Chemical Engineering Science 63 (2008) 3802 -- 3809

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: w w w . e l s e v i e r . c o m / l o c a t e / c e s

Phenomenological study and modelling of wick debinding

I.M. Somasundram a , A. Cendrowicz b , D.I. Wilson a , M.L. Johns a,∗
a Department of Chemical Engineering, New Museums Site, Pembroke St, Cambridge CB2 3RA, UK
b Ross Ceramics Limited, Derby Road, Denby, Derbyshire DE5 8NX, UK

A R T I C L E I N F O A B S T R A C T

Article history: Wick debinding employs capillary suction (via a surrounding wicking powder) to remove the liquid
Received 4 February 2008 binder phase from powder injection moulded parts (known as a compact). Experimental measurements
Received in revised form 14 April 2008 of binder distribution within the compact during debinding highlight flaws in previous wick debinding
Accepted 23 April 2008
models. The spatially uniform distribution of binder observed consistently during debinding indicates that
Available online 3 May 2008
it is removed in order of pore size regardless of location in the compact. A model is proposed which gives
Keywords:
good agreement with 1-D experimental data of binder distribution. Key parameters of the model are the
Debinding permeability of the wicking powder and the relationship between the capillary pressure, saturation and
Wicking relative permeability of the compact.
Capillary pressure © 2008 Elsevier Ltd. All rights reserved.
Ceramic injection moulding

1. Introduction
Powder injection moulding (PIM) is used to manufacture preci-
sion high density items using extrusion (Mutsuddy and Ford, 1995).
The injected material is a paste formed by mixing a powder with a
binder phase: the latter is typically solid at ambient conditions but is
liquid at elevated temperature. The paste is injection moulded above
the binder melt temperature to form the green product, the resul-
tant item removed from the mould and cooled. Before the powder
can be sintered to form the final product, the binder phase must be
removed without deforming the item.
This stage of the PIM process, known as debinding, can take place
over several days in order for the binder to be removed in a con-
trolled manner. There are several debinding methods (Lewis, 1997),
the most common being to burn off the binder, leaving the solid pow-
der. This can be performed in a single operation using a controlled
temperature-time ramp. Solvent extraction is also used, where a
large part of the binder is removed by dissolution in a suitable sol-
vent, with the remaining binder subsequently removed by burn-out.
A third method is wick debinding, where the green compact item
is placed in contact with wicking powder and the temperature in-
creased so that the binder melts. The wicking powder then extracts
the liquid binder by capillary action as a result of the larger capillary
suction pressures of the pores in the wicking powder compared to
the compact. As well as removing the binder in a potentially more
controlled manner than burning out, the wicking powder is often
∗ Corresponding author. Tel.: +44 1233 334767; fax: +44 1233 362017.
E-mail addresses: mlj21@cam.ac.uk, mlj21@cheng.cam.ac.uk (M.L. Johns).
used to embed the item, allowing extraction of binder from all di-
rections as well as affording the item significant structural support
(German, 1987). These effects combine to improve the product qual-
ity. Defects can still occur, however, especially in parts with complex
shape.
The understanding of the fundamental mechanisms of wick de-
binding is not well developed, and much of the published work in
the literature builds on the model published by German (1987) with
later evidence (Vetter et al., 1994a) indicating its limitations. Here
we report an experimental investigation of wick debinding to elu-
cidate the governing mechanisms and thereby develop a numerical
predictive model of the process. The study employs both a commer-
cial paste and wick debinding powder; the results are thus expected
to apply to many industrial systems.
2. Background---modelling of wick debinding
German (1987) suggested that wick debinding follows one of two
modes, namely compact-controlled and wick-controlled debinding.
The former was preferable as the latter was considered to be inef-
ficient due to the restriction of the wicking powder on debinding
times. For the compact-controlled case to occur, the permeability of
the wicking powder, Kw , should be higher than that of the compact,
Kc . However, German also stated that for the wicking powder to pro-
vide capillary suction, its pores should be smaller than those in the
compact. Where the surface energies of the two powders are sim-
ilar, this ordinarily leads to Kc > Kw , thereby contradicting the first
condition.
German nevertheless proposed a quantitative model of wick de-
binding where a compact is in contact with wicking powder as shown

