Processing of nano-scaled silicon powders to prepare slip cast structural ceramics

/

MATB]IUALS SCIEIRE & ENGINEERING

ELSEVIER

Materials Science and Engineering A204 (1995) 107 112

A

Processing of nano-scaled silicon powders to prepare slip cast structural ceramics C. Bossel a, J. Dutta a'*, R. Houriet", J. Hilborn b, H. H o f m a n n " ~Laboratoire de Technologie des Poudres, Dkpartement des Matdriaux, Ecole Polytedmique Fkdbrale de Lausanne, 1015 Lausanne, Switzerland bLaboratoire des Polymbres, Dbpartement des Matbriaux, Ecole Polytechnique Fbdkrale de Lausanne, 1015 Lausanne, Switzerland

Abstract

For slip casting of ceramic powders it is necessary to have a well-defined and thus disagglomerated colloidal suspension. Proper selection of the solvent is required in order to achieve separation of the particles to obtain a homogenous mixture of the powders which is necessary for shaping complex geometrical structures, often used in structural ceramics. Here we report a preliminary investigation of the deagglomeration phenomena of nano-scaled silicon powders obtained by plasma induced dissociation of silane and compare it with silicon nitride powders prepared by laser induced condensation reactions and a commercial product (UBE SE E-10). The size dispersion of aggregates in colloidal suspensions, determined by photon correlation spectroscopy and sedimentation particle size analysis techniques, varies from 20 to 500 nm. Variation in the deagglomeration properties of the particles in different solvents depends on the surface property of the powders, and on the inter-particle interactions. These are studied with respect to the variations in the surface property of the powders in different solvents. Ethanol was found to be a suitable solvent for the colloidal suspension as the average aggregate radii of the silicon powders could be reduced to 80 nm.

Keywords: Silicon powders; Ceramic powders; Nano

1. Introduction Nano-structured materials are of interest as they show the potential for making ceramics with enhanced mechanical properties due to the fine microstructures. It is thus important to control the processing of the powders to be able to make ceramics with a controlled microstructure [1]. Silicon based advanced ceramics are being studied in great detail for mechanical and optoelectronic applications. In ceramic processing a good control of the processing parameters can be achieved through slip casting of the powders. Slip casting permits the homogeneous consolidation of powder and the realisation of complex shapes with a high density green body. The processing of nanometric powders hardly allows green body densities over 40%, as the high specific surface leads to the formation of hard agglomerates in the colloid [2]. As synthesized nanoparticles can be completely deagglomerated as long as the primary particles are merely held together by surface forces rather than being partially sintered by neck

2. Experimental methods Silicon nano-powders (a-Si:H) were prepared by gas phase reaction of silane in a capacitively coupled radiofrequency plasma-enhanced chemical vapour deposition (rf-PECVD) system and collected from the electrodes after the synthesis, in Balzers's laboratories (Balzers SA, Palaiseau, France) and CRPP (EPFL, Lausanne, Switzerland) [3]. Silicon nitride (Si3N4) powders were synthesized by laser induced reaction between silane and ammonia (as shown in reaction (1)) at ENEA (Area Innovazione, CR Frascati, Italy) [4]. 3Sill4 + 4NH 3 = Si3N4(s) + 12H2

(1)

Commercial Si3N4 (UBE SN E-10) from UBE (Japan) was also studied for comparison.

* Corresponding author. 0921-5093/95/$09.50 © 1995 S S D I 0921-5093(95)09946-8

formation, which are usually referred to as hard agglomerates. In this work we report on the dispersion behaviour of nano-structured silicon powders in various solvents and examine the conclusions which can be drawn from it.

