Size dependence of thermal stability of TiO 2 nanoparticles

Transcription

1 JOURNAL OF APPLIED PHYSICS VOLUME 96, NUMBER 11 1 DECEMBER 2004 Size dependence of thermal stability of TiO 2 nanoparticles W. Li and C. Ni Department of Materials Science and Engineering, University of Delaware, Newark, Delaware H. Lin and C. P. Huang Department of Civil and Environmental Engineering, University of Delaware, Newark, Delaware S. Ismat Shah a) Department of Materials Science and Engineering, University of Delaware, Newark, Delaware and Department of Physics and Astronomy, University of Delaware, Newark, Delaware (Received 2 July 2004; accepted 24 August 2004) Anatase TiO 2 nanoparticles with average particle size ranging between 12 and 23 nm were synthesized by metalorganic chemical vapor deposition. The structure and particle size were determined by x-ray diffraction and transmission electron microscopy. The specific surface areas were measured by Brunauer-Emmett-Teller and ranged from 65 to 125 m 2 /g. The size effects on the stability of TiO 2 in the air were studied by x-ray diffraction and transmission electron diffraction for isochronally annealed samples in the temperature range of C. Only anatase to rutile phase transformation occurred. With the decrease of initial particle size the onset transition temperature was decreased. An increased lattice compression of anatase with the raising of temperature was observed by the x-ray peak shifts. Larger distortions existed in samples with smaller particle size. The calculated activation energy for phase transformation decreased from 299 to 180 kj/mol with the decrease of initial anatase particle size from 23 to 12 nm. The decreased thermal stability in finer nanoparticles was primarily due to the reduced activation energy as the size related surface enthalpy and stress energy increased American Institute of Physics. [DOI: / ] I. INTRODUCTION TiO 2 has three main crystalline structures: anatase (tetragonal I4/ amd), brookite (orthorhombic Pcab), and rutile (tetragonal P4 2 /mnm). The total free energies have to be minimized for these structures to be thermodynamically stable. 1 Rutile is usually stable at high temperatures. Different structures lead to different physical properties which, in turn, lead to different applications. For instance, anatase is commonly employed for photocatalysis because of its high photoreactivity 2 and rutile is largely used for pigments due to its effective light scattering. 3 Therefore, understanding of the mechanism and the factors affecting phase stability and phase transformation is important to design and controllably manipulate phase types and concentrations for more efficient uses. With the increase of temperature, the possible phase transformation sequences among three TiO 2 main structures can be anatase to rutile, anatase to brookite to rutile, brookite to rutile, and brookite to anatase to rutile. The actual transformation behavior is determined by initial particle size, 4 7 impurity content, 8 12 starting phase, 5,13,14 reaction atmosphere, 10,15,16 temperature, pressure, etc. As nanosized TiO 2 becomes more desirable due to its unique electronic, optical, and photocatalytic properties, initial size inevitably becomes crucial in the phase evolution. It is known that the driving force for phase transformation is the difference between chemical potential of the initial and final phases. Activation energy is the minimum energy requirement for overcoming the energy barrier for phase transformation between a) Electronic mail: ismat@udel.edu the two phases. In nanoparticles, surfaces non-negligibly contribute to the chemical potential by creating surface free energy and surface stress. Thus, varying initial size of TiO 2 nanoparticles alters the phase transformation reaction path, including transformation type, transition onset point, phase nucleation, and growth rate, etc. Moreover, the thermal behavior might also not be identical in TiO 2 nanoparticles with the same initial size but synthesized via different methods. It is believed that the variations of surface characteristics caused by the preparation methods can cause the formation of different number of nucleation sites which change the growth rate of the second phase, or even the transformation sequence. 