0009-2509/$ - see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2008.04.040
 I.M. Somasundram et al. / Chemical Engineering Science 63 (2008) 3802 -- 3809 3803

Air Zone
Compact
1
height H
l
Binder

Wicking powder

x
Fig. 1. Schematic of wick debinding mechanism proposed by German (1987).

schematically in Fig. 1. Debinding was modelled as the binder mov-

ing as a continuous body of liquid from its original location within
the compact into the wicking powder, leaving the atmosphere to
occupy the vacated pore space in the compact. The rate of removal
Equation (6)
was modelled as flow through a porous medium following Darcy's
law, where the driving force is the difference in capillary pressures Experiments
(Pc − Pw ), viz.
tD
Kc (Pc − Pw ) Debinding time
u=− (1)
L
Fig. 2. Comparison of Eq. (6) with typical experimental results (such as those
where u is the superficial velocity of the binder,  its viscosity, and reported by Patterson and Aria, 1989; Waikar and Patterson, 1986).
L is the height (or length) of binder in the compact. The capillary
pressure is determined by the surface tension, , the contact angle 
and the pore neck diameter dp as presented in Liu et al. (1986):
Bulk
binder
10 cos  Compact
P= (2) height H
dp
Air zones
while the permeability is based on the following approximation for
packed beds of voidage ; (German, 1987) Wicking
powder
3c dc2
Kc = (3)
90(1 − c )2
Fig. 3. Schematic of wick debinding mechanism proposed by Vetter et al. (1994b).
A volume balance gives

dL Kc (Pc − Pw ) 4 dc (dc − dw )