Elsevier Science S.A. All rights reserved

108

C. Bosse/ el al. ' Metier&Is Science and Engineerhlg A 2 0 4 (1995) 107 112

Mass loss and heat flow as a function of increasing temperature were recorded in a thermogravimetric differential analysis system ( T G A / D T A Setaram 680), allowing estimation of the phase transformations of the powders. The use of different atmospheres, namely oxygen and Forming gas (92%N2- 8%H2, gave valuable information on the oxidation behaviour of the powders. The gas adsorption technique using nitrogen (BET method) and helium picnometry were carried out to determine, respectively, the specific surface and the density of the powders. Crystallinity and crystallite sizes were evaluated from X R D [5]. High resolution transmission electron microscopy ( H R T E M ) was carried out in a Phillips EM 430 on samples sprayed on the carbon coated copper grids. Estimation of the particle size was made from the TEM measurements. Bonding and chemistry of the surface was monitored by a Nicolet 510 F T I R spectrometer from pellets made by mixing the powders with potassium bromide. Electron spectroscopy for chemical analysis (ESCA) was used to determine the atomic stoichiometry of the powder surface. Colloids were prepared with a powder concentration between 15 and 50 vpm (0.03 and 0.1 g 1 ~). All samples were dispersed for 10 min in an ultrasonic water bath. Solvents used were: distilled water, ethanol, isopropanol, ethylene glycol, methyl ethyl ketone, acetyl acetone, acetone, triethylamine and dimethyl acetyl acetate. The agglomerate or aggregate sizes in the colloids were determined by two methods. The first method was by sedimentation (Horiba Capa 700), which is based on the attenuation of transmitted light during sedimentation and which can measure particles larger than 100 nm. The second method used was photo correlation spectroscopy (PCS, on a Malvern Zetasizer 4 or BrookHaven Instrument), based on scattering of light by particles, which allows the estimation of particles as small as 5 nm.

heat treatment at elevated temperatures [6]. The crystallite size estimated from X R D for the as synthesized a-Si:H powders ( ~ 3 nm) is at the limit of thermodynamic equilibrium between the amorphous and the crystalline states. Short range ordering of the bond angles in the silicon matrix of the a-Si:H powders was interpreted from Raman spectroscopic measurements

[3]. BET gives information on the specific surface of the powder, which can be utilised to calculate the average diameter of the primary particles if we suppose that they are spheres with regular surfaces. These assumptions are justified from the observations made in TEM [6]. The specific surface area was 160 m z g ~ for the a-Si:H powders, which corresponds to primary particle sizes of 20 nm, comparing well with the particle sizes determined by TEM (20 30 nm).

3.2. Chemical analysis" a-Si:H powder was heated at low temperatures (between 80 and 200 °C) in a thermogravimetric analyser (TGA) system, in air and in a reducing atmosphere. Aggregate sizes after annealing were measured by PCS, to check the influence of oxidation and adsorbed moisture on the surface chemistry and the aggregation phenomena. The loss of water around 100 °C was observed but no difference in aggregate sizes was observed. This suggests that the surface is oxidized, and that the adsorbed water has no influence on colloidal dispersion.

3. Results and discussion

3.1. Morphological analysis A typical transmission electron micrograph showing the particle aggregation in a-Si:H powders is shown in Fig. 1. It can be observed that the'particles are aggregated into larger blocks, probably due to the electrostatic interactions (Fig. 1). X R D allows the estimation of the average crystallite diameter [6] by the use of the Debye Scherrer law (not considering any strain effects), from the broadening of the crystalline peaks. Crystallites form a small part of the primary particles, mostly amorphous, which grow into larger sizes after

Fig. 1. TEM micrograph showing the microstructure of the powder.

C. Bossel el al. / Materia£" Science and Engineering A204 (1995) 107. 112

110 10~I-~ .~

3.3. Ultrasonic dispersion

~i-O

90 ~l r

80

~11Si-H ....

V (sIHSn)I~ 70 2800

109

I stretch l,

2400

2000

I

1600

I Si-O

1200

I bend

800

I

400

wavenumber (cni 1) Fig. 2. Infrared spectrum of the silicon powder.

