Continuous Generation of TiO₂ Nanoparticles by an Atmospheric Pressure Plasma-Enhanced Process

Abstract

A novel method for the continuous generation of titanium dioxide (TiO₂) nanoparticles by dielectric barrier discharge process is presented using titanium tetraisopropoxide (TTIP) and water as precursors. The aerosol generator employs an atmospheric pressure plasma enhanced nanoparticle synthesis (APPENS) process of alternative current (AC). The influences of applied voltage, frequency and precursor molar ratio on the generated particles were described by the SEM, XRD, and SMPS analyses. The results showed that TiO₂ particles appear to be in a broad size range of bi-modal distribution when no voltage is applied. While after applying the AC plasma they become uni-modal distributed with average sizes range from around 30 to 60 nm. The applied electric frequency can be adjusted to either generate nanoparticles after the plasma reactor or develop a thin film in the reactor. An increase in the precursor molar ratio leads larger particles with a broader size distribution.

1. INTRODUCTION

The TiO₂ nanoparticles have attracted many studies on their potential applications including photocatalytic destruction of various environmental pollutants, photocatalytic sterilization, photocatalytic cancer treatment, antifogging, self-cleaning, and solar cell materials (Fujishima et al. 2000; Hoffmann et al. 1995; Nakada et al. 2002). In addition to the wide application of TiO₂ nanoparticles, their implications on human health and biological impact are also of concerns. For the purposes of studying whether on the application or implication of TiO₂ nanoparticles, the continuous production of size controllable nanoparticles is demanded.

Most of the aerosol generators are approached by physical processes where the chemical composition of generated particles is the same as the precursors. For example, the vibrating orifice (Berglund and Liu 1973), evaporation-condensation aerosol generator (Veranth et al. 2003; Peineke et al. 2006), spray (Chein and Lundgren 1995) or electrospray (Chen et al. 1995) aerosol generator, modified fluidized bed (Prenni et al. 2000), gas-metal arc welding of steels (Zimmer et al. 2002), and the spark discharge particle generator (Harvath and Gangl 2003; Evans et al. 2003). The species of physically generated particles are subjected to their dissolubility or evaporation properties.

On the contrary, the chemical process in which the chemical composition of generated particles is different from the precursors is widely applied for the production of semiconductor materials, e.g., flame synthesis (Almquist and Biswas 2002; Arabi-Katbi et al. 2001; Lee and Choi 2002; Pratsinis et al. 1996; Tsantilis et al. 2002; Wang et al. 2001; Yang et al. 1996), chemical vapor deposition (Kim et al. 2005; Nakaso et al. 2003; Park et al. 2001), thermo decomposition furnace aerosol reactor (Cho and Biswas 2006; Nakaso et al. 2001; Okuyama et al. 1986; Spicer et al. 2002; Tsantilis and Pratsinis 2004a), and plasma process (Kim et al. 2003). However, the chemical methods have not been utilized as an aerosol generator due to either their high temperature or vacuum operation condition.

The atmospheric pressure plasma of dielectric barrier discharge (DBD) was developed for ozone production by Siemens in 1852. It has also been applied to the air pollutant removals (Chang et al. 1992; Lin and Bai 2001), surface treatment (Thyen et al. 1997), and the SiO₂ thin film deposition (Foest et al. 2003; Martin et al. 2004). The atmospheric pressure plasma has also been utilized for enhancing the capture of ultra-fine particles (Kulkarni et al. 2002). In the authors' previous studies, uniform N-doped TiO₂ particles were generated by the DBD technology and successfully acted as visible light photocatalysts, it was denoted as the Atmospheric Pressure Plasma Enhanced Nanoparticle Synthesis (APPENS) process (Bai et al. 2004; Chen et al. 2007). However, the effects of the operation parameters on the generated particle size distribution have not been investigated and therefore their potential as a continuous aerosol generator has not been clarified.

In this study, the operation parameters of APPENS process which may influence the particle morphology, sizes, and distributions are studied. These include the applied voltage, frequency, and the precursor molar ratio employed in the APPENS process. And possibility of the process as a chemical-approached aerosol generator for the continuous generation of TiO₂ nanoparticles (< 100 nm) is demonstrated.

