutional nitrogen doping.
To test the hypothesis that substitutional doping increases when particles start to nucleate in an oxygen-poor environment, the oxygen inlet was moved further downstream to 1.5″ to provide more space for the particles to grow prior to reaching the oxygen inlet. As hypothesized, this produced particles with predominant substitutional doping as long as the position of the electrode was at or upstream of the location of the oxygen inlet, as shown in Figs. 3(b) and 3(c). Figure 3(d) demonstrates that the optical absorbance for the three tested electrode positions is comparable, though slightly decreasing in the visible range for the case that the electrode is located at the same position as the oxygen inlet (1.5″). Raman spectroscopy shown in Fig. 3(d) shows signatures of an anatase phase. The geometric mean of the particles in Figs. 3(e)–3(g), determined from TEM (see Fig. S3(c) available in the Supplemental Materials on the ASME Digital Collection), increases from 4.0 nm to 4.8 nm as the position of the electrode is moved from 0.5″ to 1.5″.
At an electrode position of 0.5″, the initial nucleation and particle growth through TDMAT decomposition in the oxygen-poor plasma. This resulted in a significant amount of the carbon incorporation into the growing particles, as shown in Fig. S2(c) available in the Supplemental Materials on the ASME Digital Collection. The inset images of the powder in Figs. 3(e)–3(f) and the XPS survey scans in Fig. S2(c) available in the Supplemental Materials on the ASME Digital Collection also highlight this significant carbon incorporation, as evidenced by the almost black appearance of particles shown in Fig. 3(e). It is believed that the primary source of absorption is due to interstitial and substitutional nitrogen; carbon is not believed to be a significant source of absorption as the three samples at 0.5″, 1.0″, 1.5″ show similar absorption features shown in Fig. 3(a). By positioning the electrode closer to the inlet of the oxygen at 1.0″ and 1.5″ electrode distances, carbon incorporation decreased significantly, Fig. S2(c) available in the Supplemental Materials on the ASME Digital Collection, while the dominant substitutional component of the nitrogen doping did not change significantly, Figs. 3(b) and 3(c). The band gaps of these materials are shown in Fig. S4(c) available in the Supplemental Materials on the ASME Digital Collection as roughly 3.77 eV for the 0.5″ electrode position, 3.86 eV for the 1.0″ electrode position, and 3.45 eV for the 1.5″ electrode position. The resulting particles had much larger bandgaps and much lower absorption in the visible than those that were produced at 0.5″ oxygen inlet position and 0.5″ electrode position.
Influence of Plasma Power.
The influence of plasma power was most significant when the position of the electrode and oxygen inlet were set to 0.5″. The particle properties and particularly the optical absorbance, Fig. 4(a), change drastically when the plasma power is changed between 10 and 110 W. Particles produced at 10 W are amorphous, Fig. 4(d), and contain only interstitial nitrogen as indicated by Fig. 4(c). These samples exhibit little absorption in the visible, Fig. 4(a) and appear visually white, as shown in the inset of Fig. 4(e).
Particles produced at 50 W had stronger signatures of an anatase TiO2 phase, Fig. 4(d) and also had the largest amount of substitutional nitrogen doping, Figs. 4(b) and 4(c). This behavior suggests that there might be a limit to the ability of crystalline samples to incorporate substitutional nitrogen into their lattice. This hypothesis is further supported by the particles produced at an oxygen inlet position of 1″, Fig. S5 available in the Supplemental Materials on the ASME Digital Collection. These particles have significant anatase peaks in Raman, Fig. S5(c) available in the Supplemental Materials on the ASME Digital Collection, and display similar doping trends as the 0.5″ electrode and oxygen inlet position case, Figs. S5(b) and S5(c) available in the Supplemental Materials on the ASME Digital Collection. However, the crystalline particles with substitutional nitrogen doping at this condition have significantly less absorbance, Fig. S5(a) available in the Supplemental Materials on the ASME Digital Collection, in the visible compared to samples that are more amorphous such as in the 50 W power case at an electrode and oxygen inlet position of 0.5″. These particles, produced at an oxygen inlet position of 1″, have identical properties other than their dopant concentration and absorption profile. This further supports the hypothesis that nitrogen doping is responsible for the increased absorption profile while particle scattering is not.
The quantity of nitrogen relative to the other elements found at 50 W is the highest among all produced samples, Fig. S2(a) available in the Supplemental Materials on the ASME Digital Collection. This maximum in doping is reflected in the UV-Vis, Fig. 4(e), in that the absorption of the 50 W sample in the visible exceeds that of samples produced at lower as well as higher powers. At 110 W, a strong indication of an anatase phase is noticed, Fig. 4(d), while the overall absorption in the visible is significantly reduced compared to the 50 W sample, Fig. 4(e), likely due to the decrease in substitutional nitrogen doping and overall nitrogen incorporation, Fig. 4(c).
The geometric mean of the particles, determined from TEM and shown in Fig. S3(b), decreased slightly with increasing plasma power from 8.1 nm at 10 W, to 5.9 nm for the 50 W case, and 6.1 nm for the 110 W case, Figs. 4(e)–4(g).