−c =u=− = c (4) Although Lograsso and German (1990) claimed to confirm Eq. (6)
dt L 9dw (1 − c )2 L experimentally, Vetter et al. (1994a) found that this model under-
The time required to remove all the binder from the compact, tD , is estimated the debinding time by two to three orders of magnitude.
given by Furthermore, German's model prediction of tD is of limited use as
complete removal of binder from a compact is unlikely because there
 H  often exists a minimum x in the form of an irreducible liquid binder
3 dc (dc − dw ) 0
L dL = − c dt (5) saturation. It is impossible to remove all the binder from a porous
0 9dw (1 − c )2 tD
network due to: (i) small pores retaining binder that have a higher
which yields capillary suction pressure than the wicking powder and (ii) snap-off
where ganglia form in the porous network (Dullien, 1972), creating
(1 − c )2 H 2 dw small pores which cannot be drained due to the surrounding larger
tD = 4.5 (6)
3c dc (dc − dw ) pores having been completely drained; there is thus no route for
binder transport to the surface.
where H is the compact height. Despite these criticisms, German's model is frequently used as
German reported that this model gave good agreement with the the basis for research in wick debinding. Chen and Hourng (1999)
experimental data of Waikar and Patterson (1986) which showed a conducted finite element numerical simulations based on Eq. (6) that
linear relationship between the square of the total mass loss against agreed with the theory as expected. Their 2D simulations showed
time. However, he assumed the total debind time (as predicted by binder moving from a compact into wicking powder with both lead-
his model) could be equated to the time for partial debinding (as ing and trailing fronts as in Fig. 1. However, their results were
measured experimentally). German equated the mass loss with H, not validated experimentally. Lin and Hourng (2005) simulated the
which was subsequently critiqued by Vetter et al. (1994a) since the movement of binder from a compact into wicking powder by Ger-
mass of binder removed does not relate directly to H for partial man's binder removal mechanism using a pore network approach.
debinding. Fig. 2 schematically illustrates this discrepancy between They compared the predicted location of the binder front with cor-
German's model and typical experimental results, such as those of responding experimental observation of the front in the wicking
Patterson and Aria (1989) and Waikar and Patterson (1986), in terms powder. Although their results demonstrated reasonably good agree-
of x verses debind time, where x is the fraction of binder removed ment, they again assumed a trailing binder front leaving the compact
from the compact. on completion of debinding.
3804 I.M. Somasundram et al. / Chemical Engineering Science 63 (2008) 3802 -- 3809
German's model and subsequent adaptations assume that there
is a route available for air to enter the compact behind the departing
binder. This is unlikely to be the case in industrial practice where
the compact is buried in the wicking powder and binder is removed
from the compact in all directions. Vetter et al. (1994b) proposed an
alternative model in which air enters the compact through channels
in the wicking powder, as shown in Fig. 3. The binder is removed
counter-current to the air flow and is fed from the bulk binder by
arteries through to the wicking powder. Vetter et al. (1994b) devel-
oped an appropriate relationship for the debinding time in a similar
fashion to Eq. (6) except that they included the wicking powder
permeability. Their model provided a better fit to the experimental
debinding profile shown in Fig. 2. However, the description pre-
sented in Fig. 3 is not necessarily intuitive. Although their ex-
perimental results showed that there was less binder near the
compact/wick interface than in the centre of the compact, the binder
distribution was not described in any detail.
Bao and Evans (1991) reported that instead of there being distinct
binder rich and binder lean regions in the compact during wick de-
binding, the binder distribution decreases uniformly along the length
of the compact. They attributed this observation to the redistribu-
tion of binder due to local pressure differences within the compact,
suggesting that the capillary pressure of the compact will increase
as the saturation decreases.
Improving the efficiency of debinding operations requires a
quantitative understanding of the process and the influence of vari-
ous process parameters. The importance of local binder content in
defect generation means that an improved model of binder distri-
bution during debinding will help predict and hence avoid such
developments. With the exceptions of Vetter et al. (1994b) and Bao
and Evans (1991), the spatial distribution of binder was not inves-
tigated experimentally as it seems the emphasis of most work was
to establish suitable debinding times.
Our work reports an experimental investigation of wick debind-
ing in nominally one-dimensional geometries and the identifica-
tion of a quantitative model based on evolving capillary pressures
and saturation which will be applied in due course to 2- and
3-dimensional wick debinding systems. The observed phenomenol-
ogy indicates that models based on evolving capillary pressures and
saturation offer a more reliable and versatile approach.
Table 1
Properties of ceramic paste
Property Value
Mean particle size 47 m
Powder bulk density 1450 kg m−3
Powder porosity, c 0.38
Solid binder density 900 kg m−3
Liquid binder density 800 kg m−3
Binder viscosity,  at 100 ◦ C 60 mPa s
Mass fraction of binder in paste 0.18
Wicking powder
Al foil Compact
Liquid binder in
wicking powder
Fig. 4. Wick debinding experimental setup.
3. Methods and materials
Debinding experiments were performed using a commercial ce-
ramic paste based on a zirconia powder and a paraffin wax binder
as summarised in Table 1. The wicking powder was a commercial
alumina based powder. DSC analysis of the binder indicated that it
softened around 70--80 ◦ C, and its viscosity was confirmed as New-
tonian at 100 ◦ C using a Bohlin CVO 120 controlled stress rheometer.
Thermogravimetric analysis of the binder indicated that it started to
evaporate at 130 ◦ C. Cylindrical compacts of length 40 mm and di-
ameter 16 mm were generated by compaction in a cylindrical mould
using a strain frame at 70 ◦ C. The uniformity of binder distribution
along the compacts was tested by sectioning the compact along its
length and ashing the 5 mm slices to remove all the binder by heating
in air at 850 ◦ C. The percentage mass of binder across the compact
was found to be uniform within the accuracy of the measurements.
Compacts were weighed and placed inside borosilicate glass tubes
of length 160 mm and id. 16.1 mm as shown in Fig. 4. The com-
pacts were located halfway along the tubes and wicking powder was
loaded at either end, tapped to ensure good contact with the com-
pact, and held in place with crushed aluminium foil.
Several tubes were placed simultaneously into a pre-heated oven
at 100 ◦ C and heated for debinding times ranging from 30 min to 14
days. This temperature was selected as it is just below the tempera-
ture at which the binder begins to evaporate. An unsteady state heat
transfer calculation for an infinite cylinder indicated that it would
take approximately 7 min for the compact to reach the oven temper-
ature (assuming a thermal diffusivity of 0.56 m2 s−1 and including
the resistance of the tube wall to heat transfer, both of which were
ascertained from separate experimental tests). After the desired pe-
riod the contents of the glass tubes were removed carefully. The
compact was re-weighed and the binder distribution determined by
the method described above for verifying the original binder dis-
tribution. Binder content is reported in terms of mean saturation,
calculated by the following method.
3.1. Calculating the saturation of a sample from the mass of binder
removed
Initially the solid binder partially occupies the porosity of the
compact, c = 0.38. The mass fraction of binder in the paste is ap-
proximately 0.18. Therefore the fraction of the bulk volume of the
paste occupied by molten binder is calculated as
mb mb Vbl bl
= 0.18 = = (7)
mT mb + mp Vbl bl + Vp p
where subscript bl refers to molten (liquid) binder and p refers to
the compact powder. Therefore:
0.18p
Vbl = Vp = 0.40Vp > c Vp (8)
0.82bl
Wicking powder
Al foil
Glass tube wall
 I.M. Somasundram et al. / Chemical Engineering Science 63 (2008) 3802 -- 3809 3805

