Sei sulla pagina 1di 12

Available online at www.sciencedirect.

com

Acta Materialia 57 (2009) 54545465


www.elsevier.com/locate/actamat

Direct comparison between hot pressing and electric


eld-assisted sintering of submicron alumina
Jochen Langer a,*, Michael J. Homann b, Olivier Guillon a
a
Institute of Materials Science, Technische Universitat Darmstadt, Petersenstrasse 23, D-64287 Darmstadt, Germany
b
Institute of Ceramics in Mechanical Engineering, Universitat Karlsruhe (TH), Haid-und-Neu-Strasse 7, D-76131 Karlsruhe, Germany

Received 21 January 2009; received in revised form 23 July 2009; accepted 24 July 2009
Available online 26 August 2009

Abstract

This study compares hot pressing (HP) and the electric eld-assisted sintering technique (FAST) of two dierent electrically insulating
Al2O3 submicron powders with median particle sizes of 150 and 500 nm. Sample geometry, heating schedule, applied pressure and atmo-
sphere were identical for both sintering methods. The densication behavior and characterization of the microstructure revealed that
FAST sintered samples reached a higher density compared with HP, in particular for the ner powder. It was found that an increase
in dwell time was required to reach the same nal density by HP. However, analysis of the sintering curves showed that the densication
mechanism for both sintering methods was grain boundary diusion. Increasing the heating rate up to 150 K min1 did not modify the
densication mechanism. The sintering trajectory showed that the grain size was only dependent on density and was insensitive to the
sintering method, in addition to showing a lack of preferential grain orientation.
2009 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Alumina; Electric eld-assisted sintering (FAST); Hot pressing (HP); Densication

1. Introduction ment of new materials, relatively few papers have been


published on the specic mechanisms involved during
The rst investigations and development of the spark FAST, either thermally, electrically or mechanically [5
plasma sintering technique (SPS) started in the 1960s 10]. It is not surprising that materials processed by FAST
and 1970s in Japan. Because plasma formation could often show better properties than those sintered in con-
not be validated, other terms were used for this technique, ventional furnaces, as the processing parameters are not
such as pulse electric current sintering or the eld-assisted identical [11,12]. Neglecting all possible eects of electrical
sintering technique (FAST) [1,2]. Due to the new possibil- eld and direct heating oered by FAST, there is still one
ities, interest in this method has increased over the last signicant dierence from the free sintering case: the
10 years. High heating rates of up to 1000 K min1 and applied compressive stress, which signicantly enhances
direct heating of electrically conductive densifying speci- densication [1315]. It is thus sensible to make a direct
mens have been the driving force for this expansion, comparison of FAST with its closest alternative, hot
allowing shorter production cycles for materials typically pressing (HP). The required equipment is similar: powder
dicult to fully densify [1] with limited grain growth is uniaxially pressed between two graphite punches in a
[3,4]. Although many researchers have focused on the graphite die and heated in a dened atmosphere (vacuum
advantages of FAST over free sintering and the develop- or inert gases). The primary dierence is how thermal
energy is transferred to the powder pellet. In the case of
FAST an electrical current ows through the graphite
*
Corresponding author. Tel.: +49 6151 16 6313. fax: +49 6151 16 6314. punches and die so that the pressing tool acts as the heat-
E-mail address: langer@ceramics.tu-darmstadt.de (J. Langer). ing element by the Joule eect. In the case of specimens

1359-6454/$36.00 2009 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2009.07.043
J. Langer et al. / Acta Materialia 57 (2009) 54545465 5455

