Available online at www.sciencedirect.com

Journal of the European Ceramic Society 33 (2013) 1917–1928

Low temperature synthesis of ultrafine non vermicular a-alumina from aerosol decomposition of aluminum nitrates salts Yann Aman a,b , Cécile Rossignol a , Vincent Garnier b , Elisabeth Djurado a,∗ a

Laboratoire d’Electrochimie et de Physico-Chimie des Matériaux et des Interfaces, LEPMI, UMR 5279, CNRS – Grenoble INP – Université de Savoie – Université Joseph Fourier, BP75. 38402 Saint Martin d‘Hères, France b Université de Lyon, INSA Lyon, Laboratoire MATEIS UMR CNRS 5510, 7 avenue Jean Capelle F-69621 Villeurbanne, France Received 19 September 2012; received in revised form 22 January 2013; accepted 25 January 2013 Available online 19 February 2013

Abstract In this paper, spherical, smooth and unagglomerated ultrafine amorphous powder particles were prepared by ultrasonic spray pyrolysis (USP) of easy-handling aqueous aluminum nitrate salts increasing the precursor solute concentration to 0.5 mol L−1 and reducing the pyrolysis temperature to 700 ◦ C. The transformation of the USP alumina powders into a-Al2 O3 was studied using combination of X-ray diffraction, electron microscopy, infrared spectroscopy, BET surface area, thermogravimetry and differential thermal analysis. A downward shift of the onset temperature of a-phase transformation to 900 ◦ C has been detected using a larger precursor solution concentration and performing a milling before calcination due to an increase in the surface density of defects, in surface area and in anisotropic particle shape. Additional post-milling of the low calcined powders allowed the preparation of agglomerate-free pure ultrafine a-Al2 O3 powder particles (∼100 nm, 28 m2 g−1 ), free of vermicular microstructures. © 2013 Elsevier Ltd. All rights reserved. Keywords: a-alumina; Ultrasonic spray pyrolysis; a-phase transformation; Aluminum nitrates

1. Introduction The widespread use of advanced a-Al2 O3 ceramics for thermo-mechanical, optical, electrical or chemical applications depends on the ability to control the microstructural development during processing and to obtain submicron grain size at full density.1 To achieve such a dense and submicron sintered microstructure, the process has to start with ultrafine/nanosized alumina powder provided that (i) sufficient quantities of high purity agglomerate-free nanopowders are available, (ii) the compaction behavior does not induce flaws before sintering, and (iii) grain growth is minimized during sintering.2 However, the synthesis of a-Al2 O3 defects-free nanosized powder is difficult and still remains a challenging issue. Indeed, the reported chemical methods such as gas-phase processes (including aerosol and vapor phase syntheses)3–7 sol–gel routes8–12 emulsion calcination synthesis13 or alum-derived synthesis14,15 often produce nanosized powders consisting of amorphous or metastable

∗

Corresponding author. Tel.: +33 4 7682 6684; fax: +33 4 7682 6777. E-mail address: [email protected] (E. Djurado).

0955-2219/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jeurceramsoc.2013.01.035

crystallographic alumina phases (e.g. transition aluminas such as boehmite, g, d, u). Upon heating, dehydration of water structure, dehydroxylation and structural changes of the intermediate metastable phases occur16–18 followed by nucleation and growth14,19 of the ultimate stable phase a-Al2 O3 (corundum). These phases transformations evolve through topotactic changes with low activation energy among the face-centered cubic (fcc) arrangements of oxygen anions in transition aluminas (g → d → u), and cations ordering over tetrahedral and octahedral sites in the oxygen sublattice. In contrast, the hexagonal close-packed (hcp) arrangement of oxygen anions and the ordering of Al cations solely over 2/3 of the octahedral interstices in the ultimate stable phase a-Al2 O3 require a reconstructive mechanism, and therefore a relatively high activation energy and high transformation temperature above ∼1200 ◦ C.19,20 This high temperature a-phase transformation is accompanied by a detrimental decrease in specific surface area. An investigation reported by McHale et al.21 suggested that polycrystalline nanostructured a-Al2 O3 is energetically unstable with respect to polycrystalline nanostructured g-Al2 O3 at surface areas of ∼125 m2 g−1 , i.e. a grain size of ∼8 nm. Hence, this would also affect the grain growth of the primary a crystallites through a