For the paste studied here the liquid binder occupies more space 1.0
than available in the voids of the compact powder and therefore
on melting a certain fraction of the binder seeps into the wicking 0.9
powder. The calculation in Eq. (9) shows that this fraction is 5%.
Because of this effect the binder saturation is calculated based on the 0.8
mass of binder occupying the compact after this expansion period

Mean Saturation
prior to wick debinding. 0.7

c 0.38
mb,i = m = m = 0.95mb (9) 0.6
0.40 b 0.40 b

The binder saturation S is then calculated from the mass of binder 0.5
removed mb,r as
0.4
mb,i − (mb,r − 0.05mb )
S=
mb,i 0.3
0.95mb − (mb,r − 0.05mb ) mb − mb,r
= = (10) 0.2
0.95mb 0.95mb
0 2 4 6 8 10 12 14 16
4. Results and discussion Time / days

Fig. 5. Mean saturation of compacts during wick debinding at 100 ◦ C.

Fig. 5 shows the change in total binder content, presented as
binder saturation, S over time. The mean saturation here is equivalent
to 1 − x in Fig. 2, showing that the data follow the trends reported 1.0
in previous studies. The rate of debinding decreases as debinding 0.5 h
progresses. Vetter et al. (1994b) explained this decrease as owing to 0.9 1h
an increase in the artery length. Furthermore, Fig. 5 shows that the 3h
equilibrium saturation of this system is around 0.37. 0.8 6h
The axial distribution of binder within the compact under these 8h
0.7
conditions is plotted in Fig. 6. Averaging the binder saturation over
Saturation