2. EXPERIMENTAL METHOD

The APPENS process for generating TiO₂ nanoparticles is sketched in Figure 1. The aerosol reactor was a wire-tube type DBD reactor operated at atmospheric pressure and low temperature conditions. Titanium tetraisopropoxide (TTIP, Ti(OC₃H₇)₄) and H₂O were used as the precursor solutions for generating TiO₂ nanoparticles in this study. The precursors were vaporized and sent into the plasma reactor via passing the carrier gas through two series of impingers, they were controlled at temperatures of 150–200°C and 25°C, respectively for TTIP and water. The carrier gas was N₂ with flow rates being controlled by mass flow controllers (Brooks, 5860E), and the total flow rate of precursor vapors was 137.4 sccm. The plasma reactor was made of Pyrex glass with 21 mm I.D., 23 mm O.D., and 200 mm in length. One electrode was a stainless steel rod of 2 mm in diameter positioned along the central line of the reactor. The other electrode was a sheet of stainless steel mesh wrapped around the reactor with wrapping length of 135 mm.

FIG. 1 Schematic diagram of APPENS reactor for producing TiO₂ nanoparticles.

The applied voltage, frequency and the precursor molar ratio were the controlling factors of particle size distribution generated by the APPENS process. The applied voltage was adjusted at the range of 0–10.20 kV, the frequency of alternative current (AC) was either 60, 120, or 240 Hz, and the precursor molar ratio was controlled by TTIP/H₂O volumetric flow rate ratio of 0.125 to 8. The corresponded TTIP/H₂O molar ratios were calculated from the vapor pressure of TTIP and H₂O to be 0.4 to 76.1 based on the equilibrium equation (Duminica et al. 2004) at a certain TTIP bubbling temperature,

where P_TTIP is the TTIP vapor pressure (Torr), and T is the TTIP bubbling temperature (°K).

The nanoparticles deposited on the reactor wall were collected for further analyses by SEM (Scanning Electron Microscope; HITACHI S4700) and XRPD (X-ray Powder Diffractometer; Rigaku; Cu target, 30 kV, 20 mA). The geometric mean diameter (Dp_g) and geometric standard deviation (σ_g), were determined based on averaging over 80 particles as observed from the SEM images. And for on-line measurement of particle size distribution, a SMPS (Scanning Mobility Particle Sizer, TSI model 3080L) was connected after the APPENS reactor to determine the Dp_g, σ_g, and the total mass concentration (TMC) of generated particles. In order to achieve the minimum required flow rate of the SMPS, a N₂ dilution gas stream was introduced after the plasma reactor with a N₂/aerosol dilution flow rate ratio of 1.0 (i.e., the inlet flow rate to the SMPS is double of the reactor outlet flow rate). And to further prevent agglomeration of the TiO₂ particles, a N₂/aerosol dilution flow rate ratio of 12.0 was sometimes introduced (i.e., the inlet flow rate to the SMPS is 13 times of the reactor outlet flow rate).

3. RESULTS AND DISCUSSION

3.1 Effect of Applied Voltage

For the study of the effect of applied voltage on the size distribution of generated particles, the TTIP/H₂O precursor molar ratio was 11.9 as bubbled at the temperature of 150°C and the frequency was 60 Hz in the aerosol generator. Figure 2a, Figure 2b, Figure 2c, Figure 2d show the SEM photo images of the particles generated at applied voltage of 0, 6.48, 8.58, and 9.60 kV, respectively. It was observed that the generated particles were in a broad size distribution as no voltage was applied, the particle sizes could be as large as 1 μ m or smaller than 100 nm. On the other hand, as the voltage was applied the particle sizes became smaller and more uniformly distributed.

