Ceramics International 49 (2023) 4578–4585

Contents lists available at ScienceDirect

Ceramics International
journal homepage: www.elsevier.com/locate/ceramint

3D printing of cordierite materials from raw reactive mixtures

B. Dorado a, L. Moreno-Sanabria b, E. García c, M. Belmonte b, P. Miranzo b, M.I. Osendi b, *
a
School of Mining and Energy, University Polytechnic of Madrid (UPM), 28003, Madrid, Spain
b
Institute of Ceramics and Glass (ICV), Spanish National Research Council (CSIC), 28049, Madrid, Spain
c
Center for Thermal Spray Research, Stony Brook University, Stony Brook, NY, 11794, USA

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

Keywords: Porous cordierite materials are 3D printed by robocasting from two kaolin containing raw materials mixtures.
Cordierite Water suspensions of both mixtures at variable solid concentrations (40–67 wt%) are characterized by rheo­
3D printing logical measurements, showing good printability for concentrations >60 wt% without the need of any printing
Ink rheology
additive. The mixtures react during sintering (at 1250 ◦ C) giving indialite as the main phase in the structures,
Compression strength
Thermal conductivity
which differ in minor phases. Three types of lattices are printed for both compositions with a logpile inner
Thermal expansion coefficient structure. Properties of interest like the coefficient of thermal expansion (CTE), the thermal conductivity (KT) and
the compression strength (σ) of the printed cordierites are determined and compared with published data. Re­
sults evidence that printing of clay containing reactive mixtures is a straightforward and cost-effective way to
achieve porous complex shaped cordierite with CTE~ 2–3 x10− 6 K-1, KT ~ 0.4–0.6 W m− 1 K− 1 and maximum σ of
24 MPa.

1. Introduction magnesite (MgCO3) are usual sources for MgO; these mixtures must be
adjusted sometimes with extra additions to achieve the stoichiometric
Cordierite is a Mg–Al silicate ceramics (Mg2Al4Si5O18) particularly cordierite composition [12]. During the heat treatment, the selected
appreciated for its thermal shock resistance owing to its low thermal mixture of raw materials experiences a series of reactions until the
expansion coefficient, which also has a remarkable low thermal con­ cordierite is formed at temperatures in the range of 1200–1300 ◦ C [11,
ductivity, low dielectric loss and high refractoriness and corrosion 12].
resistance. All these characteristics make cordierite materials appealing Honeycomb-type structures are of prime interest for several of the
in diverse industrial uses, such as supports in catalytic converters, par­ mentioned applications, being the extrusion of reactive mixtures the
ticle filters, heating tubes in furnaces, kiln furniture, infrared gas usual processing route [13]; nonetheless, a number of articles very
burners or as electrical insulator [1–6]. Besides, in the case of recently referred the use of additive manufacturing (AM) routes for
glass-ceramic compositions, applications include radomes for antenna getting structured cordierite ceramics. It is well known that AM tech­
protection in missiles and low-temperature co-fired ceramic substrates niques have the advantage of making possible complex shapes and
for electronic packaging [7,8]. structures. In fact, methods like digital light processing (DLP) [14] and
Cordierite materials are cost-effectively produced from adequate direct ink writing (DIW) [9,15] using previously reacted cordierite
mixtures of diverse raw materials, although other routes such as the sol- powders have been described. Reportedly, there is just one paper that
gel method [9] or the stoichiometric mixture of pure oxides [10] are also details the DIW of a reactive mixture consisting of talc, clay and
employed for controlling the impurity presence. Cordierite melts α-alumina [16], making use of methyl cellulose and polyethylene glycol
incongruently and its stability zone in the phase equilibrium diagram is as printing additives. Compared to other AM methods like DLP, SLA
delimited by six eutectic points; accordingly, single-phase dense cordi­ (stereolithography) and FDM (fused deposition modeling), DIW offers a
erite bodies are difficult to get because of the narrow sintering window great versatility as any material can be printed without using large
and the common presence of secondary compatible phases [11,12]. amounts of photoresins or thermoplastics that often require of slow
Among the possible raw materials mixtures, clay is a common source debinding processes [17,18]. Fidelity to models, multi-materials print­
for alumina (Al2O3) and silica (SiO2), whereas talc (Mg3Si4O10(OH)2) or ing and 4D methods are additional advantages of DIW [18,19]; besides,

* Corresponding author.
E-mail address: miosendi@icv.csic.es (M.I. Osendi).