In order to further understand the underlying crystal structure of the particles, high-resolution TEM and EDX images were also taken, shown in Fig. 5. In Fig. 5(a), power is shown to correlate with increasing crystallinity of the particles, as witnessed by the inset selected area electron diffraction (SAED) patterns, with crystal grains appearing on the order of 2–3 nm for the 110 W sample; this increasing crystallization behavior is consistent with that observed in Fig. 4(d). Figure 5(b) shows a representative EDX mapping of the Ti, O, and N signals that confirm the presence of nitrogen throughout the powder. Figure 5(c) shows a zoomed out picture of the 110 W sample that demonstrates that the particle size is consistent with 6.1 nm mean size determined on the lower resolution T12 microscope, Fig. S3 available in the Supplemental Materials on the ASME Digital Collection, although individual primary particles may consist of multiple crystal grains of smaller size. Finally, the band gap for the particles was determined via a Tauc plot, shown in Fig. S4 available in the Supplemental Materials on the ASME Digital Collection, and decreased from 3.62 eV at 10 W, to 3.16 eV at 50 W, and increased again to 3.25 eV at 110 W.
In order to obtain a better understanding on the interdependence of plasma power, electrode position, and oxygen inlet position, samples were collected at multiple electrode and oxygen inlet positions ranging from 0.5″ to 1.5″, respectively, Fig. 6. Only when the electrode and oxygen inlet were at the same position did we find a strong power dependence in visible light absorption of the particles, Fig. 6. However, this dependence was strongest for the 0.5″ electrode position and 0.5″ oxygen inlet position, and much weaker for the 1.0″ and 1.5″ cases which had far weaker control over particle properties and visible light absorption. When the oxygen inlet position was located upstream of the electrode position, i.e., when particle nucleation was delayed until after the location of the oxygen inlet, the particles showed poor absorption in the visible regardless of input power, indicative of interstitial nitrogen doping. When the oxygen inlet was located downstream of the electrode position, the powders were brown and substitutionally doped regardless of input power, as evidenced by the absorption behavior of the powder in Fig. 6.
When the electrode was positioned upstream of the oxygen inlet, particles were anatase in nearly every case and exhibited no dual phase composition, see subfigures above the upward sloping diagonal in Fig. S6 available in the Supplemental Materials on the ASME Digital Collection. A strong power dependence of the crystal phase was witnessed for powders that were grown when the electrode was positioned downstream or at the same location as the oxygen inlet, see subfigures on and below the diagonal in Fig. S6 available in the Supplemental Materials on the ASME Digital Collection. At low powers, most particles were amorphous and grew into either an anatase phase or, in a few cases, a combined anatase/rutile phase at increasing powers while higher powers produced only anatase phase particles. We were unsuccessful in producing pure rutile particle films.
This phase transformation behavior might be due to the small crystal sizes of anatase joining at the faces which seeds the nucleation of larger rutile particles . At high powers, the rutile phase may not be achievable due to the significant nitrogen doping which acts to inhibit the formation of the rutile crystal structure . Furthermore, growing the particles in the presence of oxygen is probably key to producing small anatase particles that are capable of joining other anatase particles to produce large rutile particles. When the particles nucleate upstream of the oxygen inlet, the initial particles contain only titanium, nitrogen, and carbon species and likely crystallize only after they are exposed to oxygen. These larger particles still have significant time to oxidize within the oxygen plasma but have already nucleated to larger sizes that enable them to charge negatively and resist agglomeration within the plasma [76,77] which prevents them from joining other small particles and forming rutile phases.
The photocatalytic performance of the TiO2 nanoparticles was evaluated through photocatalytic degradation of the organic dyes methyl orange (MO) and methylene blue (MB) under solar light illumination and compared to commercial P25 powder. Both MO and MB dyes are common for benchmarking the photocatalytic efficiency of TiO2 [2,45,48,51,56,60,78–80]. While we do not expect our materials to outperform P25, which is widely seen as gold standard for photocatalytic activity, the intent of this study is to understand how the tunable properties of our materials translate into tunable photocatalytic performance.
Generally, photodegradation of MO and MB relies on the photogeneration of electron-hole pairs in TiO2 under light irradiation. The photoinduced electrons and holes participate in redox reactions with the adsorbed oxygen and water molecules to produce highly reactive oxidative species, such as superoxide anions and hydroxyl radicals, which are responsible for the degradation of the organic dyes . Figures S7(a)–S7(f) and S8(a)–S8(f) available in the Supplemental Materials on the ASME Digital Collection illustrate the gradual degradation of MO and MB dye solutions at certain time intervals for TiO2 nanoparticles produced at 10 W, 50 W, and 110 W at an oxygen inlet position of 0.5″ and electrode position of 0.5″. Also shown are results for particles produced at an electrode position of 0.5″, 1.0″, and 1.5″ and an oxygen inlet position of 0.5″. Catalyst-free photolysis performed as control experiment revealed a certain degree of self-degradation under light irradiation, Figs. S7(g) and S8(g) available in the Supplemental Materials on the ASME Digital Collection.