the compact gave mean saturations consistent with the values re-
ported in Fig. 5. The plot shows that the binder is not located in two
regions (a fully saturated and an air zone) as suggested by German
(1987). Rather, the distribution is generally uniform across the com-
pact, and remains uniform as the saturation decreases. The void space
generated must be occupied by vapour and air. Given the short heat-
ing time of 7 min it is most unlikely that the binder could redistribute
itself in the time associated with removal from the oven and cool-
ing to the binder solidification point. Separate tests indicated that
there was an even distribution of binder with height, indicating that
gravity-driven flow was not significant. Uniform binder distributions
were observed during interrupted wick debinding experiments by
Bao and Evans (1991) using similar techniques. There is arguably a
small drop in binder saturation towards the edge of the compact in
Fig. 6, which agrees with the observation of Vetter et al. (1994b).
However, the drop in saturation is not as dramatic as that suggested
by their model, depicted in Fig. 3.
The uniformity in binder saturation indicates that the binder must
be removed at a uniform rate along the length of the compact. A typi-
cal ceramic powder such as the one used here contains a distribution
of particle sizes and therefore results in a continuous distribution
of pore sizes in the compact. The most reasonable explanation for
the consistent uniformity of binder distribution is that large pores,
which offer the least resistance to capillary suction, are evacuated
first along the entire length of the compact and that progressively
smaller pores are subsequently evacuated thereafter. This explana-
tion is used by Shaw (1986) and many others for the reverse process
of liquid entering a porous medium.
Mercury porosimetry was performed on slices of debinded com-
pacts to see if changes in evacuated pore size distribution and air
porosity could be tracked. The porosity data supported the above hy-
pothesis but the pore size distribution did not show any consistent
variation. This null result can be attributed to the destructive nature
of mercury porosimetry as a pore size measurement technique and
it is postulated that the high pressures exerted by the mercury on
the sample causes pores to break open, altering the pore sizes in the
24 h
0.6
48 h
0.5
0.4 14 days
0.3
0.2
0 5 10 15 20
Centre Length along compact / mm Edge
Fig. 6. Binder distribution from the centre to edge of compact during wick debinding.
sample. Another possibility is that the high pressure increases the
temperature such that remaining binder softens and is displaced by
the mercury.
The binder was observed to leave the compact and move through
the wicking powder with a distinct front, as shown in Fig. 7. This
movement is consistent with both the German and Vetter et al. mod-
els (Figs. 1 and 3, respectively). Little, if any, fingering of binder
within the pores of the wicking powder was observed. The satura-
tion level of the binder in the partially saturated region of the wick-
ing powder was uniform, at 0.4, confirming that motion through the
wicking powder was also controlled by capillary suction via small
pores, mercury porosimetry was conducted before and after wick
debinding on the (saturated where relevant) wicking powder. From
this it was clear that there were significantly fewer smaller pores in
the saturated wicking powder than in the non-saturated powder. It
was also apparent that the level of larger pores remained relatively
unchanged. Unlike in the compact, the pore structure of the wick-
ing powder was unaltered by mercury porosimetry due to its sin-
tered nature. From this it can be concluded that the binder travels
through the smaller pores of the wicking powder, allowing for air to
permeate in the opposite direction into the compact, replacing the
evacuated binder.
3806 I.M. Somasundram et al. / Chemical Engineering Science 63 (2008) 3802 -- 3809

Wicking powder L = 20 mm Wicking powder and therefore combining Eqs. (11) and (14) yields Poisson's
equation, viz.,
Compact
j2 P
=0 L<z<W (15)
jz 2

t=1h and the velocity of boundary W is given by Eq. (11).

Time
(ii) Compact: Binder movement in the compact is affected by the
change in saturation. The appropriate form of Darcy's law is (Mirzaei
2h
and Das, 2007)

Kc Kr jP
3h uz = − (16)
 jz

where Kc is the compact permeability and Kr is the relative perme-

4h ability of the compact. In this domain binder is being lost throughout
the compact and a volume balance gives
Fig. 7. Schematic and photographs showing the extent of binder ingress (marked jS juz
by yellow region) into the wicking powder during wick debinding. c =− (17)
jt jz