with higher electrical conductivity the electric current can 2. Experimental procedure
also contribute to direct heating of the sintering powder.
In contrast, graphite heating elements in HP surround 2.1. FAST and HP experiments
the pressing tool, transferring heat by radiation/
convection. All experiments were made with HP (model HPW 150/
At the macroscopic scale several nite element method 2002200-100-LA) and FAST (model FCT HP D 25/1)
simulations have been performed to model temperature equipment manufactured by the same company (FCT Sys-
and current distributions in insulating and conductive teme, Germany). This allowed a better comparison
materials during FAST [1618], highlighting possible tem- between the two methods, as FAST units have been derived
perature gradients for high heating rates. However, addi- from pre-existing HP set-ups. Used materials were a-Al2O3
tional phenomena induced by the electric eld which powders: (i) purity 99.99% and median particle size 150 nm
aect densication were not taken into account, except (TMDAR, Taimei Chemicals, Japan) [24] and (ii) purity
recently in a constitutive model including electromigration 99.78% and median particle size 500 nm (CT 3000 SG,
by Olevsky and Froyen [19]. It is, however, known that Almatis, Germany). Raw powders were sieved through a
FAST has an eect on the reactivity between Si and Mo, 100 lm mesh, allowing better packing and reproducibility
as investigated by Munirs group [5,8]. Their results of the green bodies. The FAST graphite tool, with a wall
showed a higher growth rate of the reacted layer in the thickness of 10 mm (graphite type R7710, q = 1.88 g cm3,
presence of current, without changing the reaction mecha- SGL Carbon GmbH, Germany) and an internal diameter
nism. It seems that neither current direction nor pulsing of 20 mm was lled with 7.5 g of the sieved powder and
aects the diusion kinetics. Nonetheless, in insulating mounted into the FAST equipment. To avoid shifting of
materials like alumina no direct current ow through the the mobile die at the beginning of the sintering process a
powder is expected [20]. gauge was used. The interior of the die and the surfaces
Alumina is a well-known ceramic material and is usually of the punches were covered with a compressible graphite
used as a model powder for solid-state sintering experi- foil (type N998, FCT Systeme) with a thickness of
ments. The advantages of using a-Al2O3 powder are its 0.37 mm. This maximized the contact area between the
(i) availability at high purity in a wide range of particle rough powder compact surface and the punches. Further-
sizes and (ii) thermodynamic stability at high temperature. more, this graphite foil had a lubricating eect, reducing
However, under a critical size of about 100 nm metastable the friction between the movable parts and signicantly
transition phases may signicantly aect the sintering reducing temperature inhomogeneities [25,26]. In addition,
behavior when transforming into the a-phase [21]. Zhou the exterior of the die was covered by graphite felt with a
et al. [4] and Shen et al. [22] investigated the evolution of nal thickness of 10 mm to reduce possible temperature
the microstructure of FAST sintered a-alumina as a func- gradients. A similar procedure was used to ll the die in
tion of heating rate. They found that a high heating rate the HP graphite tool (graphite type R7500, q =
leads to a smaller grain size for a given density. Wang 1.78 g cm3, SGL Carbon GmbH, Germany).
et al. [23] focused on microstructure, highlighting how het- The initial sample height was 11 mm, with a relative
erogeneities depend on the position within the sample. green density of 52 2% (measured geometrically). To
More recently, Stanciu et al. [12] found that during the ini- facilitate comparison, the same heating rate of 10 K min1,
tial stage of sintering there was a greater rate of neck sintering temperatures of 1100, 1150 and 1200 C, dwell
growth below 900 C in comparison with microwave and times of up to 2 h and under vacuum (0.4 mbar) were cho-
conventional sintered alumina. sen for the sintering experiments. In the case of FAST a
In the present work FAST and HP experiments were run pulse pattern of 25:5 was adopted for heating. The cooling
under the same conditions: unaxial shrinkage was mea- rate was 10 K min1 until a temperature 150C below the
sured at temperatures in the range of 11001250 C with maximum temperature was reached. The temperature con-
a dwell time of 12 h, a xed heating rate of 10 K min1 trol unit was shut o after this. To investigate densication
and a macroscopic uniaxial pressure of between 15 and behavior assisted by mechanical stress loads ranging
50 MPa. Special care was taken to check the validity of between 4.5 and 15.7 kN were chosen (corresponding to
these experimental parameters in both cases, including tem- 15 and 50 1.6 MPa, respectively). Axial shrinkage was
perature measurement. This enabled the identication of measured at a resolution of 10 lm for FAST and 5 lm
the densication mechanism involved in each process by for HP. Thermal expansion of the machine was taken into
determining the activation energy, stress exponent and account through reference measurements obtained by plac-
grain growth exponent. The heating rate was varied up to ing a dense alumina sample in the pressing tool and sub-
150 K min1 to highlight any possible changes in the den- tracting these measured values from the original sintering
sication mechanism in FAST. Further, microstructure curves. The eect of heating rate on the densication mech-
analysis was carried out to determine any gradient or pref- anism in FAST was also analyzed. Various heating rates
erential orientation of pores and grains, as well as to ratio- from 35 up to 150 K min1 were used, together with a sin-
nalize similarities and dierences between FAST and HP tering temperature of 1200 C, a soak time of 10 min and a
specimens obtained from the macroscopic measurements. stress of 50 MPa.
5456 J. Langer et al. / Acta Materialia 57 (2009) 54545465