FIG. 2 SEM images of TiO₂ particles collected in the reactor under different values of applied voltage. (a) No applied voltage. (b) 6.48 kV. (c) 8.58 kV. (d) 9.60 kV. The frequency was 60 Hz, the TTIP/H₂O precursors molar ratio was 11.9 as TTIP bubbled at 150°C. The scale bar in the SEM photo is 1 μ m. (Continued)

The SMPS on-line measurement data for the particle size distribution are shown in Figure 3. The experimental conditions for the APPENS reactor were the same as that shown in Figure 2. As no voltage was applied, the TiO₂ particles show bi-modal distribution with two mode sizes at around 106 nm and 372 nm, respectively. And the particle size appears to be uni-modal distributed as the voltage was applied. The mode of particle size decreases from 163 nm to 27.9 nm as increasing the applied voltage from 6.48 to 8.58 kV. But further increases in the values of applied voltage to 9.60 and 10.20 kV result in increases in the mode size instead. The lognormal fittings (dashed-lines) seem to obtain good correlation with the SMPS online measurement data (step-lines) for the plasma produced particles. The calculated values of σ_g ranged from 1.43 to 1.83 for particles generated at the applied voltage of 6.48 to 10.20 kV.

FIG. 3 Particle size distributions measured by SMPS after dilution ratio of 1. The operation condition was the same as that of Figure 2. Step-lines are SMPS measured data obtained at different applied voltages, and dash-lines are their lognormal fittings.

Values of Dp_g of generated particles at applied voltage of 6.48, 8.58, 9.60, and 10.20 kV were also determined from the SEM images and they were 75, 31, 39, and 43 nm, respectively, with their corresponding values of σ_g of 1.5, 1.2, 1.4, and 1.3. By comparing values of Dp_g and σ_g determined from the SEM images to those from SMPS data, one can see that particle sizes measured by the SMPS were larger and the distribution were broader. This is opposite to those predicted by the diffusion and electrostatic mechanisms that particles deposited on the reactor wall should have broader size distribution than those penetrating the reactor. The larger Dp_g and σ_g of particles penetrating the reactor may be mainly caused by that loosely agglomerated particles were counted by the SMPS as one single particle. This is confirmed by that values of σ_g measured by the SMPS (shown in Figure 3) are larger than the value predicted by the self preserving size distribution theory for coagulating spheres, 1.33 (Vemury et al. 1994). Larger values of σ_g are typically the result of particle agglomeration without coalescence in the gas phase.

Nakaso et al. (2003) indicated that faster reaction rate of TTIP conversion led to narrow particle size distribution. In this study, the TTIP conversion rate is enhanced by applying the plasma in this system. In addition, the non-thermal plasma kept the gas temperatures to be low within and after the aerosol generator. This particular condition is similar to the effect of cooling gas which can restrain the coalescence between the particles (Joshi et al. 1990; Tsantilis and Pratsinis 2004b). On the contrary of without applying any voltage or at low applied voltage, the bi-modal size distribution or broad distribution may be due to the slow reaction rate of precursors or particles (Nakaso et al. 2003; Kuster and Pratsinis 1995; Ahonen et al. 1999). Thus the application of plasma not only avoids the further coalescence of the agglomerated particles, but also enhances the hydration rate of the precursors. But if the applied voltage is increased to be too strong (∼ 10.2 kV), the melting possibility of smaller particles in the agglomerates is increased. This leads to larger product particles or hard agglomerates of small particles. Hence the optimal applied voltage is 9.6 kV for better production of catalytic particles with nanosize distribution, high production yield and less agglomeration.

3.2 Effect of Electric Frequency

The effects of electric frequency at 60, 120, and 240 Hz on the morphology and size distribution of particles were investigated by the SEM and SMPS analyses and results are shown in Figure 4. The applied voltage was 10.20 kV and the TTIP/H₂O precursor molar ratio was 11.9 by bubbling of the TTIP at 150°C. A very interesting phenomenon was observed from the SMPS data that as increasing the frequency from 60 Hz to 240 Hz, the particle concentration decreases significantly. The SMPS distributions in Figure 4 for the 120 Hz and 240 Hz are so low in concentration that they may not be representative of the actual size distribution of the particles produced.

FIG. 4 Particle size distributions measured by SMPS after dilution ratio of 12 under different electric frequencies. Also shown are SEM images of the reactor-deposited particles. The applied voltage was 10.20 kV, the TTIP/H₂O precursor molar ratio was 11.9 as TTIP bubbled at 150°C.