1918

Y. Aman et al. / Journal of the European Ceramic Society 33 (2013) 1917–1928

considerable ‘sporadic’ coarsening during the a-phase transformation. More problematic, the polymorphic crystal change during the transformation to a-Al2 O3 is accompanied by an increase of ∼17% in relative density.22 With a low nucleation density it would result in large spacing between nucleation events and the formation of micrometer scale, single crystal a-Al2 O3 grains with dendritic protrusions surrounded by continuous pore channels also known as the vermicular microstructure.14 This wormlike microstructure would require high sintering temperatures (>1600 ◦ C) to reach high densities and that would inevitably lead to exaggerated grain growth.14 Therefore, lowering the temperature of a-phase transformation should be a prerequisite to limit the growth of primary a-crystallite, and to minimize the development of vermicular microstructures. Hence, the complete transformation to the stable corundum phase at low temperature should ensure the synthesis of ultrafine unseeded a-Al2 O3 . Factors affecting either the temperature and transition kinetics or the sequence of phase transformation depend on the synthesis method and the starting precursor,6,23,24 in terms of structure25–27 thermal history and residual hydrol groups,18,28,29 morphology and particle size,21,30,31 initial compact density32 initial particle packing33 doping additives or impurities34–36 seeding and calcining atmosphere.15,37 In the present investigation, we report the synthesis of ultrafine a-Al2 O3 from aerosol decomposition (also known as spray pyrolysis) of aluminum nitrates precursors. Spray pyrolysis methods are well-suited for large-scale, continuous synthesis of multicomponent materials and nanoparticles.4,38 They offer many advantages for processing of alumina ceramics such as the production of nano to submicron scale easily dispersible powders with no hard agglomerates 39 the availability of a wide range of precursors solutions (alkoxides, salts, colloids)6,38,39,40 which allows a wide range of molecular mixing and costs, and the potential to control product homogeneity and purity, size, shape and distribution of the resulting powder particles. To date, there are few extensive reports on the USP synthesis of this powder in the literature review.38,40–42,43 Messing’s group38 reported the influence of precursor nature and concentration on particle morphology and porosity. They obtained spherical solid submicron alumina particles from sulfate salts, instead of macroporous micronic particles in presence of ethylene glycol. Martin et al41 have studied the microstructural and phase evolution for different post annealing (calcination) conditions of nitrate-derived alumina USP-synthesized at 400 ◦ C. The phase crystallization into a-Al2 O3 was promoted by additional thermal treatment up to 1300 ◦ C. However, systematic investigations of the combination of different USP parameters were only proposed by Okada,43 Jokanovic42 and Vallet-Regi.40 These authors have characterized the influence of the nature and concentration of precursors, SP temperatures, ultrasonic frequencies and gas carrier flow rates on the subsequent phase transformation of alumina particles after calcination. In Okada’s investigation, the atomization frequency was not mentioned. Nitrate salts were reported to favour the transition to alpha alumina at temperature as low as ∼1000 ◦ C43,42 and it was reported to be faster in the case of nitrate ions than in the case of chloride ions. In addition, submicron to micronic particles were produced, but average crystallites