where S is the liquid binder saturation. Combining Eqs. (16) and (17)
The existing models attribute the decrease in debinding rate to yields
the increase in distance between the front in the wicking powder  
and the binder in the compact. As this length increases the pressure jS Kc j jP
c = Kr (18)
gradient decreases and according to Darcy's law (Eq. (11)) this will jt  jz jz
reduce the flux of binder. On the basis of our saturation measure-
where Kr and P are both functions of saturation S, which when sub-
ments, we postulate that the change in dimension is accompanied
stituted into Eq. (18) yields a partial differential equation for P. By
by a change in the difference in capillary pressures between the
comparing Eqs. (15) and (18) it follows that both domains can be
binder front in the wicking powder and the binder in the compact.
modelled using the same equation with different parameters, viz.,
The decrease in saturation in the compact means that the remain-
ing binder is held in progressively smaller and smaller pores, with  
jP j B jP
the result that the capillary suction of the compact increases and A = (19)
jt jz  jz
the driving force for debinding decreases. Another retardation effect
arises in that the permeability of the unsaturated compact decreases where the coefficients A and B are given for each domain in Table 2.
as debinding occurs so that progressively smaller and smaller pores The relationship between the capillary pressure difference, satu-
are involved in transporting the binder along a more tortuous route, ration and relative permeability of the compact (P.S.Kr ) can be ob-
effectively increasing the resistance to binder motion. These effects tained experimentally. However, for ceramic pastes with wax-based
are non-linear with saturation and have not been considered in the binders this is not straightforward as measurements need to be ob-
existing models. tained at temperatures where the wax is molten. For the purpose of
demonstrating the validity of this new model the following approach
4.1. Model formulation is used for the P.S.Kr relationships.
Brookes and Corey (1964) proposed relationships based on a the-
Two domains are considered in wick debinding, namely the wick- oretical evaluation of the mutual dependence of all three variables.
ing powder occupied by binder, and the compact (see Fig. 8a): They applied the following experimentally observed correlation:
(i) Wicking powder: Experiments indicate that the saturation level  
S − Sr Pd 
is constant in this domain and has a well defined linear front, in = (20)
1 − Sr Pc
agreement with Bao and Evans (1991). Movement of binder through
this domain can be approximated using Darcy's law which is used where Pc is the capillary pressure of the compact powder, Pd is the
in wick debinding by German (1987), Vetter et al. (1994b), Lin and pore pressure of the largest pores, Sr is the irreducible liquid satu-
Hourng (2005) and others. In one dimension, this takes the form ration and  is a particle size distribution index. The pressure driv-
Kw jP ing force in wick debinding is the difference between the capillary
uz = − (11) pressure of the wicking powder and the capillary pressure of the
 jz
compact, i.e.
The length of the wicking powder domain (W in Fig. 8a) increases
with time as binder empties from the compact into the wicking Pc = Pw − P (21)
powder. The boundary conditions are
Brookes and Corey (1964) used the experimental observation of Eq.
P = Pw z=W (12) (20) in combination with theory to give the relative permeability
Kr as
jP
=0 z = 0 (symmetry) (13)  
jz S − Sr 2+3/
Kr = (22)
where P at z = 0 changes as binder is removed from the compact. 1 − Sr
As there is no accumulation of binder within the wicking powder
The values of Pd , Sr and  were obtained by analysis of mercury
(except at the binder front), continuity gives
porosimetry results on the compact powder according the method
juz used by Brookes and Corey (1964). The capillary suction pressure of
=0 L<z<W (14)
jz the wicking powder was obtained from capillary rise experiments.
 I.M. Somasundram et al. / Chemical Engineering Science 63 (2008) 3802 -- 3809 3807

Binder
occupied

Compact

z
W
L

Wicking
powder

Binder
velocity

Compact sub-domain Wicking powder

sub-domain
Boundary condition P = Pw
Insulation/symmetry
Internal boundary

Fig. 8. (a) Schematic for debinding model and (b) finite element mesh and domains used in COMSOL multiphysics.

In brief, a vertical glass tube with the base sealed with Whatman Table 2
Parameters used for each domain in Eq. (19)
Grade 1 filter paper was filled with the powder. Having preheated
both the tube of powder and a beaker of wax, the tube was dipped Coefficient Wicking powder domain Compact domain
into the wax and allowed to draw the wax up the column of powder. A 0 c dP
dS

When the wax ceased to climb the column of powder the height B Kw Kc Kr (P)
of the wax column was equivalent to the capillary suction pressure
according to

Pw = b gh (23) Table 3
Constants used in simulations
The model requires a single free parameter, namely Kw , which cannot
Constant/parameter Value
be elucidated experimentally due to the complex proprietary struc-
ture, where an ill defined network of comparatively smaller pores is w 0.28
used in binder transport. We do, however, show that the value of Kw Kc 1 × 10−10 m2
Pw /kPa 234 kPa
that reproduces the experimental data is consistent with estimates Pd /kPa 45.2 kPa
provided via predictions similar to Eq. (3).  0.621