The SEM morphology of particles deposited in the reactor which generated at 240 Hz was very different from those at 60 and 120 Hz as seen from Figure 4. The particles seemed to develop a film at 240 Hz. The approximate sizes of particles from SEM images for 60 and 120 Hz are 43 and 34 nm, respectively. But the particle sizes of 240 Hz were difficult to calculate from this SEM image due to its film structure. This result is similar to literature data on the SiO₂ thin film (Thyen et al. 1997; Foest et al. 2003) deposited at high frequency (20–100 KHz). Therefore, most of the precursor vapors deposited in the plasma reactor and formed a thin film at frequency of 240 Hz. This may explain why a lower particle concentration was measured after the reactor by SMPS as the applied frequency was increased. As a result, the APPENS could work as an aerosol generator at lower frequency of 60 Hz. And it can also be applied for producing thin film at higher frequency of 240 Hz. But for a frequency of 120 Hz, the SMPS detected too low particle concentration while the SEM did not observe a nice film developed so it maybe applied for neither purpose.

3.3 Effect of Precursor Molar Ratio

The crystalline phase of generated particles was determined with the XRPD pattern and the results are shown in Figure 5. The XRPD peaks at around 25°, 38°, 48°, 55°, and 63° corresponding to anatase crystal are observed for all synthesized samples, with the most significant peak at around 25°. And the peak intensities at all angles seem to be increased with increasing the molar ratio. The effect of TTIP/H₂O precursor molar ratios on the crystalline sizes of particles was estimated from XRPD data by Scherrer equation (Musić et al. 1997),

where D (nm) is the crystalline size, λ (nm) is X-ray wavelength for the XRPD tests, θ is Bragg angle, and β is full width of diffraction at half of maximum intensity (FWHM). The calculated crystalline sizes are listed in Table 1. As seen in Table 1, the average crystalline size of particles generated at TTIP bubbling temperature of 200° is around 23.5 ± 2.4 nm, which is slightly larger than the value of 22.2 ± 1.1 nm at 150°C. And the average of all data is 22.8 ± 2.0 nm. The data are within the experimental error and thus one can say that the effect of molar ratio on crystalline sizes of particles is not significant.

FIG. 5 The XRPD patterns of generated particles synthesized in this study. The operation condition was the same as that of Table 1 with TTIP/H₂O precursor molar ratio (MR) from 0.4 to 76.1 and TTIP bubbling temperature of 150 or 200°C.

TABLE 1 Crystalline size of TiO₂ particles generated via different TTIP/H₂O precursor molar ratios. The crystalline sizes of primary particle were calculated based on Scherrer equation with parameter values obtained from XRPD patterns at X-ray wavelength (λ) of 0.154056 nm.

TTIP bubbling temperature (°C)	Flow ratio (TTIP/H2O)	Molar ratio (TTIP/H2O)	FWHM	2θ (degree)	Crystalline size (nm)
150	8	23.9	0.0064	25.33	22.2
150	2	6.0	0.0067	25.33	21.3
150	1	3.0	0.0065	25.28	21.8
150	0.125	0.4	0.0061	25.36	23.3
150	Average size	–	–	–	22.2 ± 1.1
200	4	76.1	0.0067	25.31	21.1
200	1	19.0	0.0057	25.29	24.8
200	0.5	9.5	0.0058	25.30	24.5
200	0.25	4.8	0.0061	25.35	23.5
200	Average size	–	–	–	23.5 ± 2.4
	Total average size	–	–	–	22.8 ± 2.0

The primary particle can be composed of more than one crystal or can be partly crystalline and contain some amorphous TiO₂. The effect of precursor molar ratio on the primary particle size is also evaluated via averaging over 80 particles in the SEM imagines for each test condition. The results are shown in Figure 6 with symbols indicate the SEM averaged size distribution data and lines are logarithmic regression results of Dp_g and σ_g with respect to the precursor molar ratio.

FIG. 6 Effect of precursor molar ratio on the size distributions of particles collected in the reactor and measured by SEM images. The applied voltage was 9.60 kV and the TTIP bubbling temperatures were at 150 or 200°C.

One can see in Figure 6 that decreasing the precursor molar ratio results in a smaller and uniform particle size. Based on the regression curves, Dp_g of particles is decreased from around 60 nm to 30 nm and σ_g of particle distribution is decreased from 1.6 to 1.2 as decreasing the TTIP/H₂O precursor molar ratio from 76.1 to 0.4. The precursor molar ratio has a slightly stronger effect on Dp_g at TTIP bubbling temperature of 200°C. But values of σ_g for data obtained at 200°C do not seem to be increased further as increasing the TTIP/H₂O molar ratio. It must be noted that at very low TTIP/H₂O precursor molar ratio the generated particle number concentration is much lower. Hence the decreasing of σ_g to monodisperse aerosol by decreasing TTIP/H₂O precursor molar ratio is accompanied by that less particles are generated. This may not be acceptable for some particle generation application.