size and microstructure were not mentioned systematically in those previous works. In the present paper, easy handling, low cost aqueous solutions of aluminum nitrates precursors were investigated based on this USP literature background. The dissolved precursor salts were ultrasonically aerosolized before being transported into the reactor furnace. As evaporation of water solvent proceeded, solute precipitation occurred within droplets when the critical supersaturation solute concentration was achieved as reported by Messing et al.38 This was followed by drying, pyrolysis of the precipitates, and eventually sintering at high temperatures to form amorphous or polycrystalline spherical ultrafine alumina particles. The effects of the operating ultrasonic spray pyrolysis (USP) parameters, i.e. reaction temperature and solute concentration, on the powder physical-chemical characteristics and the effect of an additional milling step, on the morphology and the phase transformation kinetics of the as-prepared powders were then systematically investigated, in order to provide a versatile approach for low temperature synthesis of ultrafine unseeded aAl2 O3 powders. The post-milling treatment is not advantageous in terms of time-consuming because it is an additional step to the one-step USP process but even with this additional step, USP is easy handling and is a low cost process starting from aqueous solutions of aluminum salts. 2. Experimental 2.1. Materials and precursor solutions Aluminum nitrate salts (Al(NO3 )3 ·9H2 O, 99.1% purity) were purchased from Sigma Aldrich. Three different aqueous nitrates solutions depending on reagent concentrations were prepared by dissolving the appropriate amounts of salts (0.02, 0.1 and 0.5 mol L−1 respectively) in 1000 mL of deionized H2 O, under vigorous magnetic stirring at room temperature. The using of the lower solute concentration was motivated by the obtention of powder with high quality38 (the smallest particles and narrower distribution of particle size) but the main disadvantage is a low production. These samples will be respectively referred to as USP 0.02 M, 0.1 M and 0.5 M in the text below. 2.2. USP operating parameters The reactor system used in the present experiments has been reported previously.44 It consists of an ultrasonic atomizer (Gapusol type, RBI, France) equipped with three piezoelectric ceramic transducers for which the resonant frequency was fixed at 2.5 MHz, and a quartz tube (55 mm in diameter, 1090 mm heated length) located inside a cylindrical furnace (Carbolite TZF, UK). This furnace is divided into three zones, which first one (inlet) is always programmed at an inlet temperature 50 ◦ C lower than the second and the third zones maintained at the same temperature (pyrolysis reactor temperature Tpy ) in order to favor a progressive solute diffusion in the droplets to avoid crust formation on the final spherical particles. The gas carrier of the generated aerosol droplets into the pyrolysis zone was synthetic

1919

Y. Aman et al. / Journal of the European Ceramic Society 33 (2013) 1917–1928 Table 1 USP synthesis conditions for a constant carrier gas flow rate of 6 L min−1 . Solution concentration (M)

Atomisation rate (cm3 min−1 )

Tpy (◦ C)

Q react (L min−1 )

Heating rate (◦ C s−1 )

0.02 0.1 0.5

2 1.26 0.66

1000 700 700

273 191 191

3433 1666 1666

air (80% N2 + 20% O2 ). Synthesized alumina particles were then collected downstream on Mo wire in stainless steel electrostatic filter tube under an applied voltage of 10 kV. The collecting zone was pre-heated at about 200 ◦ C to avoid water condensation. The as-prepared powders were then stored in a bell-jar containing moisture absorbents. For all different synthesized powders, the ultrasonic atomization frequency was fixed at 2.5 MHz, and the gas carrier flow rate was maintained at 6 L min−1 . The influence of the pyrolysis temperatures (700 ◦ C and 1000 ◦ C) and the precursor solutions characteristics (see Table 1 for details) were investigated. These experimental conditions were defined on the basis of literature background.40,42,43 The choice of 1000 ◦ C was motivated by the obtention of a crystalline as-prepared powder in one step while the temperature of 700 ◦ C was preferred in order to avoid any silica pollution that is expected to form at high temperature. 2.3. Post-synthesis treatments. Calcination and attrition milling process Due to the limited temperature of the reactor furnace (≤1000 ◦ C), the short reaction time of pyrolysis (typically less than 19 s), and the higher temperature required for aphase transformation, annealing of the as-prepared powders was necessary. Hence, the collected powders were isothermally calcined (up to 18 h) in a conventional electric furnace in air, at different temperatures (900 ◦ C–1200 ◦ C). These annealing temperatures and dwell durations were determined such that they ensure complete formation of the ultimate stable phase aAl2 O3 . Additionally, attrition ball milling was conducted only for the USP 0.5 M powder, the most agglomerated powder. This milling step aimed to reduce agglomeration or break the weakly bonded polycrystalline particles, and prevent formation of vermicular microstructures. This step was performed before calcination with alumina grinding media (Dynamic Ceramic, Dynallox, UK, 92% a-Al2 O3 purity, 0.75–1.2 mm in diameter) in a nylon chamber with plastic stirrer. The available mill volume was filled with ∼66% of the small alumina balls, and 34% of the powders previously dispersed in ethanol solution (Aldrich, 95% purity). The powder was milled at 1200 rpm for 3 h. This milling time was found to give a reduce contamination due to the abrasion of grinding media. After milling, the mixture (balls + powder) was dried at 120 ◦ C for 24 h. Then, a 200-mesh nylon sieve was used to separate the dried milled powder from the alumina balls. This milled powder will be referred as USP 0.5 MBC (milled before calcination) in the text below.