The temperature was measured in two dierent ways: in


HP a thermocouple was used, while in FAST a pyrometer
was used. In both cases a hole was drilled in the surround-
ing graphite allowing temperature measurements approxi-
mately 5 mm from the sample. In the FAST set-up the
temperature was measured axially on the surface of the
punch [18]. The pyrometer was not able to measure temper- Fig. 2. (a) Sintered sample with the cutting direction. (b) Sample cut-out
atures under 450 C, limiting control to temperatures showing four representative areas.
above 450 C. The power input was set to 35% of the max-
imum nominal power. As a consequence, initial overheat- lower than the maximum temperature during the sintering
ing up to 600 C could be observed. In HP the experiment. Micrographs were then taken at four dierent
temperature was measured with an S-type thermocouple positions on the sample (Fig. 2b) with a high resolution
at the edge of the center plane perpendicular to the direc- scanning electron microscope (XL 30 FEG, Philips, The
tion of loading of the powder compact. Additionally, ther- Netherlands). The grain size was measured by the linear
mal homogeneity through the sample was critical. To this intercept method using an image analysis program (Lince
end, 1 wt.% copper powder (mean diameter 0.751.5 lm, v. 2.31, Ceramics Group, Technische Universitat Darms-
Alfa Aesar) was mixed with the alumina powder in an inert tadt, Germany), taking at least 2400 grains into account.
atmosphere and sintered. It was veried that this small Finally, additional samples were sintered under 50 MPa
amount of copper metal well below the percolation with a heating rate of 10 K min1 to a density between
threshold did not alter the densication behavior of the 55% and 65% by both methods. They were further charac-
alumina powder. For both set-ups calibration experiments terized at room temperature by means of intrusive mercury
were started at temperatures 20 C lower than the melting porosimetry (Pore Sizer 9320, Micromeritics, USA) and the
point of copper (1084 C) and then increased in 5 C steps. laser ash method (ZAE Bayern, Wurzburg, Germany).
After each experiment a photograph of the sectioned spec-
imen was taken. Fig. 1a shows a HP sintered sample with a 2.3. Macroscopic analysis of sintering
dwell time of 60 s at 1085 C. Fig. 1b shows a schematic of
the behavior of the samples kept at constant temperature During HP and FAST a constant uniaxial mechanical
for dwell times of 190 s. The outer area of the respective load was applied so that lateral shrinkage of the sample
ellipse indicates the area of molten copper, which increases remained negligible. Densication took place only along
with time. These experiments showed that the HP and the thickness direction. As large strains were observed,
FAST sintered samples needed 90 and 60 s at the use of true strain ez was preferred:
1085 C, respectively, for the copper to become completely  
h
molten. Temperature accuracy was better than 5 C, ez ln 1
because specimens heated at 1080 C did not show any h0
melting. The densities of all sintered specimens were mea- where h is the instantaneous sample height and h0 is the ini-
sured by the Archimedes method in water at room tial height of the green body.
temperature. Instantaneous density q can be computed from:
q q0 expez 2
2.2. Microstructural analysis
where q0 is the initial sample density.
In order to obtain representative information over the Densication was enhanced by the compressive loading
whole sample volume specimens of TMDAR alumina were applied during HP and FAST. By limiting grain growth
prepared as follows: a bar was cut out of the middle of the both sintering techniques provided an eective method
sample (Fig. 2a), subsequently ground, nely polished and for investigating densication behavior. When moderate
then thermally etched for 20 min at a temperature 50 C loads are applied and deformation is controlled by a

Fig. 1. (a) Photograph of polished copper/alumina sample (sintered at 1085 C for 60 s by HP). (b) Schematic of copper melting as a function of dwell time
at constant temperature (1085 C).
J. Langer et al. / Acta Materialia 57 (2009) 54545465 5457

diusional process (assumptions validated for alumina by Therefore, the following procedure was adopted to iden-
Coble [27]) the true strain rate can be described by the gen- tify the exponents n and m. If negligible grain growth is
eral equation [15]: assumed and temperature is kept constant the stress expo-
nent n can be calculated using:
dez 1 dq HD  
_ez  m /pa n 3 1 dq
dt q dt G kT ln / n ln/pa G and T fixed 6
q dt
where e_ z is the true strain rate (the inverse is equivalent to
the normalized densication rate), H is a numerical con- To calculate the grain size exponent m, the eective
 stress /pa and temperature T are assumed to be constant,
stant, D D0 exp E RT
a
the diusion coecient of the
rate-controlling species, D0 a pre-exponential factor, Ea is giving:
 
the activation energy, R is the gas constant, G the grain 1 dq
ln / m ln G T and /pa fixed 7
size, k the Boltzmann constant, T the absolute temperature, q dt
/ is the stress intensication factor, pa the uniaxial applied
stress and the so-called stress exponent n and grain size Finally, to calculate the activation energy of the diu-
exponent m depend on the mechanism of densication sion coecient grain size has to be xed:
 
