Thermal Stability of Brookite

Mater. Res. Soc. Symp. Proc. Vol. 1146 © 2009 Materials Research Society

1146-NN04-02

Thermal Stability of Brookite - TiO2 Nanoparticles with Controlled Size and Shape: in-situ studies by XRD Céline Perego1, Isabelle Clemençon1, Bernadette Rebours1, Renaud Revel1, Olivier Durupthy2-3, Sophie Cassaignon2-3, Jean-Pierre Jolivet2-3 1 IFP-Lyon, 69360 Solaize, France. 2 UPMC Univ. Paris 06, UMR 7574, Chimie de la Matière Condensée de Paris, Paris, France. 3 CNRS UMR 7574, Chimie de la Matière Condensée de Paris, Paris, France. ABSTRACT TiO2 is a material of great interest for many technological applications, such as photocatalyst or catalyst support. As the material properties depends on the polymorph used: rutile, anatase or brookite, it is of the foremost importance to know the thermal stability of each polymorph. The anatase to rutile phase transition has been largely studied. On the contrary, very few work refer to the brookite to rutile transition. In this study, a set of nanoparticles (~ 10 nm) of pure brookite with an isotropic shape was synthesized and sintered under various conditions. In situ X-ray diffraction (XRD) measurements were carried out from 600 to 750°C, without detection of the anatase polymorph. The influence of the annealing gas atmosphere on the brookite to rutile transformation, and the influence of adsorbed cations on the particles and the type of aggregation on the transformation kinetics are presented and discussed. INTRODUCTION Understanding the mechanism of the phase transformation of a material is fundamental to the control of the microstructure and, thus, the material properties. Nanocrystalline titania (TiO2) is a material widely used in the electronics, ceramics, pigments and catalysis domains. Three polymorphs of titania are mainly observed: rutile (thermodynamic phase), anatase and brookite (metastable phases). All three structures are built up of TiO6 octahedra, stacked in different ways. Rutile can be described as arrays of linear chains of edge-sharing octahedral, brookite as arrays of zigzag chains of octahedral sharing three edges and anatase also as arrays of zigzag chains but octahedra share four edges. The thermodynamic stability of the polymorphs was proposed to depend also on the size of the particle (D) at the nanometer scale [1] : at D < 11 nm, anatase is the most stable phase; for 11 < D < 35 nm, brookite is the most stable phase; while for D > 35 nm, rutile is the most stable phase. Those results come from thermodynamic considerations that take into account calculated surface energy of as cleaved surface considered at 0K. Therefore, results may vary slightly with temperature and pressure conditions [2,3]. The anatase to rutile phase transformation has been widely studied over the last decade, and many kinetic models have been proposed [4,9]. On the contrary, the brookite polymorph was less studied, probably due to the difficulties encountered in obtaining phase-pure brookite. To date, only two studies of brookite to rutile have been published, using monodispersed synthetic nanocrystals [10] and natural materials with an average composition of (Ti0.966, Nb0.016, Fe0.01, V0.008)O2 [11]. In this work, pure-phase brookite nanoparticles with a size of ~10 nm and isotropic shape were studied. Two preparation modes of the powder from the colloidal solution to the powder were used and they influence on sintering analyzed. No anatase polymorph was observed by X-