14,17,18 There are few previous reports investigating the role of particle size on the thermal stability of TiO 2 nanoparticles. Banfield and co-workers theoretically and experimentally investigated particle size dependence of phase stability and demonstrated that the macrocrystalline rutile is more stable than macrocrystalline anatase, but the stability reverses when particle size becomes less than 14 nm. 4,7 Zhang and Banfield studied initial size effects on the transformation behavior of a mixture of anatase (5.1 nm) and brookite phases (8.1 nm) and found that first anatase transforms to brookite and/or rutile and eventually the whole mixture becomes rutile. 5 However, So et al. 14 reported that the phase transformation occurred from brookite to anatase, and eventually to rutile. The above studies are based on the TiO 2 particles prepared by sol-gel process. Therefore, further studies are still necessary in order to get a better understanding of the impact of size on the phase transformation, especially for TiO 2 nanoparticles synthesized by various other methods /2004/96(11)/6663/6/$ American Institute of Physics

2 6664 J. Appl. Phys., Vol. 96, No. 11, 1 December 2004 Li et al. We use metalorganic chemical vapor deposition (MOCVD) method to prepare pure polycrystalline anatase TiO 2 nanoparticles with different sizes by varying O 2 flow rate, growth pressure, and temperature. The nanoparticles with sizes 12±2, 17±2, and 23±2 are selected to investigate the thermal behavior of anatase in air. The objectives of the study are to determine phase transformation type and the onset temperature of the phase transformation as a function of particle size. The lattice changes of anatase are also quantified during the phase evolution. The surface energy, surface stress, and activation energy, which are the important factors to study the phase transformations, are evaluated to reveal the effect of these parameters on the thermal stability of anatase nanoparticles. II. EXPERIMENT MOCVD method was used to prepare TiO 2 nanoparticles with different average particle sizes. Details of the process are given elsewhere All particles were collected on 5 cm diameter disks made of several layers of stainless steel screen mesh, which were designed for high collection efficiency. Prior to the deposition, the screen was cleaned by acetone, methanol, and de-ionized water to reduce any native contamination on the surface. The screen was then dried by compressed air. A 3 cm diameter and 12.5 cm long quartz tube was used to hold substrate perpendicular to the flow direction in the central part of reaction chamber. After the system was pumped down to 0.5 Pa by a mechanical pump, the reactor with the screen was baked at 500 C with 665 Pa oxygen for 15 min to further remove residual contaminants, including moisture that could cause unexpected hydrolysis. For deposition, the substrate temperature was raised to 600 C. In order to obtain particles with different sizes, O 2 flow rate was varied. In this experiment, O 2 (99.999%) flow rates applied were 20, 25, and 35 standard cubic centimeter per minute, at a constant O 2 pressure of 1.33 kpa, for obtaining particles with sizes of 23, 17, and 12 nm, respectively. The flow rate and pressure was measured at the inlet and outlet of the chamber, respectively. For all depositions, TTIP (titanium tetraisoproxide, Aldrich, 97%) was used as Ti precursor. The temperature of the precursor bath was kept at 220 C. Ar (99.999%) was used as the carrier gas and the total initial deposition pressure was about 2 kpa. After deposition, the reactor chamber was cooled down to room temperature in an atmosphere of 665 Pa oxygen to prevent surface contaminations. The deposited nanoparticles were then removed from the screen for annealing and characterization. Isochronal annealings were carried out to investigate the phase transformation behavior. For annealing, each asdeposited sample was divided into several portions. The annealing temperatures ranged between 700 and 800 C. Annealing experiments were carried out in air for 1hina1200 W box furnace. The structure and particle size analysis were done by x-ray diffraction (XRD) and transmission electron microscopy (TEM). Using a Rigaku D-Max B diffractometer, XRD 2 scans were recorded with Cu K radiation and a FIG. 1. XRD patterns for TiO 2 