and grain growth, respectively. q_ T
ln / Ea =RT 8
The stress intensication factor describes how the mac- q /pa n
roscopic applied stress is magnied at the microscopic
scale. For a low density body the contact area between par-
ticles where stress is transmitted is small and stresses are 3. Results
locally much higher than the macroscopic applied stress.
Recently, a numerical study by Montes et al. [28] showed 3.1. Characterization of the densication behavior
how critical is the choice of coecient on the eective
stress. By considering an ideal powder aggregate consisting Fig. 3 shows the densication curves of TMDAR sam-
of spherical particles arranged in a simple cubic packing, ples obtained during sintering in FAST and HP equipment
but avoiding the rigorous tensorial stressstrain relation- at two maximum temperatures with all parameters kept
ship at the particle contacts and neglecting friction and wall identical above 450 C (same heating rate, applied stress
constriction, they developed a simplied expression for uni- and dwell time). The densication curves were obtained
axial compression: from Eq. (2) by calculating backwards from the nal den-
sity. The t to the initial geometrically measured density
  1
1q was veried. Curves for each sintering method present a
/ 1 4 similar trend, in particular during the heating ramp, which
1  qM
shows good reproducibility.
where qM is the minimal equilibrium density without defor- First, dierences can be observed between the two pro-
mation, which is assimilated in real packings to q0. As a cesses. Densication for FAST sintered samples occurred
comparison, the formula proposed by Helle [29] for hot at lower temperatures than those hot pressed (from
isostatic pressing was also used: 860 C, about 25 C lower than for HP). Final densities
1  q0 reached with the FAST equipment were between 94%
/ 5 and 99.5%, with a density of 93% already obtained upon
q2 q q0
reaching the isothermal temperature of 1200 C. In con-
In order to properly apply Eq. (4) a uniaxial stress state trast, hot pressed specimens showed a reduced densica-
in the samples has to be ensured. After full densication the tion: only limited densication had taken place at
specimen geometry was a at disc approximately 5.5 mm 1100 C, the isothermal temperature (70%, instead of
thick. Although this nal geometry does not correspond 76% for FAST specimens). The nal density range for
to the standard compression aspect ratio, previous work HP samples lay between 88 and 97%, which illustrates
has shown that a uniaxial stress with these boundary con- how signicant thermal activation is to densication. At
ditions is developed. Wang et al. [26] indeed developed a 1200 C a longer dwell time would be required for HP to
nite element model to investigate homogeneity of the tem- achieve the same densities as obtained by FAST. As a ref-
perature distribution and stress eld components in alu- erence, dry pressed specimens made of the same powder
mina and copper discs (20 mm diameter and 4 mm and sintered freely in air have a nal density of 85% after
height) during FAST. They highlighted the eect of ther- 4 h at 1200 C [30].
mal expansion mismatch between the sample and the die/ In comparison, Fig. 4 shows strain curves for the coarser
punch materials on the stress state in the specimen. Because alumina powder CT 3000 SG sintered by FAST and HP as
alumina and graphite have similar coecients of thermal a function of time. The process parameters previously
expansion, all stress components, except the uniaxial described for the TMDAR powder were used. The sinter-
compressive stress, were negligible during changes in ing behavior was qualitatively similar to that of TMDAR.
temperature. Here initial densication and separation of the curves took
5458 J. Langer et al. / Acta Materialia 57 (2009) 54545465

Fig. 3. Relative density curves obtained for both methods (FAST and HP) at two dierent temperatures as a function of time (heating rate: 10 K min1,
applied stress: 50 MPa and dwell time: 1 h) for TMDAR alumina.

Fig. 4. Curves obtained for both methods (FAST and HP) at 1200 C as a function of time (heating rate 10 K min1, applied stress 50 MPa, dwell time
1 h) for CT 3000 SG.

place at 965 C. Final densities were 89% for FAST and the dwell time). Dq reaches a maximum at 1200 C and
87% for HP. The higher temperature needed to initiate den- decreases with dwell time, because signicant densication
sication is easily explained by the dierences in average takes place with HP. It can be seen that Dq is larger for the
starting particle sizes (a ratio of about 3 between CT ner alumina TMDAR (Dq = 0.076) than the coarser
3000 SG and TMDAR). However, it seems that FAST (Dq = 0.027) CT 3000SG.
induces a smaller positive eect on densication for the Further analysis was done on TMDAR specimens sin-
coarser alumina. Further analysis can be made by plotting tered by FAST and HP at 1150 C for a maximum dwell
in Fig. 5 the dierential relative density Dq = q(FAST) time of 90 min with densities lower than 95%. Fig. 6
q(HP) of the FAST and HP sintered powders as a function displays the logarithm of the normalized densication rate
of temperature (during the heating ramp) and time (during as a function of the logarithm of the eective stress (the
J. Langer et al. / Acta Materialia 57 (2009) 54545465 5459

Fig. 5. Relative dierential density Dqrel between FAST and HP sintered alumina powders TMDAR and CT 3000 SG.

product of applied stress multiplied by intensication fac- age value of n using the simplied Eq. (4) was 1.04, whereas
tor). It has been found that minimal grain growth occurs with Eq. (5) it was 1.06. A value of 1 is associated with lat-
up to 80% density for HP and 85% density for FAST. As tice diusion, grain boundary diusion or viscous ow [15].
expected from the proportional relationship in Eq. (6), a Due to the high purity of the investigated alumina, no
linear behavior was observed for all chosen densities. For glassy phase formed during sintering, thus allowing us to
both FAST and HP processes a stress exponent of n  1 reject the third possibility. It has previously been observed
was calculated. No signicant dierences were observed that submicron powder densies by grain boundary diu-
for the stress exponent values obtained using the models sion and not lattice diusion at temperatures up to
of Montes et al. [28] and Helle et al. [29]. In the range of 1400 C [31]. Validation of this assumption can be made
applied stress and density investigated deviations in the by computing the grain size exponent as well as the activa-
results lie within the error bars in Figs. 6 and 7. The aver- tion energy during sintering.