ray diffraction (XRD) measurements carried out in situ from 600 to 750°C. The influence of the annealing gas atmosphere on the brookite to rutile transformation was also studied. Finally, the choice a kinetic model for the brookite to rutile phase transformation was discussed. EXPERIMENTAL DETAILS The brookite particles were obtained through the co-thermohydrolysis at 60°C of the aqueous precursors TiCl3 and TiCl4 [12]. An equimolar solution of Ti3+ and Ti4+ (0.1 mol.L-1) was prepared and the pH adjusted to 4.5 with sodium hydroxide. The suspension was aged one week at 60°C without stirring. After the synthesis the particles were collected and washed either by filtration (sample #1) or by centrifugation (sample #2). The particles of the sample #1 were flocculated, in order to facilitate the filtration, by increasing the pH up to 6 (close to the isoelectric point) [13] by addition of NaOH, and then washed with deionized water. In sample #2, the particles were separated and washed several times with deionized water by centrifugation at 10000 rpm. The powders were then dried at 120°C for one day. The obtained brookite particles have an isotropic morphology with an average size of ~10 nm. In situ Bragg–Brentano X-ray diffraction (XRD) measurements were carried out in an Anton Paar XRK900 cell, using a PANalytical X’Pert Pro diffractometer equipped with a copper tube, a nickel beta filter, a position sensitive detector (X'Celerator). The samples were placed on the sample holder over a sintered glass plate, allowing the gas to flow through the sample. The sample holder is located in an oven ensuring a proper homogeneity of the sample temperature. Two types of thermal treatment were applied: (a) the temperature was raised at 5°/min from 100°C up to 900°C with 50°C steps at which a X-ray diffractogram was collected (b) the temperature was raised at 5°/min up to the desired annealing temperature (600, 650, 700 and 750°C) and maintained several hours. X-ray data were collected all along the calcination time, and all measurements were done under a gas flow (air or nitrogen) of 2L/h. Data were collected over a 5-90° 2θ angular range with a step size of 0.033° and a collection time of 150 s per step, with a scanning mode and fast detector with a 2.1° active length. XRD patterns obtained were used for the determination of crystallites size (applying the Scherrer formula) and for the calculation of weight fraction of the phases, using the relative area of the (110) and (121) diffraction lines of rutile and brookite respectively [14]. Particle morphology was observed by transmission electron microscopy (TEM, Model FEI Tecnai 20F Ultratwin). DISCUSSION Influence of the gas atmosphere The evolution of the weight percentage of rutile and the average particles sizes of brookite and rutile in the step by step annealing of brookite nanoparticles is presented in Figure 1. The results of sample #1 display the same evolution of the weight percentage of rutile under air and under nitrogen. The first rutile particles are detected at 700°C, and the transformation into rutile is completed at 850°C. The average particles size evolution of brookite and rutile in sample #1 doesn't depend on the nature of the atmosphere. In sample #2 the brookite to rutile transformation starts at lower temperature (600°C) and is completed at 700°C. The size evolution of brookite particles with temperature is the same under air and under nitrogen (superimposed points on plot). On the contrary, a difference can be observed concerning that of rutile particles.

They grow bigger when annealed under nitrogen: 85 nm instead of 60 nm under air. The same trend is observed in sample #2 sintered at 800°C for 3h: the average particle size of rutile is 40 nm under air and more than doubled under nitrogen with 90 nm. A possible explanation concerns the amount of oxygen vacancy defects in the particles. Oxygen vacancy defects, for example, are formed by annealing TiO2 rutile (110) surfaces above 550°C; these vacancies have been detected with a variety of techniques [15]. Conversely, oxygen vacancies can be filled by exposure to molecular oxygen. In the presented experiments, differences in rutile particles size are observed from 700°C and the difference increase with temperature. We assume that the particles under nitrogen atmosphere may develop more oxygen vacancy defects. Grain growth mechanisms include the step of anionic vacancy creation due to the movement of hydroxyl species and this step is needed for the diffusion of titanium ions [16,18]. The greater number of oxygen vacancies in the particles under nitrogen may promote the grain growth of the rutile particles. As mentioned earlier, the thermodynamic stability of the polymorphs is size dependant [1]. It is observed here that, in both samples, the brookite starts to transform into rutile as soon as the size is larger than 25 nm. Then the temperature of the phase transition between the two samples may be more understood as the temperature for which the critical brookite particles size is reached. Consequently the 100°C shift for the phase transition observed between the two samples is related to a slower brookite growth. 100

120

Size (nm)

% Rutile

%R-air %R-nitrogen

60 40 20

60 40

0

0

200

400

a)

600

800

0

1000

200

400

b)

T(°C) 100

600

800

1000

600

800

1000

T(°C) 120 B-air B-nitrogen R-air R-nitrogen

100

80

%R-air %R-nitrogen

Size (nm)

% Rutile

80

20

0

60 40 20

80 60 40 20

0

0

0

c)

B-air B-nitrogen R-air R-nitrogen

100

80

200

400

600

T(°C)

800

1000

0

d)

200

400

T(°C)

Figure 1: Weight percentage of rutile and average particle sizes of brookite and rutile respectively for sample #1 (a and b) and sample #2 (c and d).