https://doi.org/10.1016/j.ceramint.2022.09.343
Received 29 July 2022; Received in revised form 20 September 2022; Accepted 26 September 2022
Available online 4 October 2022
0272-8842/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
B. Dorado et al. Ceramics International 49 (2023) 4578–4585

Fig. 1. XRD pattern of the original batches with corresponding phase ascrip­
tion, where T stands for talc; K is kaolinite; G, gibbsite; Q, quartz; E, enstatite
and Mg, magnesite.

the possibility of printing fine features and overhanging parts has been
proved by using fine nozzles and no wetting solvent baths [20]. For Fig. 2. The CAD models (a) and view of the sintered structures for SP100
composition (b).
certain applications that require exact dimensions and smooth surface
finish, i.e. bulk parts components for engines and prosthesis, the method
may not be appropriated or entail post-processing steps [21].
Table 1
While the development of commercial ceramic parts and products Geometric specifications (CAD models and after sintering) and labelling of the
using AM methods is still a long way off compared to metals and poly­ 3D printed samples.
mers, continued progress has been made in reaching the market [21].
batch outer shape side (L) or diameter height (H) labelling
However, it has been recognized in recent reports that further advances (Φ) (mm) (mm)
in this sense are linked to the development of reliable inks/feedstock,
CAD sinter CAD sinter
increasing the knowledge and reliability of the whole process, producing
a wide database of properties of the 3D printed parts and the develop­ SP100 prism 12 11 6 5 3dSP100-sp
prism 8 7.5 12.5 10.2 3dSP100-lp
ment of appropriate standard methods [20–22].
cylinder 8 7.4 12.5 10.2 3dSP100-c
Presently, we achieve logpile cordierite structures by robocasting MG300 prism 12 10.5 6 4.5 3dMG300-sp
using two different reactive mixtures, both containing kaolin but prism 8 7.1 12.5 9 3dMG300-lp
different MgO sources. We show that both batches can be printed cylinder 8 6.9 12.5 9 3dMG300-c
without any additives, and that the raw materials react during heating to
form mainly cordierite bodies retaining their printed structures.
(pH = 7). Suspensions were first conditioned with a paddle mixer in a
Important parameters for cordierite materials such as the coefficient of
glass beaker within an ultrasonic bath, and then, in a planetary cen­
thermal expansion, the thermal conductivity and the mechanical
trifugal mixer (ARE-250, Thinky Co.) at 1500 rpm for 40 s. To determine
strength are directly determined in the 3D printed materials and dis­
their rheological characteristics in terms of flow curves for increasing
cussed in relation to their structure and porosity, and also compare to
shear rates, and the storage (G1) and loss moduli (G2) under oscillating
published data for porous and dense cordierites.
amplitude sweep conditions at constant frequency of 1 Hz, corre­
sponding measurements were done in a rheometer (CVO 100 D, Bohlin
2. Experimental
Instruments) using a cone-and-plate geometry (diameter: 40 mm; cone
angle: 5◦ ) covered with a fitting tool to reduce evaporation. A pre-
Two commercial batches were used as starting materials, labelled as
shearing at 8 s− 1 for 10 s followed by 10 s stabilization was the stan­
MG300 and SP100 (supplied by Vicar SA, Spain). These batches consist
dard procedure before each measurement.
of a mixture of raw materials according to their powder x-ray diffraction
Once the optimal suspensions were selected according to the rheo­
(XRD) patterns (Fig. 1), which were recorded with the D8 Advance
logical results and printability proofs, structures with different shapes
diffractometer (Bruker Inc.). Peak adscription was completed by
and geometries (Fig. 2) were filament printed with a three-axis robo­
matching the spectra with database included in the equipment software.
casting system (A3200, 3-D Inks LLC) and using CAD software (Robo­
In particular, gibbsite (Al(OH)3), talc (Mg3Si4O10(OH)2) and kaolinite
CAD 4.2, 3-D Inks LLC). The printing syringe was filled with the
(Al2Si2O5(OH)4) are present in the case of the SP100 batch, whereas the
corresponding ink that was extruded through the nozzle tip (Precision
MG300 contains magnesite (MgCO3), kaolinite and enstatite (MgSiO3);
Tips; EFD Inc., USA), using inner tip diameter of 610 μm (D) and a
certain amount of quartz was detected as well in both lots. The average
printing speed of 10 mm s− 1. The specific geometric parameters of the
particle size of both powder compositions were determined by means of
printed cylinders and long and short square prisms are displayed in
the laser diffraction technique (Mastersizer, Malvern Panalytical)
Table 1. The internal pattern of the CAD model structures was logpile
showing a median value around 9.5 μm but extending over a wide scale
type with a distance between filament centers a = 1.8 mm for the short
of sizes (~ 0.5–100 μm) (see Supplementary information (SI), Fig. S1).
prism and 1.2 mm for the other geometries, and the distance between
Water suspensions of the powders mixtures were prepared at
centers of alike rods in the z direction was h = 0.958 mm in all cases,
different concentrations in the 40–65 wt % range using ultra-pure water
which gives an overlapping (0.21D) between connecting filaments. After