nanoparticles with average sizes of 12, 17, and23nm. graphite crystal monochromator (K 1 = Å and K 2 = Å). XRD patterns were obtained for all asdeposited and annealed samples. Full scans were obtained in the range of 20 to 60 2 which includes most major anatase and rutile diffraction peaks. The scan speeds and point intervals were kept at 1 /min and 0.04, respectively /min low scan speed and short point interval were used to obtain high-resolution anatase (101) and (200), as well as rutile (110) peaks, for peak position and peak area measurements. XFIT (Ref. 22) software was used for peak fitting. Particle sizes were estimated by Scherrer s formula 23 based on the precise 2 positions and full width at half maxima of the diffraction peaks. Anatase lattice changes were confirmed from the (101) and (200) peak shifts. A JEOL 2000 FX TEM was employed to obtain bright field images, dark field images, and diffraction patterns of the TiO 2 nanoparticles. The acceleration voltage was 200 kv. Particle sizes were measured from both bright and dark field images. Crystal structures were studied using the electron diffraction patterns. Specific surface area was measured by Brunauer- Emmett-Teller (BET) method. The BET measurements were performed in a NOVA2000 gas sorption analyzer (Quantachrome Corporation). Before the sorption analysis, the samples were degassed and calcinated at 100 C for 24 h. N 2 was chosen as the adsorbate for isothermal adsorption. The relative pressure P/ P 0 used was within the range of P 0 is the gas pressure required for saturation at the temperature of experiment and P is the pressure of the gas at the equilibrium with a solid. The average pore radius was also recorded. III. RESULTS Figure 1 shows the XRD patterns of as-deposited TiO 2 nanoparticles with different grain sizes and the highresolution (101) peak scans for these samples (inset). All peaks belong to the anatase phase and no other phase is detected within the x-ray detection limit. The measured particle sizes were 12±2, 17±2, and 23±2 nm. TEM observations also confirmed that the particles were polycrystalline anatase with the average sizes in good agreement with the

3 J. Appl. Phys., Vol. 96, No. 11, 1 December 2004 Li et al FIG. 2. Electron diffraction patterns, bright field images, and dark field images for 12 nm sample with conditions of as-deposited (a) (c), annealed at 700 C with 1 h(d) (f); and annealed at 800 C and 1 h (g) (i), respectively. XRD results. Figure 2 shows the diffraction pattern (a), the bright field image (b), and dark field image (c) for asdeposited 12 nm sample. As the surface area of nanoparticles has non-negligible contribution for the phase evolution, particles with same average size but different surface area may have different thermal behavior. Thus, the surface areas of these samples must be measured. Although surface area is directly related to the particle size, agglomeration decreases the effective surface area. The surface areas of the different sized samples measured by BET are shown in Table I. Commercially available standard Degussa P25 TiO 2 that has an average size of about 27 nm was also characterized for comparison. The measured areas were 125, 95, 65, and 60 m 2 /g for 12, 17, 23 nm particles, and P25, respectively. The area of 23 nm sample is slightly larger than that of P25. In order to study the phase transformation behavior, the particle growth and phase concentrations at various annealing temperatures (including starting anatase nanoparticles) were qualitatively determined by TEM and XRD. Figures 2(a) 2(i) are diffraction patterns, bright field, and dark field TABLE I. BET specific surface areas of 12 nm, 17 nm, 23 nm as-deposited particles, and Degussa P25. Samples 12 nm 17 nm 23 nm P25 Surface areas ±5 m 2 /g images for as-deposited, 700 C, and 800 C annealed 12 nm samples, respectively. By the comparison of bright field and dark field images, it is observed that the particle sizes become larger and nonuniform with the increase of the annealing temperature. Such nonuniformity was undetectable by XRD measurements. The nucleation and growth of the rutile phase in the matrix of anatase at 700 C and the phase transformation is reflected in the difference of the diffraction patterns [Figs. 2(a), 2(d), and 2(g)]. However, when the two phases coexist, the