Heating at higher temperatures showed exothermic peaks at 350 and 620 °C, which are believed to be the signature of the loss of weakly bonded hydrogen and the beginning of crystallization, respectively. The chemistry of the surface controls the colloidal properties. Post-oxidation analysis of silicon nitride in TGA suggests that the powder contains about 6'V0 oxygen, presumably at the surface as silicon suboxides. ESCA measurements further confirm that oxygen accounts for 30 at.% of the superficial layers of a-Si:H powders, and 25% of Si3N 4 powders. Carbon is present as carbides or as carbonates in the surface as well, 1.5% in a-Si:H powders, and 6% in Si3N 4, evidently due to the contamination from exposure to the atmosphere. Also, due to the synthesis from silane the a-Si:H powders contain about 20 at.% hydrogen. In Fig. 2 a typical infrared spectrum of the silicon powder studied here is presented. Since the small particles are very reactive due to their large specific areas, exposure to atmosphere immediately oxidizes the powders, as can be observed from the Si-O absorption band centered at 1100 c m - 1. The oxygen is mostly bonded to the surface of the powders, as can be interpreted from the presence of Si-O3 absorption at 2240 cm-1. Silicon and hydrogen are bonded in clusters, as can be observed from the absorption peaks at 880 cm t, 2080 cm-~ and 2100 cm-1, which are attributed to S i - H 2 bending, S i - H 2 stretching and (Si-H2), stretching, respectively. Furthermore, the presence of Si-O absorbtion at 1100 cm l suggests bridging-type oxygen bonding on the surface. The characteristics of the oxygenated surface would be different for different amounts of oxygen in the powder. This can strongly influence the dispersion characteristics and so the results obtained in these experiments cannot necessarily be extrapolated to oxygen free samples.

In dry powder an agglomeration of primary particles is almost inevitable. In the colloidal state, several binding forces have to be counteracted in order to achieve an acceptable disagglomerated state. The major binding forces which are experienced by the particles in a low viscosity solvent are: (1) binding by adhesion forces between the particles, essentially due to Van der Waals interaction; (2) binding by wetting liquids: these boundary forces are in addition to the capillary underpressure in the liquid bridges and they are stronger than the adhesion forces but less influenced by the surface roughness. Dispersion of the powders essentially requires counteracting these two forces by physico-chemical and mechanical means to an extent that minimizes the aggregate sizes. In order to achieve this situation, it is absolutely essential to achieve good wetting of the solid particles by the liquid, deagglomeration of the aggregates and stabilization of the suspension to prevent renewed agglomeration. Ultrasonic dispersion was used to disperse the aggregates from the agglomerates. Above a power level of 1 W, ultrasonic waves in liquid form cavities which implode, producing an intense pressure gradient (20 GPa cm-~) [7]. We monitored the size of agglomerates in the suspension after various time delays of ultrasonic dispersion in the bath at a constant power. The agglomerate size decreases during the first 3 min of ultrasonic dispersion, and then reaches a minimum value which is believed to be the best dispersion possible under the present conditions. For all the samples we carried out the ultrasonic dispersion for 10 min. It has been reported that if ultrasonic dispersion is carried out for a longer time, aggregation of particles during implosions is observed, which we did not see under our present conditions.

3.4. Choice oj the solvent To obtain non-agglomerated sterically stabilized suspensions, strongly attached and dense layers of dispersant molecules are required on the powder surface. A strong adsorption is obtained by the interracial Lewis acid-base interactions between dispersant and surface site of the powder [8]. Another condition to obtain a stabilized suspension requires a solvent of low surface energy so as to wet the surface, which essentially means that the solvent must have a low dipolar moment. For example, distilled water could not be used to disperse our powders because of its high polarity, even after changing the polarity in pH or ionic content. Organic additives such as polyoxyethylene (Tween 80) had to be added in water to achieve wetting. But these additives create bubbles during the ultrasonic dispersion process, sometimes adversely influencing the PCS measurements.

C. Bossel et al./ Materials Science and Engineering A204 (1995) 107 112

110

A parameter F is often used to characterize the suitability of dispersing solvents, where F is given by [9]

1000

-

. Si3N4, EtOH + Water

qe

F-

-

.

.

.

.