Figure 7 shows the degradation efficiency of the TiO2 samples produced at powers of 10, 50, and 110 W with a fixed electrode position of 0.5″ and oxygen inlet position 0.5″ compared against P25 nanoparticles. Also shown are results for samples produced at a fixed oxygen inlet position of 0.5″ with different electrode positions of 0.5″, 1.0″, and 1.5″ at a power of 50 W compared against P25 nanoparticles. The amorphous particles produced at 10 W exhibited the highest degree of photodegradation of the MO dye solution among other nitrogen-doped powders and were comparable to P25 after 240 min. The samples produced at 50 W an electrode position of 0.5″ and oxygen inlet position of 0.5″ produced the highest degree of photodegradation of the MB dye solution after 120 min when compared against nanoparticles produced at other electrode positions. However, these powders produced at varying electrode distances did not quite reach the performance of P25.
We believe that this behavior might be explained by the disordered band structure caused by the lack of a crystal structure in amorphous TiO2 particles, which likely causes more mid-gap states than in a crystalline material. This may enhance the production of photoexcited electrons and thus enhance the photocatalytic activity [8,82]. Additionally, the introduction of substitutional and interstitial nitrogen narrows the bandgap of the TiO2 nanoparticles and can also decrease the recombination efficiency of the photogenerated electron-hole pairs, which leads to the enhanced solar light catalytic activity . However, it is still controversial which mode of nitrogen doping is more beneficial with regards to the photocatalytic performance [17,60,84].
Among the amorphous samples, the sample produced at 10 W contained only interstitial nitrogen doping, showing a better degradation efficiency for MO than the other samples. The sample produced at 50 W at an electrode position and oxygen inlet position of 0.5″, which contained significant substitutional nitrogen doping along with interstitial nitrogen doping, showed the best degradation of MB, Figs. 7(a) and 7(c). The samples produced at 10 W and 50 W show comparable dye degradation with the P25 TiO2 reference. Overall dye degradation rates between these two samples are comparable. The P25 nanoparticles contain both rutile and anaphase phases which are known to increase dye degradation rates when used together [5–7]. It is currently unknown whether rutile nanoparticle additions to our samples would further increase their dye degradation rate.
Prior research on photocatalytic degradation of MB and MO of similar concentration () revealed that typically a catalyst concentration of around 0.5 g/L is required to reach an efficient degradation of 70–90% within 120–240 min [45,79]. Furthermore, additional supplements such as the addition of H2O2 or metal ion doping are usually used to enhance radical production or adjust surface charges in order to degrade different types of organic dyes [45,46]. Compared to those studies, our TiO2 catalysts show efficient degradation of MO and MB to around 30% of the initial concentration that occurs at lower concentration of nanoparticles per liter of dye solution (0.2 g/L for MO and 0.1 g/L for MB) without any additional chemical supplements or incorporation of metal ions.
Titanium dioxide nanoparticles with colors varying from white to brown with optical absorbances in the visible were produced in a single step flow through nonthermal plasma reactor. The plasma reactor design allowed for control over the position of the oxygen inlet position and exciting electrode relative to the main precursor inlet TDMAT which allowed for control over particle properties.
Results of our study are schematically summarized in Fig. 8. When the electrode was located upstream of the oxygen injection port, the resulting TiO2 particles were highly absorbing in the visible due to the substitutionally doped nitrogen, regardless of the plasma power supplied. In this configuration, particles started to grow in an oxygen deficient environment and were then oxidized in an oxygen plasma. When the electrode was located downstream of the oxygen inlet, the resulting particles had little absorption in the visible due to a lack of substitutional nitrogen which, again, was independent of applied plasma power, because particles started to grow in a well-mixed oxygen/TDMAT plasma environment. When the electrode and oxygen inlet were positioned 0.5″ downstream of the TDMAT inlet, particle properties were found to have a strong dependence on the plasma power.
Particle properties were tunable from white amorphous TiO2 with interstitially doped nitrogen to brown anatase TiO2 with substitutionally doped nitrogen. The amorphous TiO2 particles produced at 50 W were the most efficient at decomposing methylene blue while the amorphous TiO2 particles produced at 10 W were the most efficient at decomposing methyl orange and both samples were comparable to P25 dye degradation performance. The low concentrations of TiO2 particles required for MO and MB dye degradation, without using supplemental metal ions or hydrogen peroxide, make them potential candidates for photocatalytic degradation of organic dyes.
This work was primarily supported by NSF through the MRSEC program under Award DMR-1420013. Part of this work was carried out in the College of Science and Engineering Characterization Facility, University of Minnesota, which has received capital equipment funding from the NSF through the UMN MRSEC program under Award DMR-1420013. U.R.K. acknowledges partial support through the Army Research Office grant W911NF-18-1-0240. The authors would like to thank Jacob T. Held and Andre K. Mkhoyan for their advice and work regarding TEM.
Conflict of Interest
There are no conflicts of interest. This article does not include research in which human participants were involved. Informed consent not applicable.
Data Availability Statement
The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.