To further verify the precursor effect on the particle size distribution, TEM images were also taken and shown in Figure 7(a) and Figure 7(b), respectively, for TTIP/H₂O precursor molar ratios of 1.1 and 76.1. The TEM images reveal similar trend to the SEM observed data shown in Figure 6 that increasing the precursor concentration leads to a broader size distribution. At a low TTIP/H₂O precursor molar ratio of 1.1, the particle sizes are near uniform with size of around 30–40 nm. From Figure 6 one knows that at TTIP/H₂O precursor molar ratio of 1.1 and TTIP bubbling temperature of 150°, value of Dp_g is around 35 nm and σ_g is around 1.2. Based on the lognormal distribution, 95% of the particles for such a size distribution should fall within 29 and 42 nm. This size range corresponds well with the TEM observation shown in Figure 7(a). But at a high TTIP/H₂O precursor molar ratio of 76.1, the TEM observed a much broader particle size distribution with particle size ranges from around 15 to 80 nm.

FIG. 7 TEM images (scale bar = 20 nm) of particles collected in the reactor as obtained at two different TTIP/H₂O precursor ratios of 1.1 and 76.1, respectively. The applied voltage was 9.60 kV and the TTIP bubbling temperatures were at 150 and 200°C, respectively. The low TTIP/H₂O precursor ratio of 1.1 clearly indicates higher uniformity of TiO₂ particles.

The observation that higher precursor concentration leads to higher particle size distribution seems to be opposite to the TiO₂ thermo-decomposition result of Pratsinis and Spicer (1998), where surface reaction is the dominating mechanism for the change of particle size distribution. This can be explained by that the hydrolysis reaction of TTIP with H₂O is much faster than the thermo-decomposition rate of TTIP (Okuyama et al. 1990; Seto et al. 1995). Thus hydrolysis reaction is the dominating mechanism for the formation of TiO₂ particles at low TTIP/H₂O molar ratio, thus the particle size distribution is narrower. On the other hand, at high TTIP/H₂O molar ratio (above the stoichiometric ratio of hydrolysis reaction) the formation of TiO₂ particles was first induced by hydrolysis reaction, then excess of TTIP forms TiO₂ monomers by thermo-decomposition mechanism, and the monomers either nucleate as new particles or condense onto the TiO₂ particles which were previously formed by hydrolysis reaction. As a result, the particle size distribution is broader at high TTIP/H₂O precursor molar ratio due to both hydrolysis and thermo-decomposition reactions occurred for the formation of TiO₂ particles.

3.4 Long-Term Stability

The size distributions of particles generated from APPENS reactor were measured continuously by SMPS for 1 hr to test the long-term stability of the aerosol generator, and variations of Dp_g and σ_g with respect to time are shown in Figure 8. The generator was operated under conditions of TTIP/H₂O precursor molar ratio of 0.7 (with TTIP bubbled at 150°C), applied voltage of 9.60 kV, frequency of 60 Hz and without any dilution. The average value of Dp_g of generated particles is 40.7 ± 0.5 and σ_g of them is 1.43 ± 0.1. The result indicated that the particle size distribution is stable and the APPENS generator can be used for long-term operation.

FIG. 8 Stability of geometric mean particle diameter (Dp_g) and geometric standard deviation (σ _g) measured by SMPS during one hour of operation time. The frequency was 60 Hz, the TTIP/H₂O molar ratio was 0.7 as TTIP bubbled at 150°C and the applied voltage was 9.60 kV.

3.5 Comparison of APPENS Generator with Commercial Aerosol Generators

Table 2 compares the properties of particles generated via the APPENS generator to the furnace reactor as well as to commercial nano-size range (< 100 nm) aerosol generators. Both the APPENS and the furnace reactor produce particles via chemical process of gas to particle conversion. The furnace reactor produces TiO₂ particles at high temperature via thermo-decomposition mechanism and the particle number concentration is very high (10¹⁰ #/cm³) with σ_g of around 1.3–1.5. The APPENS generator is operated at room temperature and it produces TiO₂ particles via hydrolysis and thermo-decomposition mechanisms. It is capable of generating particles with number concentration of around 10⁶ #/cm³ and σ_g of around 1.2 to 1.5.