500 where r/(cP) is the viscosity, e is the dielectric constant and/~ (D) the dipolar moment. A solvent with a low F is characterized by the tendency to create electrical charges on an initially uncharged surface to electrostatically stabilize the particles. Without this stabilization the particles would tend to re-agglomerate after the ultrasonic dispersion process. We measured the mean diameter of particles in several colloids (Fig. 3) with the same amount of a-Si:H powder, but different solvents, and monitored the reagglomeration phenomenon as a function of time. The mean diameter of agglomerates was not proportional to F, which might mean that the surface was already charged before the wetting treatment. One of the best solvents to achieve nearly total deagglomeration seems to be ethanol. We studied the mixtures e t h a n o l + M E K (methyl ethyl ketone) and ethanol + w a t e r with the powders a-Si:H and Si3N4 (Fig. 4). From this study it appears that the best stabilization is achieved when the ethanol content in the solvent is maximum, essentially thus motivating our choice of ethanol as the solvent for dispersing this type of powder. The measured diameter of the aggregates in all the powders considered in this study after the ultrasonic dispersion is shown in Table 1. The ultrasonic treatment of the colloidal suspension leads to the rupture of weak bonds between aggregates but stronger chemical bonds between primary particles are difficult to break. The estimated particle size in the colloids is then the size of the strong aggregates between primary particles. Knowing the diameter of primary particles, dpr~m, and the diameter of aggregates, can calculate the maximum number of particles dagg

,

w e

300 • Acetone 250 I~propanol ,5 = 200 MEK

Ethylene

~D

.~

Ethanol

150

.

.

.

.

.

.

E

~

CD

er

Si, EtOH + MEK 100

]

I

]

I

[

0.2

0.4

0.6

0.8

1

vol (EtOH/(EtOH+solvent)) Fig. 4. Diameter of a-Si:H and Si3N ~ aggregates in a mixture of varying a m o u n t of ethanol and M E K or water,

per aggregate or the fractal dimension of aggregate by using [10]

The average diameter in the colloid, as measured for a-Si:H, is 150 nm, which means that no more than 420 primary particles are contained per aggregate. Following similar reasoning, we estimate that no more than 1400 particles of S i 3 N 4 form an aggregate. The particles are necked to each other rather than sintered into one mass; thus the particles are obviously smaller in number in the aggregates. Observation by H R T E M confirms these estimations (Table 2).

3.5. Formation of the aggregates In the rf-plasma, large hydrogenated silicon clusters are formed by condensation reactions resulting in negatively charged clusters having highly cross-linked threedimensional structures [11]. The formation of small clusters, at least up to about 40 atoms of silicon, arises from negative ion polymerization in low pressure silane discharges, while anion-neutral and neutral-neutral condensation leads to homogeneous nucleation that ultimately results in the formation of the powders. The growth of the clusters is limited and when they get larger they may get ejected and attach to one of the

Glycol Table 1 Diameter (nm) of particle agglomerates in three different solvents

100

10

100

;00

parameter F 3. Diameter of a-Si:H in various solvents. No correlation is observed between the standard F = qE/l~ and the agglomerate diameter. Fig.

a-Si:H Si3N 4 (ENEA) Si3N 4 (UBE E-10)

EtOH

Water

M EK

146+__11 470 ___30 650

~720 _+ 50 630

162+15 530 +_ 50

C. Bossel et al. / Materials Science and Engineering A204 (1995) 107 112

111

Table 2 Morphological sizes observed by different analytical methods

Observed object a-Si:H, average diameter a-SigN 4 (ENEA), average diameter

XRD

BET

PCS

TEM

crystallites 3 nm 1 nm

primary particles 20 nm 40 nm

aggregates 150 nm 450 nm

all 20 150 nm 40 450 nm

electrodes where they are eventually collected. As the larger clusters are in continuous motion in the plasma they have a high probability of collision, which might lead to larger chemically bonded clusters. The overall mechanism of agglomeration is shown in Fig. 5. The charged clusters create an electrostatic repulsion between the particles, and in addition, face the gravitational force, drag due to gas flow, along with the force due to the electrical field across the electrodes [12]. Above a given size, the balancing forces are disturbed and the aggregates can no longer be sustained by the plasma; thus they get ejected. In the case of laser driven reactions, the primary agglomeration phase occurs in the gas phase and once the balancing forces consisting of the thermophoretic forces, the drag due to gas flow and the gravitational forces are overcome, the agglomerates are ejected from the reactor for collection.