Fig. 6. Plot of the logarithm of the densication rate as a function of the eective pressure /pa at dierent relative densities.
5460 J. Langer et al. / Acta Materialia 57 (2009) 54545465

for the HP sintered alumina were obtained. These values


correspond to the well-known value of 440 45 kJ mol1
for grain boundary diusion in alumina [31].
In addition, Fig. 9 shows a plot of the relative density as
a function of temperature for dierent heating rates
obtained for specimens sintered in the FAST equipment
(all other parameters were kept constant). It can be seen
that the starting temperature for densication was shifted
to higher temperatures with increasing heating rate. This
is most likely due to the dierence between the measured
temperature and the real temperature seen for the sample.
With increasing heating rate the temperature dierence
became larger, highlighted by nite element method tem-
perature gradient simulations [16,17]. Nevertheless, the
slopes of the densication curves in their linear regime
Fig. 7. Plot of the logarithm of the densication rate as a function of the (Fig. 9) are similar, as shown in Table 1. The lowest values
logarithm of the normalized grain size.
in this series for the two largest heating rates can be attrib-
uted to the thermal bias. A constant slope is typically
Fig. 7 shows the logarithm of the densication rate as a observed for constant heating rate experiments [15] and
function of the logarithm of the grain size normalized by indicates that there is no change in the densication
the grain size measured on samples with a relative density mechanism, provided that the temperature bias does not
of 65%. Temperature and eective stress were kept con- depend on temperature in the range considered. This
stant (1100 C and 185 MPa). The complete sintering tra- means that grain boundary diusion remains the primary
jectory (grain size as a function of density) is presented in driving force for densication at higher heating rates (up
Fig. 12. As anticipated from the proportional correlation to 150 K min1). All of the samples showed a similar rela-
in Eq. (7), a linear behavior can be observed and the grain tive density of 99% after the sintering process and a dwell
size exponent m can be calculated. For both FAST and HP time of 10 min.
a grain size exponent of m  3 was obtained. This value is
characteristic of grain boundary diusion, whereas a value 3.2. Analysis of the microstructure and sintering trajectory
of 2 would have indicated lattice diusion [15].
Fig. 8 shows the logarithm of the normalized densica- The local grain size was measured at dierent areas in
tion rate as a function of the reciprocal of absolute temper- the sample, as shown in Figs. 2 and 10a and b shows typical
ature (Arrhenius plot). Using Eq. (8), activation energies of microstructures of FAST and HP sintered TMDAR alu-
420 35 kJ mol1 for the FAST and 430 50 kJ mol1 mina under the same conditions, which are comparable

Fig. 8. Arrhenius plot of FAST and HP sintered TMDAR alumina at 1100 C, 1150 C and 1200 C with a heating rate of 10 K min1 and a dwell time of 1 h.
J. Langer et al. / Acta Materialia 57 (2009) 54545465 5461

Fig. 9. Comparison of dierent heating rates of FAST sintered TMDAR with a maximum temperature 1200 C, a stress of 50 MPa and a dwell time of
10 min.

the reference grain size (at 65% density). This result was
Table 1
expected, as it has been shown that the temperature distri-
Slopes of the densication curves in the linear regime obtained from
Fig. 9. bution in the sample was homogeneous and a low heating
rate was used. High heating rates (several hundreds of
Heating rate (K min1) Slope (103 K1)
degrees per minute) could lead to thermal gradients and
10 1.7 0.2
subsequently gradients in microstructure, as previously
35 1.6 0.2
50 1.7 0.2 observed [23].
100 1.5 0.2 The average grain size aspect ratio Gy/Gx was 1, which
150 1.4 0.3 shows no marked grain anisometry and preferential grain
orientation, as can be seen in Fig. 10. This low degree of
anisotropy had already been observed in hot pressed alu-
with those obtained by other researchers [4,32]. No local mina at similar stress levels with a sintering temperature
variation in density was observed. of 1450 C, as long as signicant grain growth did not
As highlighted in Fig. 11, samples of a given density had occur (grain size smaller than 2 lm) [33].
a constant grain size. It was found that grain size did not Fig. 12 shows the sintering trajectory of FAST and HP
depend on the measurement position. The measured grain sintered alumina, with normalized grain size plotted as a
size was homogeneous throughout the sample for the range function of relative density. No dierences can be found
of densities investigated by both FAST and HP. There was between FAST and HP samples. The grain size remained
a limited amount of grain growth, as the maximal nal constant up to 90% relative density. A steep increase was
grain size (at 99% density) was no larger than 2.5 times then observed starting from 95%, explaining the dierences

Fig. 10. High resolution SEM pictures of polished and thermal etched (a) FAST and (b) HP sintered alumina at 1200 C, with a stress of 50 MPa, a
heating rate of 10 K min1 and a dwell time of 1 h.
5462 J. Langer et al. / Acta Materialia 57 (2009) 54545465