Presence of sodium on the particles surface Samples #1 and #2 are different because of the way there were washed after the synthesis. The flocculation performed in sample #1 by addition of NaOH may have provoked the adsorption of more sodium ions at the nanoparticles surface than in the sample #2 washed by successive centrifugations with water. This is confirmed by ICP analysis as the sodium amount in sample #1 is 580 ppm while it is five time smaller in the sample #2 (106 ppm). Moreover the flocculation method may have induced an ill-ordered stacking of brookite particles whereas, in the centrifugation method, particles may have time to stack more compactly. The difference in specific surface measurement of the two samples confirms that hypothesis: Sspe = 230 m²/g for sample #1 and only 140 m²/g for sample #2. The presence of adsorbed sodium on the brookite particles have an impact on the grain growth and phase transition [19]. Indeed the delay in phase transition under air may actually be attributed to a delay in the grain growth of the brookite particles in presence of sodium, which may stabilize the initial brookite particles. Moreover, an ill-order stacking of the particles may prevent the formation of twinning at the interface between two brookite particles. The relation between sodium cations and the presence of oxygen vacancy defects is not fully understood, but the sodium adsorption on the surface may prevent the grain growth of rutile under nitrogen. Phase transformation kinetics Table 1 Kinetics models references 1. Standard 1st order [20,22]0] 2. Standard 2nd order [23]1] 3. Contracting spherical interface [18,24,25]3] 4. Nucleation and growth of overlapping nuclei [18,25]3] 5. One dimensional, linear, branching nuclei and constant growth [18,25]3] 6. Random nucleation and rapid growth [18,25]3] 7. JMAK model [4,5,26]4] 8. Interface nucleation and constant growth [8]6] 9. Pure interface nucleation [9]7] 10. Combined interface and surface nucleation [9]7] 600°C

1,0

650°C

700°C

750°C

2

lnα = kt + c ln(1-α) = kt + c ln[-ln(1-α)] = n lnk + n lnt 1/((1-α)(D0/D)3] - 1 = k2t(m-1)/m (D/D0)3 / (1-α) - 1 = (k2N0)t (D/D0)3 / (1-α) - 1 = (1 + (k2N0/k1))(ek1t - 1) y = 0,6742x - 0,2288 2

1

ln(-ln(1-α))

0,8

Wt % Rutile

Equation ln(1-α) = -kt (1-α)-1-1 = kt (1-α)1/3 = kt + c [-ln(1-α)]1/3 = kt

0,6 0,4 0,2

R = 0,9964

y = 0,5867x - 1,8737 2

R = 0,9985

0 -1 -2 -3 -4

y = 0,6704x - 3,5059

y = 0,8955x - 6,13

2

0,0 0

10

20

30

40

50

60

-5

2

R = 0,9961

-2 -1 0 1 2 Time (h) ln t (a) (b) Figure 2: Weight fraction of rutile (a) and JMAK plot of transformation data (b).