observed in amorphous silicon as determined by photocorrelation spectroscopy. The silicon nitride powders obtained by laser driven reactions form aggregates of about 250 nm radii, while the commercial UBE powders are dispersed in aggregates of 300 nm sized radii, which agrees well with the literature. The smaller aggregate sizes of the powders synthesized by plasma processing could be due to the inherent properties of the plasma. Detailed study of the cluster growth and the development of the powder in the plasma will add more light to this observation. The a-Si:H powders obtained by the plasma reaction of silane qualify well for slip cast because of the uniform shape to the spherical aggregates. However, further studies have to be carried out to verify the advantages of small sized agglomerates of powders obtained by plasma induced condensation reactions, and to clarify whether these conclusions are universally true for the other silicon alloys.

4. Conclusions The properties of the colloidal suspensions necessary to prepare slip cast ceramics depend upon the solvents used. Ethanol was found to be a suitable universal solvent for the powders studied here as it dispersed the agglomerates into the smallest aggregates when compared with suspensions made in other solvents. Particle size measurements of the dispersion of nano-sized aSi:H powders obtained by plasma driven reactions of silane, and Si:N powders obtained by laser driven reactions of silane and ammonia suggest that the primary particles form chemically bonded aggregates, which are difficult to separate. Aggregate sizes of 80 nm radii were

Acknowledgements This work has been funded by the Fonds National project of the Federal Government of Switzerland under contract No. 2100-039361.93/1. We would like to thank Dr. J.P.M. Schmitt of Balzers S. A., Palaiseau (France) for supplying some a-Si:H powders and Dr. E. Borsella for supplying the silicon nitride powder produced at ENEA (Italy).

References [1] F.F. Lange, B.V. Velamakanni, J.C. Chang and D.S. Pearson,

°°° °Co°° O.. °°O 0

Proc. llth Riso Int. Symp. on Metallurgy and Material Science,

O

0

A electrostatic repulsion

/

Iaggregates I ~ thermophoretic forces. ~ - ~

Drag due to gas flow

gravitation ,~1 force

T~ electric field

Fig. 5. Model of powder agglomeration during synthesis in plasma.

1990, pp. 57 78. [2] I.A. Aksay, G.C. Tangle and M. Sarikaya, Ceramic Powder Processing Science, Proc. 2nd Int. ConiC, Berchesgaden, FRG,

1988, p. 463. [3] J. Dutta, W.S. Bacsa and C. Hollenstein,J. Appl. Phys. (1995)in press. [4] E. Borsella, S. Botti, M.C. Cecile and A. Nesterenko, US Italy Workshop on Advanced Materials, 1993. [5] J. Dutta, I.M. Reaney, C. Bossel, R. Houriet and H. Hofmann, Nanostruct. Mater. (1995) in press. [6] B.D. Cullity, Element of X-Ray Diffraction, Addison-Wesley, New York, 1978, p. 289. [7] L.D. Rozenberg, High Intensity Ultrasonic Fields', Plenum Press, New York, 1972, p. 263.

112

C. Bossel et al./ Materials Science and Engineering A204 (1995) 107-112

[8] F.M. Fowkes, in K.L. Mitttal (ed.), SurJace and Colloidal Science in Computer Teehnology, Plenum Press, New York, 1987, pp. 3-25. [9] M. Lihrmann, in Duran and Fernandez (eds.), Third Euro-ceramies VI, Faenza Edice Iberica, Spain, 1993, pp. 27-32. [10] P.J. Van der Put, R.A.Bauer and J.Schooman, in G.L. Messing

(ed.), Ceramic Powder Science 111, American Ceramic Society, Westerville, OH, 1990, pp. 259 266. [11] C. Hollenstein, J.-L. Dorier, J. Dutta, L. Sansonnens and A.A. Howling, Plasma Sources Sci. Technol., 3 (1994) 278 285. [12] J.-L. Dorier, C. Hollenstein and A.A. Howling, J. Vae. Sci. Technol. A, (1995) in press.