Fig. 11. Grain size as a function of the position in the sample (TMDAR).

between the micrographs presented in Fig. 10. As a com- 4. Discussion


parison, the sintering trajectory of specimens sintered freely
[30] reveals that grain growth started earlier: this can be Direct comparison of hot pressing and eld-assisted sin-
rationalized by the additional driving force for densica- tering for electrically insulating alumina has shown that
tion provided by the applied mechanical pressure. The densication is mainly controlled in both processes by
microstructure of specimens sintered at dierent heating grain boundary diusion. The stress and grain size expo-
rates was not analyzed. Because the grain size is dependent nents (n = 1 and m = 3), as well as the activation energy
on the real local temperature it is dicult to compare spec- (430 kJ mol1) determined from the macroscopic behavior,
imens with increasing heating rate due to the increase in agree remarkably well. This is further conrmed by grain
thermal bias between the sample and measured tempera- size measurements done at the microscopic scale, which
tures. If the dwell time is too short the sample may not even show similar sintering trajectories for HP and FAST.
be exposed to the maximum temperature [22,34]. Therefore, the hypothesis of possible accelerated diusion

Fig. 12. Normalized grain size as a function of relative density measured on FAST, HP and conventionally sintered TMDAR alumina (heating rate
10 K min1, stress 50 MPa and dwell time between 1 and 2 h).
J. Langer et al. / Acta Materialia 57 (2009) 54545465 5463

processes due to the application of an electrical eld or pro- of 110130 C at the beginning of the FAST sintering pro-
motion of another densication mechanism, such as diu- cess. This phenomenon could inuence neck building and
sion through a melted surface or plastic deformation, neck growth at the beginning of the heating cycle and the
during FAST as proposed by other researchers [79] do subsequent nal densities, even though the densication
not seem to hold for the present experiments. It must be mechanisms are identical to those observed in this study.
noted that the sintering parameters utilized here were all The eect of this transient temperature excess seems to be
identical except for the heating rate under 450 C, which larger for smaller particles than for coarser ones, as the dif-
was, as previously mentioned, uncontrolled for FAST. fusion processes start at lower temperatures for ner pow-
All FAST sintered samples began to densify at lower ders. As a consequence, attention has to be focused on the
temperatures and, therefore, had a higher nal density than rst stage of sintering, prior to massive densication.
HP (corresponding to a temperature shift of 25 C for the Microscopic models describing the normal stress distribu-
TMDAR powder). Similar results were also observed by tion along a grain boundary between two particles show
Oh et al. [35] with a coarser alumina (mean particle size that the intensity of the sintering stress decreases with neck
0.2 lm) and a heating rate of 30 K min1. In order to growth, leading to a decrease in the densication driving
explain this moderate decrease in the initiation of densica- force and, consequently, the plating rate at the particle
tion temperature two possibilities may be considered: (i) scale and the densication rate at the macroscopic scale
temperature is locally higher than the macroscopic mea- [15]. Fast ring schedules are justied by this fact, promot-
sured temperature and (ii) smaller inter-particle necks are ing densifying mechanisms at the expense of the neck
present. coarsening phenomenon. Therefore, in order to explain
The rst point has already been addressed and is dicult the earlier densication onset with FAST we propose that
to assess experimentally. In low permittivity materials such overheating in the present FAST experiments did not lead
as alumina, electrical discharges occurring in pores under a to a simple growth of the inter-particle contacts, but more
high electric eld is not thought to increase the local tem- likely to an increase of the number of inter-particle necks,
perature more than a few degrees centigrade [36], although which had a benecial eect on densication rate. Better
this may be sucient to substantially aect mass transfer mechanical properties such as bending strength [35,39]
through thermal gradients. The possible role of the powder measured for FAST alumina specimens are in agreement
has been described by Chaim [37], who proposed that for with this hypothesis.
insulating dielectric materials the particle size and mor- Unfortunately, direct observation of the evolution of
phology have an eect on densication under an applied selected inter-particle contacts cannot be done in situ by
electric eld. It may be possible that as the surface to bulk transmission electronic microscopy and quantitative char-
conductivity ratio increases as the inverse of particle size acterization through SEM micrograph analysis is dicult
accumulation of electrical charges at the surface induces and sensitive to artifacts. Therefore, measurements were
surface breakdown (especially when particles are contami- carried out by mercury porosimetry on TMDAR speci-
nated with adsorbed water and a low oxygen partial pres- mens with relative densities of 55% and 60% heated under
sure is maintained, such as in this study). A discharge the same conditions until the required density was reached
breakdown through particles is not possible in our opinion, during the heating ramp. Fig. 13 shows the measured mer-
because the global voltage applied to the compact sample cury volume as a function of the mean pore diameter on a
was divided among all particles. Even at temperatures of logarithmic scale for FAST and HP sintered samples. For a
several hundreds of degrees centigrade and a porosity better comparison, pieces of same mass (0.45 g) were cut
above 30%, where the electrical strength of compact alu- out of the specimens. These curves conrm that the pore
mina may be under 100 MV m1 (0.1 V nm1), the elec- size and accumulated pore volume decrease as density
trical potential dierence seen for a 100 nm particle increases from 55% to 60%. For both processes a monomo-
aligned in a 10 mm thick specimen loaded under 15 V is dal pore size distribution is observed, with a homogeneous
1.5  104 V. Under the prevailing sintering conditions a pore size under 80 nm. More interestingly, there is a slight
maximum voltage of 5 V was applied by the FAST set- dierence between both types of specimens for the two den-
up. In addition, a large local elevation of temperature sities investigated: the mode of the pore size distribution is
induced by such phenomena could not be seen in copper larger for the samples produced by HP by about 5% than
particles observed under scanning electron microscopy for the FAST alumina. Of course, computation of the pore
(SEM) in the present specimens. In support of these con- equivalent size requires that we assume that pores are like
clusions, recent work done by Hulbert et al. [38] proved capillary tubes of constant and circular cross-section. In
the absence of sparking events by atomic emission spec- reality pores are more ink bottle shaped. Therefore, mer-
troscopy and ultrafast in situ voltage measurements on cury porosimetry gives a measure of the pore smallest sec-
electrically conducting or insulating materials (including tion. This, unfortunately, does not give a direct indication
alumina). of the neck area.
Therefore, a plausible explanation for the dierences Additional information was obtained by means of the
observed between the FAST and HP densication curves laser ash method [12,40]. Samples heated under the same
could be the previously mentioned transient overheating conditions with densities varying from 54% to 65% were
5464 J. Langer et al. / Acta Materialia 57 (2009) 54545465