R = 0,9326

3

4

5

The brookite particles of sample #1 that display a higher thermal stability were used for the study of the brookite to rutile transition. Indeed, a high thermal stability is required for their potential use as catalyst support. Various kinetics models, summarized in Table 1, were proposed in the literature to describe the kinetics of the phase transformation. Among the selected models, two set seemed to fit correctly the collected experimental data: the Johnson-Mehl-AvramiKolmogorov (JMAK) model [4,5,26] and a set of models proposed by Banfield et al. for the anatase to rutile transition at different temperatures (numbers 8-9-10) [8,9]. The JMAK model is widely used to describe solid-state phase transformations. If the JMAK equation is applicable, a linear relation is obtained between ln[-ln(1-α)] where α is the weight fraction of rutile (Figure 2.a) and ln(t). The slope of the line, called the Avrami exponent value, can give information about the nature of the transformation mechanism. Among the limitations of the model, it assumes that samples are infinite size, what makes it not applicable to nanocrystalline particles. Indeed, phase transformation of nanocrystalline particles may involve interface nucleation and nuclei growth which is limited to the extent of a single nanoparticle. However, the JMAK model correctly fit our results for temperature from 650 to 750°C, but not for 600°C (Figure 2.b). The obtained Avrami exponent values of n = 0.59-0.67 are close to 2/3, which correspond to the model of growth along dislocations. This is also close to the value found by Huberty [11]. The value of the activation energy can be calculated from the kinetic constants measured for the three higher temperature and is Ea = 382 kJ.mol-1 (R² = 0.9842). It is higher than the value of 213 kJ.mol-1 found by Li [10] with the JMAK model on ~25 nm brookite particles. This results are in good agreement with the proposition that, at nanoscale, the activation energy increases with a decrease in particle size (model number 8 in Table 1) [9]. 250

600°C 650°C 700°C

160

[(Da /D0)3/(1-α)]-1

[1/((1-α)*(D0/Da)3)]-1

200

120 80 40 0 0

10

20

30

Time (h)

40

50

60

700°C 750°C

200 150 100 50 0 0

10

20

30

40

50

60

Time (h) (b) (a) Figure 3: Interface nucleation and constant nuclei growth plot (a), pure interface nucleation plot at 700°C and combined interface and surface nucleation plot at 750°C (formula multiplied by 0.2) (b).

The interface nucleation and constant nuclei growth model takes into account not only the time t and the fraction of transformation α, but also the initial particle size D0, the mean size of particles D and an additional time corresponding to the time needed for the nuclei to grow to the entire volume. This first model fits well our data for 600°C, 650°C and 700°C (t < 20h), with a value of 1.5-1.7 for the exponent m, representing the grain growth behavior (Figure 3.a). The determination of the activation energy gives a value of Ea = 193 kJ.mol-1 (R² = 0.9976). At 700°C for a calcination time over 12h, the pure interface nucleation model fits experimental data (Figure 3.b) with R² = 0.9923. At 750°C, model of combined interface and surface nucleation is the more appropriate, with R² = 0.9912. The use of three different models for various