4.2. Numerical solution

The PDE of Eq. (19) for the 1D experimental case were solved us-
ing the commercial finite element method software package COM-
SOL Multiphysics (Version 3.3a, COMSOL, Inc.) using a moving mesh
to simulate the movement of the binder front in the wicking powder
domain. The mesh setup is shown in Fig. 8b. A block mesh is used
as debinding is one-dimensional. The boundary where P = Pw is a
moving boundary with velocity uz given by Eq. (11). The simulation
took less that 10 s to solve on a standard desktop PC with 3 GB RAM
and an AMD Athlon 64-bit 2.41 GHz processor.
The simulations performed with the constant measured values
reported in Table 3 are compared with experimental data in Fig. 9.
The simulation results are shown for various values of Kw ranging
from 4 × 10−18 to 1 × 10−17 m2 , covering the range of experimental
results with the simulation result for 6 × 10−18 m2 providing the
closest match based on sum of squared differences. For comparison,
the Carman--Kozeny approximation for permeability (based on the
particle size and porosity of the powder) estimates a value Kw(CK)
of 2.1 × 10−15 m2 as shown in Eq. (24) where the porosity ignores
3808 I.M. Somasundram et al. / Chemical Engineering Science 63 (2008) 3802 -- 3809

1.0 between pure alumina and the wicking powder. Using this analogy
Experiment as an approximation estimates the permeability of the binder oc-
0.9 4 cupied pores to be 2.3 × 10−17 m2 ; which is in rough agreement
5 with the best match with experimental data. Rough et al. (2002)
0.8 6 Kw / 10-18 m2 noted that the Carman--Kozeny approximation tends to overesti-
7 mate the permeability, as is observed from our simulations. Our new
Mean Saturation

0.7 10 model is currently the only way of measuring Kw for the wicking
powder.
0.6 Fig. 10 shows axial profiles for binder saturation generated by
the simulation with experimental results. The model predicts ex-
0.5 actly uniform saturation profiles for the binder distribution in the
compact. This is expected given that the permeability of the wicking
0.4 powder is much lower than that of the compact, so that the wick-
ing powder controls the rate of debinding. This result agrees with
0.3 the experimental observations reinforcing the validity of our model,
demonstrating that this approach provides a better description of
0.2 what happens to the binder during wick debinding.
0 4 8 12 16
Time / days
5. Conclusions
Fig. 9. Mean saturation from experiments compared with simulations based on
Brookes and Corey relationships. Previous models of wick debinding have oversimplified the prob-
lem by making assumptions as to how the binder leaves the compact
as it is drawn out by the wicking powder. The experiments conducted
1.0 in this study show that these mechanisms are incorrect and so a
new wick debinding model is proposed. The model has been tested
0.9 using numerical simulations that produces results closely agreeing
with the experimental behaviour. The model requires a single free
parameter, the permeability of the wicking powder, which may be
0.8
found by comparing simulation results to those of experiments. This
approach leads to a sensible estimation of the wicking powder per-
0.7 meability, validating the wick debinding model.
Saturation

Further work is required to apply the simulation to more complex

0.6 geometries in both two and three dimensions with the ultimate aim
of predicting regions of increased local strain and thus potential
0.5 defect formation.

0.4 Notation

0.3 A coefficient in Eq. (19)

B coefficient in Eq. (19)
0.2 d mean particle diameter, m
0 5 10 15 20 dp pore neck diameter, m
Centre Length along compact / mm Edge g acceleration due to gravity, m s−2
h height of binder in powder column, m
Experiments Simulation H compact height, m
1h K permeability, m2
6h Kr relative permeability
24 h Kw(CK) Carman--Kozeny permeability of wicking pow-
48 h der, m2
14 days l length (or height) of binder column in compact,
−18 m
Fig. 10. Saturation profiles generated by simulations for Kw = 6 × 10 m2 showing
uniform binder distribution. L length of compact, m
mb mass of binder in original sample, kg
mb,i mass of liquid binder after initial phase change
the large pores that are not used for binder transport. expansion, kg
mb,r mass of binder removed from sample, kg
2 3
dw (1 × 10−6 )2 × 0.473 mp mass of powder in original sample, kg
Kw(CK) = w = =2.1 × 10−15 m2 (24)
180(1 − w )2 180 × 0.532 mT total mass of paste (sample), kg
P capillary pressure, Pa
However, this approach assumes there are no large pores in the Pd initial capillary pressure of compact, Pa
wicking powder. As binder only travels through the small pores S saturation
the permeability of the wicking powder will be lower than calcu- Sr irreducible saturation
lated in Eq. (24). A reduction factor due to the connectivity of the t time, s
small pores may be estimated from the ratio of thermal conductivity tD debinding time, s
 I.M. Somasundram et al. / Chemical Engineering Science 63 (2008) 3802 -- 3809 3809