Fig. 13. Mercury intrusion as a function of the mean diameter of the pores of FAST and HP sintered samples (TMDAR, heating rate 10 K min1, stress
50 MPa, dwell time 10 s, 750 C to reach 55% density for FAST and HP, 925 C for FAST and 950 C for HP to reach 60% density).

characterized by cutting out 10 mm radius discs from tering process (temperatures below 800 C) the total neck
FAST and HP sintered TMDAR and grinding them down area for FAST sintered samples is higher than for HP spec-
to a thickness of 0.5 mm. Fig. 14 shows that the measured imens. In the absence of proof of electric eld/current
thermal diusivity increases as a function of the relative assisted mechanisms this fact may be explained by the tem-
density. This is due to the microstructural changes that perature overshoot in FAST when the pressing tool and the
occur on the formation of sintering necks during the rst powder are heated from room temperature to 450 C. At
stage of sintering, before massive densication takes place. this stage of the sintering process the FAST sintered sam-
According to simple isotropic models the thermal diusiv- ples experienced dierent heat treatment in comparison
ity is roughly proportional to the neck radius [12,40] for a with the HP samples, due to a rapid transient overheating.
given unit cell with a constant number of inter-particle con- Enhanced by the non-spherical, elongated and distorted
tacts. In addition, thermal diusivity is higher at lower den- shape of the TMDAR alumina particles [24], we could
sities for FAST sintered samples in comparison with the envisage a larger number of contacts per particle without
HP ones (about 30% for a density of 54%). This dierence a signicant increase in density. This could be a benet
reduces as density increases, so that values are similar for a of the FAST process in this situation.
density of 65%. This means that at the beginning of the sin-
5. Conclusion/summary

Processing of submicron pure alumina by hot pressing


and eld-assisted sintering with the same parameters (heat-
ing rate, mechanical load, temperature and atmosphere)
has shown that densication is enhanced by FAST from
the beginning of the temperature cycle. However, analysis
of sintering curves reveals that densication for both meth-
ods is controlled by grain boundary diusion. A heating
rate increase up to 150 K min1 does not change the densi-
cation mechanism. Furthermore, the microstructure and
average grain size as a function of density are similar at
higher relative densities. To reach the same density as the
FAST sintered samples in the hot press the dwell time must
be increased by a few minutes. FAST specimens begin to
densify at a lower temperature, which may be due to rapid
Fig. 14. Thermal diusivity as a function of the relative density for FAST transient overheating in the initial stage, which increases
and HP sintered samples (TMDAR, heating rate 10 K min1, stress
50 MPa, dwell time 10 s, 750 C to reach 54% density for FAST and HP,
the global inter-particle area but not the average contact
900 C to reach 57% density for FAST, 950 C to reach 60% density for size. These results apply to both investigated alumina pow-
HP, 1030 C for FAST and 1060 C for HP to reach 65% density). ders, the coarser one showing a less marked benecial eect
J. Langer et al. / Acta Materialia 57 (2009) 54545465 5465