temperatures is not surprising, indeed the predominant nucleation mode may change from interface nucleation and constant nuclei growth at low temperatures, to pure interface nucleation at intermediate temperatures and to combined interface and surface nucleation at higher temperatures [8,9]. The successive mechanisms take place at higher temperature (approximately 50 to 100°C) than in the anatase to rutile transformation and at much slower rate (50 times lower). The change in model with reaction time has already been observed by Banfield et al., but only inside the unique interface nucleation and constant nuclei growth model. However, they proposed that this phenomenon could be extended to the switch from one model to another without observing it. We show here that at 700°C and for reaction time under 20h, the particles grow according to the interface nucleation and constant nuclei growth model; the pure interface nucleation model can be applied for a calcination time over 12h, when the time needed by the nuclei to grow to the entire volume is not to consider any more. Both models fit the experimental data at 700°C, between 12 and 20h, which correspond to a weight percentage of rutile of respectively 45% and 60% (Figure 2.a). More brookite-brookite interfaces are present during the first part of the calcination, while the quantity of brookite-rutile interface is more important after 20h. The interface nucleation might be faster when coming directly from a rutile particle and induces a change in the phase transition mechanism. At 600 and 650°C, the weight percentage of rutile doesn't reach 45% during our calcination time, and at 750°C, this value is exceeded in less than an hour, explaining why no change of mechanism is observed for the other temperatures. CONCLUSIONS The brookite to rutile phase transformation has been studied using particles of pure phase brookite (10 nm), with an isotropic morphology. The phase transition experiments carried out by in situ X-ray diffraction in the temperature range of 600-750°C evidenced that the brookite particles directly transform to rutile without detection of the anatase polymorph. The nature of the gas atmosphere was shown to have an influence on the growth of the particles, for temperatures over 700°C. The particles grow bigger after calcination under nitrogen than under air, probably due to the increasing amount of oxygen vacancy defects in the particles under nitrogen. The surface adsorption of sodium on the brookite particles and an ill-order stacking of the particles induce a phase transformation at higher temperature. The reaction rates of the particles transition are best fitted with the JMAK model and those proposed by Banfield et al.. The type of nucleation during calcination was shown to be dependant both on the chosen annealing temperature and on the duration of the thermal treatment. Additional kinetic studies on other brookite particles with different size, morphology and amount of sodium at the surface, are mandatory to understand more completely the thermal stability of that interesting polymorph. REFERENCES 1. H.Z. Zhang and J.F. Banfield, J. Phys. Chem. B 104, 3481 (2000). 2. C. Arrouvel, M. Digne, M. Breysse, H. Toulhoat and P. Raybaud, J. Catal., 222, 1, 152 (2004). 3. M. Digne, P. Sautet, P. Raybaud, P. Euzen and H. Toulhoat, J. Catal., 211, 1, 1 (2002). 4. A. Suzuki and R. Tukuda, Bull. Chem. Soc. Jpn. 42, 1853 (1969). 5. K.N.P. Kumar, K. Keizer and A.J. Burggraaf, J. Mater. Chem. 3, 1141 (1993).

6. K.P. Kumar, K. Keizer, A.J. Burggraaf, T. Okubo and H. Nagamoto, J. Mater. Chem. 3, 1151 (1993). 7. A.A. Gribb and J.F. Banfield, Am. Mineral. 82, 717 (1997). 8. H.Z. Zhang and J.F. Banfield, Am. Mineral. 84, 528 (1999). 9. H.Z. Zhang and J.F. Banfield, J. Mater. Res. 15, 437 (2000). 10. J.G. Li and T. Ishigaki, Acta Mater. 52, 5143 (2004). 11. J. Huberty and H. Xu, J. Solid State Chem. 181, 508 (2008). 12. Koelsch, M., PhD, Université Pierre et Marie Curie, Paris (2004). 13. M. Kosmulski, Adv. Colloid Interface Sci. 99, 255 (2002). 14. A. Pottier, C. Chaneac, E. Tronc, L. Mazerolles and J.P. Jolivet, J. Mater. Chem. 11, 1116 (2001). 15. W.S. Epling, C.H.F. Peden, M.A. Henderson and U. Diebold, Surf. Sci. 412-413, 333 (1998). 16. J.L. Hébrard, P. Nortier, M. Pijolat and M. Soustelle, J. Am. Ceram. Soc. 73, 79 (1990). 17. J.L. Hébrard, M. Pijolat and M. Soustelle, Solid State Ionics 32-33, 852 (1989). 18. K.J.D. MacKenzie, Trans. J. Br. Ceram. Soc. 74, 77 (1975). 19. K.J.D. MacKenzie, Trans. J. Br. Ceram. Soc. 74, 29 (1975). 20. A. Suzuki and Y. Kotera, Bull. Chem. Soc. Jpn. 35, 1353 (1962). 21. C.N.R. Rao and K.J. Rao, Phase Transitions in Solids, 91 (1978). 22. J.F. Banfield, B.L. Bischoff and M.A. Anderson, Chem. Geol. 110, 211 (1993). 23. A.W. Czanderna, C.N.R. Rao and J.M. Honig, Trans. Faraday Soc. 54, 1069 (1958). 24. E.F. Heald and C.W. Weiss, Am. Mineral. 57, 10 (1972). 25. R.D. Shannon and J.A. Pask, J. Am. Ceram. Soc. 48, 391 (1965). 26. S. Hishita, I. Mutoh, K. Koumoto and H. Yanagida, Ceram. Int. 9, 41 (1983).