particle sizes of anatase and rutile, as well as relative phase content, are difficult to be statistically determined from these TEM images. Thus, XRD patterns were used for quantitative evaluation. Figures 3(a) and 3(b) are the diffraction patterns of 12 and 23 nm samples, respectively. XRD patterns from as-deposited samples and samples annealed at 700, 750, and 800 C are presented. The phase composition is calculated by the following formula: 4,5 W R = A R A R =, A A A + A R where W R is rutile weight percent, A A and A R are integrated diffraction peak intensities from anatase (101) and rutile (110), respectively, and A 0 is the total integrated (101) and (110) peak intensity. These transformations have broad temperature ranges and the relative concentrations of rutile at various temperatures are listed in Table II. It can be seen that the sample with finer initial anatase particles has lower ther-

4 6666 J. Appl. Phys., Vol. 96, No. 11, 1 December 2004 Li et al. FIG. 4. Changes in particle sizes of anatase and rutile phases as a function of the annealing temperatures. FIG. 3. Comparison of XRD patterns for (a) 12 nm and (b) 23 nm TiO 2 : as-deposited and annealed at 700, 750, and 800 C for 1 h. mal stability and lower transition onset temperature. For example, at 700 C, the 12 nm sample has 7.9% of rutile whereas the 23 nm has only 1.3% rutile. For the comparison of particle size growth rates of the two phases, the particle size as a function of annealing temperature is plotted in Fig. 4. Both anatase and rutile particle sizes increase with the increase of temperature, but the growth rate is different. The rutile has much higher growth rate than anatase. The growth rate of anatase levels off at 800 C. As expected, the rutile, after nucleation, grows rapidly whereas anatase particle size remains practically unchanged. Moreover, the difference in the slopes shows that both anatase and rutile growth rates of 12 nm sample are much faster than that of the 23 nm sample. IV. DISCUSSION TABLE II. Variation of the relative area of the rutile (101) XRD peak. Samples A R /A 0 ±0.03 As-deposited 700 C 750 C 800 C 12 nm nm nm Based on the results described in the preceding section, polycrystalline anatase nanoparticles synthesized by MOCVD directly transform to rutile in the air at high temperature. Smaller size nanoparticles have lower thermal stability. This behavior is different from the high pressure TiO 2 phase transformations which include the appearance of other structures such as PbO 2 and ZrO 2 -type phases, as previously reported. 24,25 All of our phase transformation experiments were carried out under normal pressure (1 atm). For solid particles, the total free energy is determined by volume energy G v, surface energy G s, and surface stress induced energy G f. The last two terms are particularly larger in the nanoparticles compared to those of bulk in which they can be ignored. The surface energy is from the surface reconstruction with the appearance of large surface area and the BET data (Table I) indicate the importance of this surface energy. In solid state, the surface stress is different from surface free energy because of the low mobility of atoms. Therefore, the system Gibbs free energy of the nanoparticles for anatase to rutile phase transformation is usually expressed as G A R = G v,r T G v,a T + A R R A A A + P R V R P A V A = G v,r T G v,a T + 3M R R r R + 2Mf R 2Mf A, R r R A r A 3M A A r A where, subscripts A and R stand for anatase and rutile, respectively. A, V, M, r, and are surface area, molar volume, molecular weight, particle radius, and phase density, respectively. is the surface free energy and P is surface stress f related excess pressure with a value of 2f /r. Only negative G A R is a potential driving force for anatase to rutile transition. From the equation, anatase energies from both surface and stress contribute negative values to the total energy difference between rutile and anatase. Thus, the higher the anatase surface energy and surface stress energy, the more the contribution to the nonequilibrium transition from anatase to rutile. Particles with smaller sizes have larger surface areas. Therefore, in anatase with smaller size, it is easier to start the phase transformation than in the larger particles under similar conditions. The higher negative G A R value corresponds to the lower activation energy needed for ana-