u binder velocity, m s−1 Brookes, R.H., Corey, A.T., 1964. Hydraulic properties of porous media. Hydrology
uz binder velocity in axial z direction, m s−1 Papers. Colorado State University.
Chen, C.C., Hourng, L.W., 1999. Numerical simulation of two-dimensional wick
V volume, m3 debinding in MIM by body fitted finite element method. Powder Metallurgy 42,
Vp bulk volume of compact powder, m3 313--319.
Dullien, F.A.L., 1972. Porous Media---Fluid Transport and Pore Structure. Academic
W position of binder front in wicking powder, m Press, Inc., San Diego.
x fractional debinding German, R.M., 1987. Theory of thermal debinding. International Journal of Powder
z axial dimension (direction of debinding), m Metallurgy, 237--245.
Lewis, J.A., 1997. Binder removal from ceramics. Annual Review of Materials Science
Greek letters 27, 147--174.
Lin, T.L., Hourng, L.W., 2005. Investigation of wick debinding in metal injection
 binder surface tension, N m−1 molding: numerical simulations by the random walk approach and experiments.
Advanced Powder Technology 16, 495--516.
 porosity Liu, C.Y., Murakami, K., Okamoto, T., 1986. Effect of capillary pressure on
 contact angle, rad interdendritic liquid flow. Acta Metallurgica 34, 159--166.
 particle size distributionindex Lograsso, B.K., German, R.M., 1990. Thermal debinding of injection molded powder
compacts. Powder Metallurgy International 22, 17--22.
 binder viscosity, Pa s Mirzaei, M., Das, D.B., 2007. Dynamic effects in capillary pressure-saturations
 density, kg m−3 relationships for two-phase flow in 3D porous media: implications of
microheterogeneities. Chemical Engineering Science 62, 1927--1947.
Subscripts Mutsuddy, B.C., Ford, G., 1995. Ceramic Injection Moulding. Chapman & Hall, London.
Patterson, B.R., Aria, C.S., 1989. Debinding injection molded materials by melt
bl binder liquid wicking. JOM---Journal of the Minerals Metals & Materials Society 41, 22--24.
Rough, S.L., Bridgwater, J., Wilson, D.I., 2002. In situ measurements of porosities and
c compact
permeabilities of alumina pastes. Powder Technology 123, 262--274.
p powder Shaw, T.M., 1986. Liquid redistribution during liquid-phase sintering.
w wicking powder Journal---American Ceramic Society 69, 27--34.
Vetter, R., Sanders, M.J., Majewskaglabus, I., Zhuang, L.Z., Duszczyk, J., 1994a.
Wick-debinding in powder injection-molding. International Journal of Powder
Acknowledgements Metallurgy 30, 115--124.
Vetter, R., Horninge, W.R.V., Vervoort, P.J., Majewskaglabus, I., Zhuang, L.Z., Duszczyk,
Project funding from EPSRC and Ross Ceramics is gratefully ac- J., 1994b. Square-root wick debinding model for powder injection-molding.
Powder Metallurgy 37, 265--271.
knowledged, as is assistance from Zlatko Saracevic and the Paste and Waikar, R.J., Patterson, B.R., 1986. Horizons of Powder Metallurgy, Part II. Verlag
Powder Processing Group at the Department of Chemical Engineer- Schmid, Freiburg. p. 661.
ing, University of Cambridge.

References

Bao, Y., Evans, J.R.G., 1991. Kinetics of capillary extraction of organic vehicle from
ceramic bodies. Part II: partitioning between porous media. Journal of the
European Ceramic Society 8, 95--105.