TABLE 2 Comparison of the APPENS generator with commercial aerosol generators. Only nano-size (< 100 nm) aerosol generators are included

Aerosol generator	Particle Generation Mechanism	Size range (nm)	σg	Total concentration (#/cm^3)	Morphology/Component of generated particles
This study (APPENS)	Nonthermal plasma/ hydrolysis& thermal decomposition	< 10–100 (with controllable average size ranges from 30–60)	1.2–1.5	10^6	Loose agglomerates of spherical primary particles (solid phase)/ TiO2 at the present time
Furnace reactor (Nakaso et al. 2001)	High temperature flame/ Thermal decomposition	12–17 ^c	1.3–1.5	10^10	Spherical/ TiO2
		4–8 ^d	1.3–1.5	10^10–10^11
Electrospray Aerosol Generator; 3480 ^a	Electrical field spray	2–100	1.1	10^7	Spherical (liquid phase)/ Soluble solids or nonvolatile liquids
Aerosol generator; GFG-1000 ^b	Spark	20–150	> 1.6	10^7	Nature soot liked/ Graphite, metals
Constant Output Atomizer; 3076, 3079, 9302 & 9306A ^a.	High velocity jet atomizing	10–2,000	> 1.6	10^7	Spherical/ PSL ^e , DOP ^f , oils, and other aqueous or alcohol solutions or suspensions
		10–2,000	> 1.6	10^8	Spherical/ DEHS ^g , DOP, Emery 3004, Paraffines, PSL, Salt solutions
		10–2,000	> 1.6	10^7	Spherical/ Water-soluble materials (salt or sugar)
		10–2,000	< 2.0	10^7	Spherical/ PSL, DOP, DEHS, oils, and other aqueous or alcohol solutions or suspensions

As for the commercial aerosol generators, the electrospray aerosol generator is a monodisperse nanoparticle generator, while the atomizer and the spark aerosol generators generate polydisperse particles. The generated particles from both atomizer and spark aerosol generator (GFG-1000) are in a broad size range (σ_g > 1.6) and thus classifiers are needed to produce monodisperse nanoparticles. the electrospray aerosol generator produces particles from the charged solution through the capillary by electrical field pulling. The atomizer atomizes liquid particles through an orifice by high-velocity jet. The spark aerosol generator (GFG-1000) uses high voltage to generate sparks between two graphite electrodes and forms graphite fine particles by evaporation-condensation mechanism. The morphologies of generated solid particles were spherical via APPENS and furnace reactor. The electrospray and atomizer can also generate spherical liquid aerosol, but they are difficult to generate solid spherical particles. The spark aerosol generator (GFG-1000) can generate either metals or graphite particles.

4. CONCLUSION

The effects of applied voltage, frequency, and the precursor molar ratio on particle morphology and particle size distributions were described in this study. The APPENS generator can produce nanoparticles of uni-modal size distributions. It is concluded that the APPENS process has a high potential to be used as a size controllable monodisperse or polydisperse aerosol generator. The APPENS generator is simple to be composed of and it can be scaled up by using several more parallel tubes to enlarge the production rate (Lin and Bai 2001) as already approached for a commercial O₃ generator. As a result, the APPENS process could be utilized for the purpose of either implication or application studies of TiO₂ nanoparticles. Further research work might also be extended to generate other chemical components of nanoparticles such as SiO₂, ZnO, and MgO.

Acknowledgments

Hsunling Bai and Chienchih Chen acknowledge the support of National Science Council, Taiwan through grant numbers NSC 93-2211-E-009-018 and NSC 94-2211-E-009-004.

Notes

^a TSI Inc.

^b PALAS® Aerosltechnologie aerosol technology.

^c TTIP precursor.

^d TiCl₄ precursor.

^e PSL: Polystyrene latex.

^f DOP: Di-octyl phthalate.

^g DEHS: Di-ethyl-hexyl-sebacin acid.