of FAST, probably due to a lower sensitivity to overheat- [12] Stanciu L, Quach D, Faconti C, Groza JR, Raether F. J Am Ceram
ing. An improvement in the FAST measurements would Soc 2007:271622.
[13] Coble RL. J Appl Phys 1970:4798808.
be to accurately control the heating rate from room tem- [14] Vieira JM, Brook RJ. J Appl Phys 1984:2459.
perature up to 450 C. Moreover, a specimen could be sin- [15] Rahaman NM. Ceramic processing and sintering. New York: Marcel
tered rst in the FAST equipment up to 55% density and Dekker; 2003. p. 51436.
subsequently transferred to the hot press and densied fur- [16] Zavaliangos A, Zhang J, Krammer M, Groza JR. Mater Sci Eng A
ther. We could expect a better densication compared with 2004:21828.
[17] Anselmi-Tamburini U, Gennari S, Garay JE, Munir ZA. Mater Sci
a specimen sintered only in the HP set-up. Unfortunately, Eng A 2005:13948.
all partially sintered samples cracked when the minimal [18] Vanmeensel K, Laptev A, Hennicke J, Vleugels J, Van der Biest O.
mechanical load (5 kN) inherent to the hydraulic system Acta Mater 2005:437988.
was applied. Little is known about the sintering mecha- [19] Olevsky EA, Froyen L. Scripta Mater 2006:11758.
nisms of other oxide powders under FAST conditions. [20] Carmen CM, Tai-Il M. J Am Ceram Soc 2008:344850.
[21] Mishra RS, Risbud SH, Mukherjee AK. J Mater Res 1997:1869.
Thus, a similar thorough analysis has yet to be performed, [22] Shen Z, Johnsson M, Zhao Z, Nygren M. J Am Ceram Soc
for example, on ionic conductive and semi-conductive 2002:19217.
materials showing temperature-dependent electrical con- [23] Wang SW, Chen LD, Hirai T, Guo J. J Mater Res 2001:35147.
ductivity. [24] Guillon O, Weiler L, Rodel J. J Am Ceram Soc 2007:1394400.
[25] Zhang D, Zhang L, Guo J, Tuan WH. J Am Ceram Soc
2006:6803.
Acknowledgement [26] Wang X, Casolco SR, Xu G, Garay JE. Acta Mater 2007:361122.
[27] Coble RL, Ellis JS. J Am Ceram Soc 1963:43841.
This work was nancially supported by the Deutsche Fors- [28] Montes JM, Cuevas FG, Cintas J. Comp Mater Sci 2006:32937.
[29] Helle AS, Easterling KE, Ashby MF. Acta Metall 1985:216374.
chungsgemeinschaft (Emmy Noether Program GU993-1/1).
[30] Zuo R, Aulbach E, Rodel J. Acta Mater 2003:456374.
[31] Wang J, Raj R. J Am Ceram Soc 1991:11725.
References [32] Kim BN, Hiraga K, Morita K, Yoshida H. Scripta Mater
2007:60710.
[1] Munir ZA, Anselmi-Tamburini U, Ohyanagi M. J Mater Sci [33] Roy JF, Descemond M, Brodhag C, Thevenot F. J Eur Ceram Soc
2006:76377. 1993:32533.
[2] Nygren MS, Chen Z. Key Eng Mater 2004:71924. [34] Stanciu LA, Kodashund VY, Groza JR. Met Mat Trans A
[3] Zhou Y, Hirao K, Yamauchi Y, Kanzaki S. J Eur Ceram Soc 2001:26328.
2004:346570. [35] Oh ST, Tajima KO, Ando M, Ohji T. J Am Ceram Soc 2000:13146.
[4] Zhao Z, Buscaglia V, Bowen P, Nygren M. Key Eng Mater [36] Moulson AJ, Herbert JM. Electroceramics-materials, properties and
2004:2297300. applications. New York: John Wiley; 2003. p. 24385.
[5] Chen W, Anselmi-Tamburini U, Garay JE, Groza JR, Munir ZA. [37] Chaim R. Mater Sci Eng A 2007:2532.
Mater Sci Eng A 2004:1328. [38] Hulbert DM, Anders A, Dudina DV, Andersson J, Jiang D, Unuvar
[6] Anselmi-Tamburini U, Garay JE, Munir ZA. Mater Sci Eng A C, Anselmi-Tamburini U, Lavernia EJ, Mukherjee AK. J Appl Phys
2005:2430. 2008:033305.
[7] Groza JR, Garcia M, Schneider JA. J Mater Res 2001:28692. [39] Jayaseelan DD, Kondo N, Brito ME, Ohji T. J Am Ceram Soc
[8] Bernard-Granger G, Guizard C. Acta Mater 2007:3493504. 2002:2679.
[9] Chaim R, Marder-Jaeckel R, Shen JZ. Mater Sci Eng A 2006:748. [40] Homann R, Hahn O, Raether F, Mehling H, Fricke J. High Temp
[10] Olevsky EA, Kandukuri S, Froyen L. J Appl Phys 2007:114913. High Press 1997:70310.
[11] Angerer P, Yu LG, Khor KA, Krumpel G. Mater Sci Eng A 2004:12.

Potrebbero piacerti anche