2.4. Physical-chemical characterizations. Simultaneous thermogravimetry and differential thermal analyses (TG/DTA) were performed on as-prepared and milled powders (STA 409 PC Luxx, Netzsch Instruments, Germany). Samples (∼40 mg) were introduced into alumina (corundum) crucible, under flowing air (50 mL min−1 ), with a heating rate of 5 ◦ C min−1 up to 1300 ◦ C, without any dwell. As-prepared and calcined powders were characterized by Xray powder diffraction data collected in θ–2θ geometry using a Bruker D8 X-ray powder diffractometer (40 kV, 40 mA, sealed Cu X-ray tube) equipped with an incident beam Ge monochromator. Scans were continuously acquired from 20 to 80◦ 2θ, with a scanning mode of 0.05◦ per step, and the acquisition time of 1 s per step. Additionally, this X-ray diffractometer was equipped with a furnace (Anton Paar HTK1200 Oven 100–1200 ◦ C) to monitor in situ phase transition during annealing. The average crystallite size (L) was estimated by X-ray line broadening according to Scherrer formula.45 The instrumental broadening was corrected from the FWHM of a reference sample (Bruker ˚ The crystalline Corundum AXS-FF0.1–crystallite size 2000 A). phases were identified according to ICDD files. The fraction of a-phase formation aXRD versus annealing conditions was also empirically estimated from XRD pattern performed at high temperatures, according to the following formula: αXRD =

I1 0 4 It

(1)

where I1 0 4 is the integrated area of the peak reflection (1 0 4) characteristic of a-phase formation, and It the total integrated area of the whole XRD pattern. Fourier transform infrared transmission spectra were recorded on IR-Prestige 21 spectrophotometer (Shimadzu, Japan) in the frequency range of 400–4000 cm−1 at a scanning mode of 4 cm−1 . Prior to measurements, 2 wt% of powder sample was added to 200 mg of fine ground KBr powder then cold-pressed at 100 MPa to form 13 mm-diameter pellets for analysis. Particle morphology and size distribution were examined by high resolution scanning electron microscopy (FEG-SEM Ultra 55 Zeiss, Germany) and conventional transmission electron microscopy (TEM Jeol 200 CX, Japan). For TEM observations, the powder samples were ultrasonically disagglomerated and dispersed in acetone before dropping a low concentrated drop on a carbon-coated copper grid. Digital images were processed using the public domain software Image J46 Particle size measurements were carried out using pixel scale. At least 400

1920

Y. Aman et al. / Journal of the European Ceramic Society 33 (2013) 1917–1928

particles were identified as individual spherical particles and particle diameter was measured with the straight line tool. Hence, the arithmetic mean particle size (Dm ) and size distribution could be determined. Specific surface area (SSA) analysis. SSA data were obtained using Micromeritics ASAP 2010 sorption analyzer. Samples (200 mg) were degassed at 200 ◦ C overnight prior to analysis. Measurements were run at 77 K with nitrogen gas. SSA was determined by the BET 5 points method with relative pressures from 0.001 to 0.20. The average particle diameter (dBET ) was estimated from the measured SSA and the solid packing density (ρ, g cm−3 ) by assuming solid spherical particles (Eq. (2)). dBET =

6 SSA ∗ ρ

(2)

The solid packing density r was calculated from geometrical and weight measurements of cold-pressed pellets at 100 MPa. An agglomeration factor (Fag )47 is introduced to enable comparison of the synthesis conditions on the agglomeration state which is a characteristic for flawless compaction of each powder:

Fag =

D50 dBET

(3)

where D50 is the median diameter related to the size distribution also determined from image analysis. 3. Results and discussion 3.1. Influence of USP parameters on physical-chemical properties of alumina powders The final particle morphology, size distribution, and phase crystallinity result from many physical-chemical mechanisms occurring during the different stages of aerosol decomposition process. As observed on the SEM micrographs (Fig. 1), irrespective of the synthesis conditions, the three as-prepared powders present regular, spherical, smooth and presumably solid particles, with very few defects such as pinholes, distorted or wrinkled particles. Quasi-uniform monomodal narrow size distribution were obtained for low solute concentrations (0.02 M at 1000 ◦ C and 0.1 M at 700 ◦ C), whereas bimodal distribution could be distinguished from the size distribution of highly concentrated solutions (0.5 M at 700 ◦ C). The increase in precursor solute concentration from 0.02 M to 0.1 M resulted in larger particle mean diameter from 269 to 680 nm, respectively, and wider size distribution, as indicated in Table 2. The bimodal distribution of highly concentrated solutions (0.5 M) due to the size increase of the small particles could be ascribed to coalescence of the finest particles during evaporation/drying stage. Similar observations were reported by Zhang et al.48 on spray pyrolysis of TeO2 nanoparticles. Further SEM observations of USP 0.5 M showed few broken hollow particles (Inset of Fig. 1e) that probably originated from high internal stresses during particles formation and rapid evaporation of solvent from surface leading to the formation of the salt crust.38 Since in the present investigation the viscous effects during atomization can

be neglected regarding the low concentrated precursor aqueous solutions, therefore the ultrasonically sprayed droplet diameter d0 is proportional to the most unstable wavelength of the rising capillary waves49 and can be expressed as.50   πσ 1/3 (4) d0 = ρf 2 where σ (N m−1 ) is the surface tension and ρ (g cm−3 ) the density of the aqueous precursor solution, f (MHz) the atomization frequency. According to the calculated values (Table 2), similar droplets sizes are expected to be obtained for the different precursor solutions, suggesting similar physical properties of the corresponding precursor solutions. Thus quasi-monomodal size distributions are expected from ultrasonically sprayed water solute droplets, according to the maximum entropy formalism suggested by Lecompte and Dumouchel51 to model liquid spray droplet size distribution. However, this contrasts with the present results, for which bimodal size distribution of the final powder is obtained when the solute concentration is increased. Actually, since evaporation predominates over coalescence process for low gas carrier flow rate52 hence this change in particle size distribution can be ascribed to a higher collision or coalescence rate in the droplet aerosol spray for high concentrated solutions. The number of initial droplets, N0 , can be estimated from the solute volume fraction, the initial droplet size, and the atomization rate as follows: N0 =

1 ϕs × Atomization rate Vat · = 0.1 · 60 Vdroplet πd0 3

(5)

where Vat (cm3 s−1 ) is the atomized volume of solutes per unit time, ϕs the solute volume fraction, the atomization rate reported in Table 1, and d0 (cm) the droplet size, as defined by Eq. (4). The calculated values for USP 0.02 M, USP 0.1 M and USP 0.5 M powders are 7.72 × 106 , 2.38 × 107 , and 5.85 × 107 cm−3 , respectively. Accordingly, the higher the precursor solute concentration, the denser the aerosol fog is. Assuming that Brownian motion is the only reason for droplet collision in the aerosol generator (under laminar flow), since the critical collision characteristic time is inversely proportional to N0 , therefore a higher initial number density would increase the probability of droplet collisions. This simple analysis is however incomplete, since particle collision does not necessary induce coalescence. Indeed, coalescence is a diffusion-controlled process which will basically merge two collided particles of unequal sizes in one larger particle. Nonetheless, as noted by Hawa and Zachariah53 if the characteristic time of coalescence is lower than the collision one, a larger spherical particle should be produced. Otherwise, collided stuck particles are formed, resulting in multimodal size distribution in agreement with the present experimental results. According to BET measurements (Table 2), the specific surface area decreases from 31 to 3 m2 g−1 while dBET is increased when the solute concentration is ranged from 0.02 M to 0.5 M respectively. Since the ultrasonically sprayed droplet diameter is proportional to the surface tension of the precursor solution,48,49 then a decrease in the surface tension relatively to the low density of aqueous solutions should result in a finer

Y. Aman et al. / Journal of the European Ceramic Society 33 (2013) 1917–1928

1921

Fig. 1. SEM pictures of USP as-prepared powders and corresponding particle size distribution for (a, b) 0.02 M at 1000 ◦ C, (c, d) 0.1 M at 700 ◦ C, (e, f) 0.5 M at 700 ◦ C. Inset: SEM image of another zone of USP raw powder for 0.5 M at 700 ◦ C.

droplet size. Since very short residence times (