5 J. Appl. Phys., Vol. 96, No. 11, 1 December 2004 Li et al FIG. 5. Arrenhius plot of ln A R /A 0 vs 1/T for activation energy calculations as a function of the size of TiO 2 nanoparticles. tase to rutile transition. This activation energy was quantified from the experimental results. Assuming the transformation is a first-order reaction, 9,10 the activation energies can be calculated based on the results from Table II and Fig. 5 which is an Arrenhius plot of ln A R /A 0 vs 1/T, where T is absolute temperature. The values for the activation energies obtained are 299, 236, and 180 kj/mol for 23, 17, and 12 nm TiO 2 particles, respectively. Also, in Fig. 5, the values of the region of fitting (R) are given as 0.991, 0.997, and for the 23, 17, and 12 nm samples, respectively, which confirm the appropriation of the assumption of the first-order transformation. It is clear that the calculated activation energies from experiment as a function of particle sizes are in agreement with the qualitative analysis from above thermodynamics equation, which reveals the size effects on the phase stability. The activation energy for the phase transformation includes a lattice strain component. From the XRD peak positions of anatase (101) and (200), large angular peak shifts were observed with the increase of temperature and particle size. Figures 6(a) and 6(b) show the anatase (101) XRD peak positions at various temperatures for 12 nm and 23 nm samples, respectively. The peaks from K 1 and K 2 were resolved. The positions of (101) and (200) K 1 peaks and the calculated a and c lattice constants are listed in Table III. With the increase of temperature, all (101) and (200) peaks shift to higher angles. The lattice constants of tetragonal anatase were calculated from these peaks. The a values have almost no change whereas the c values have a large reduction. The tendency of c-direction change is due to the large space and distance along c axis in the tetragonal unit cell. Above a certain higher temperature, the change of c value slows down. Furthermore, with the increase of annealing temperature the reduction of c values is more evident in asdeposited 12 nm sample than that of 23 nm sample or larger changes in the c values occur in smaller size sample. For example, at 800 C, there is 0.15 Å reduction of c from initial value of Å in the 12 nm sample compared with 0.06 Å change from an initial value of Å in the 23 nm sample. These changes indicate that the anatase lattice is being compressed during the phase transformation. In other FIG. 6. Shifts in the XRD peak positions of anatase (101) peaks for (a) 12 nm and (b) 23 nm samples. word, stresses gradually buildup on the anatase surface with the increase of particle size and/or precipitation of rutile phase. A coarsening mechanism 4,5,26,27 is commonly applied for the nanoparticle phase transformation. In such case, interface boundary atomic migration is the primary source for phase growth and the migration is strongly dependent on G A R. In our case of anatase to rutile phase transformation, at temperature around 700 C, rutile begins its nucleation and growth. According to previous work, 28,29 in the temperature range of C, rutile nucleation is dominated by surface nucleation (surface nucleation interface nucleation bulk nucleation). Thus, rutile nuclei keep consuming available anatase material from the surface and interface of anatase particle(s) until its total free energy becomes equal to or less than that of anatase. At the same time, temperature controlled heat and mass transportation causes anatase to grow by merging other anatase particles, although its net amount is diminishing. Therefore, it is easier to overcome the energy barrier to start the nucleation of a new phase for TiO 2 nanoparticles with smaller sizes compared to the larger particles because more free surface area and interface can be provided in smaller particles for the rutile formation. Moreover, anatase/rutile gets packed and become denser towards the interior of the particles as the growth of the anatase phase itself and the concurrent growth of the rutile phase proceeds. Thus, anatase lattice parameters are shortened, possibly by both the larger anatase particles and the precipitation of rutile. The larger anatase particle induced decrease of its own lattice parameters is similar to the results from Zhang et al. 30,31 for CeO 2 and BaTiO 3 nanoparticles in which the increased parameter with decrease of size was attributed to the effective negative pressure created by the competition between the long-range Coulomb attractive and the short-range repulsive interaction in ionic nanocrystals. Thus, the largest

6 6668 J. Appl. Phys., Vol. 96, No. 11, 1 December 2004 Li et al. TABLE III. XRD data for lattice parameter changes of TiO 2 anatase phase as a function of the annealing temperature. 12 nm sample 23 nm sample Peak positions 2 (deg) Lattice parameters (Å) Peak positions 2 (deg) Lattice parameters (Å) Temperatures (101) (200) a c (101) (200) a c As-deposited ± ± ± ± ± ± ± ± C ± ± ± ± ± ± ± ± C ± ± ± ± ± ± ± ± C ± ± ± ± ± ± ± ±0.005 value of anatase lattice change in 12 nm sample can be because of the quickest rutile and anatase growth rates induced strongest compressive pressure on the anatase surface. V. CONCLUSIONS The thermodynamics of the size effects on the thermal stability of TiO 2 anatase nanoparticles synthesized by MOCVD was studied. 12 nm anatase had the lowest transformation starting temperature and thermal stability as compared with 17 and 23 nm samples. According to the firstorder anatase to rutile phase transformation and experimental results of XRD and TEM, the activation energies for particles were calculated to be 180, 236, and 299 kj/mol for 12, 17, and 23 nm samples, respectively. Meanwhile, the decreased lattice values along the c direction of the anatase phase were found to be due to the increased particle size and/or appearance of rutile phase. It was more evident in the finer nanoparticles because of the quicker growth rate of rutile. ACKNOWLEDGMENT The authors would like to thank NSF-NIRT (Grant No DMR ) for the financial support of this project. 1 C. N.R. Rao and K. J. Rao, Phase Transitions in Solids (MacGraw-Hill, New York, 1978). 2 A. Sclafani, L. Palmisano, and E. Davi, J. Photochem. Photobiol., A (1991). 3 J. H. Braun, J. Coat. Technol. 69, 59(1997). 4 A. A. Gribb and J. F. Banfield, Am. Mineral. 82, 717 (1997). 5 H. Z. Zhang and J. F. Banfield, J. Phys. Chem. B 104, 3481 (2000). 6 M. R. Ranade et al., Proc. Natl. Acad. Sci. U.S.A. 99, 6476 (2002). 7 H. Z. Zhang and J. F. Banfield, J. Mater. Chem. 8, 2073 (1998). 8 S. R. Yoganarasimhan and C. N.R. Rao, Trans. Faraday Soc. 58, 1579 (1962). 9 R. D. Shannon and J. A. Pask, J. Am. Ceram. Soc. 48, 391(1965). 10 F. C. Gennari and D. M. Pasquevich, J. Mater. Chem. 33, 1571(1998). 11 J. Arbiol, J. Cerda, G. Dezanneau, A. Cirera, F. Peiro, A. Cornet, and J. R. Morante, J. Appl. Phys. 92, 853(2002). 12 A. Burns, G. Hayes, W. Li, J. Hirvonen, J. D. Demaree, and S. I. Shah, Mater. Sci. Eng., B 111, 150 (2004). 13 X. S. Ye, J. Sha, Z. K. Jiao, and L. D. Zhang, Nanostruct. Mater. 8, 919 (1997). 14 W. W. So, S. B. Park, K. J. Kim, C. H. Shin, and S. J. Moon, J. Mater. Sci. 36, 4299(2001). 15 R. D. Shannon, J. Appl. Phys. 35, 3414 (1964). 16 J. A. Gamboa and D. M. Pasquevich, J. Am. Ceram. Soc. 75, 2934(1992). 17 X.-Z. Ding, Z.-Z. Qi, and Y.-Z. He, Nanostruct. Mater. 4, 663(1994). 18 Y. Hu, H.-L. Tsai, and C.-L. Huang, Mater. Sci. Eng., A 344, 209(2003). 19 W. Li, S. I. Shah, M. Sung, and C.-P. Huang, J. Vac. Sci. Technol. B 20, 2303 (2002). 20 W. Li, S. I. Shah, C.-P. Huang, O. Jung, and C. Ni, Mater. Sci. Eng., B 96, 247 (2002). 21 W. Li et al. Appl. Phys. Lett. 83, 4143 (2003). 22 Free software from 23 B. D. Cullity and S. R. Stock, Elements of X-Ray Diffraction (Prentice Hall, New Jersey, 2001). 24 J. S. Olsen, L. Gerward, and J. Z. Jiang, J. Phys. Chem. Solids 60, 229 (1999). 25 Z. Wang, S. K. Saxena, V. Pischedda, H. P. Liermann, and C. S. Zha, J. Phys.: Condens. Matter 13, 8317 (2001). 26 T. C. Chou and T. G. Nieh, Thin Solid Films 221, 89(1992). 27 P. I. Gouma, P. K. Dutta, and M. J. Mills, Nanostruct. Mater. 11, 1231 (1999). 28 H. Z. Zhang and J. F. Banfield, Am. Mineral., 84, 528 (1999). 29 H. Z. Zhang and J. F. Banfield, J. Mater. Res., 15, 437(2000). 30 F. Zhang, S.-W. Chan, J. E. Spanier, E. Apak, Q. Jin, R. D. Robinson, and I. P. Herman, Appl. Phys. Lett. 80, 127 (2002). 31 V. Perebeinos, S.-W. Chan, and F. Zhang, Solid State Commun. 123, 295 (2002).