4579
B. Dorado et al. Ceramics International 49 (2023) 4578–4585

Fig. 3. Rheological data of the inks for different solid contents (40–65 wt%). Apparent viscosity as a function of increasing shear rate of MG300 (a) and SP100 (b);
storage (G1) and viscous (G2) moduli vs shear rate for MG300 (c) and SP100 (d).

printing, structures were left to dry at ambient temperature during 2–4 with model dimension of 30 × 30 x 7 (in mm) having the usual logpile
days depending on the size and, later, they were heated at 100 ◦ C for 48 structure with distance (a) between filaments of 1.67 mm and two
h to eliminate the residual water. The dried SP100 and MG300 3D covers printed of the same material to assure an optimum contact with
structures were sintered in an electric furnace at a maximum tempera­ the TPS sensor; these samples (labelled with postfix K) were printed with
ture of 1250 ◦ C for 2 h, using a heating ramp of 1 ◦ C min− 1 up to 600 ◦ C a nozzle tip of 840 μm diameter. This procedure has proved effective for
and 5 ◦ C min− 1 to the upper set point. XRD patterns of the powered measuring the effective thermal conductivity in this type of reticulated
sintered samples were recorded to stablish the final phases. Geometric ceramics with through pores [23]. For comparative purposes, dense and
density (ρg) was determined by the mass and volume of the samples, and closed pore pellets, prepared by adding organic charge porogens, of
the apparent (ρapp) and true (ρtrue) densities were measured by the water MG300 and SP100 compositions were uniaxial pressed and sintered at
immersion method using appropriate expressions (see SI, S2). 1250 ◦ C for 2 h; the corresponding KT values were determined by the
The coefficient of thermal expansion (CTE) was measured between laser flash method (Thermaflash 2200, Holometrix-Netzsch) in disc
room temperature and 900 ◦ C at a heating rate of 5 ◦ C min− 1 in a hor­ shaped samples of 12.7 mm and 1–2 mm thick. Error in these data
izontal dilatometer (Netzsch, model 402 EP) with quartz sample holder correspond to the estimated accuracy of the laser flash technique (~7%).
and pushrod, measurements were carried on the cylindrical 3D struc­ All reported characteristics (dimensions, density, KT, σ) and prop­
tures along the c-axis (Fig. 2). To verify possible orientation effects on erties of the 3D cordierite materials are averaged for 2–3 specimens of
the CTE of the 3D structures, thermal expansion measurements were each composition.
also done along the printing plane and in the vertical z-direction for the Microstructure of the printed cordierite samples was observed in the
3dSP100-lp square prism. tabletop scanning electron microscope (TM1000, Hitachi Inc.) as well as
The two types of cordierites with the different CAD designs shown in with the field emission scanning electron microscope (FESEM, S-4700,
Fig. 2 and Table 1 were compression tested in an universal material Hitachi Inc.), in this case, the samples were sputtered with a thin gold
testing machine (ZwickiLine Z5.0 TS, Zwick-Roell) after gentle grinding coating.
the sides facing the compression platens to eliminate asperities and
produce even contact surfaces. Compression strength (σ) was calculated 3. Results and discussion
dividing the crushing stress by the contact area. Apparent elastic
modulus (E) of the samples was estimated as the slope gradient of the Rheological data for SP100 and MG300 inks are displayed in Fig. 3.
linear part of the stress/strain curves (SI, Fig. S3). Curves of viscosity (η) versus share rate (γ̇) display shear thinning
Thermal conductivity (KT) was assessed by the transient plane source behavior for both SP100 and MG300 inks, with an obvious viscosity rise
method (TPS Trident, C-Therm Technologies Ltd.) with a flexible with the increase of solid concentration in the suspensions. The SP100
double-sided hot disc sensor (6 mm of diameter) at 20 ◦ C, inside a inks always display lower viscosities than the MG300 for the same
thermal chamber. The sensor was placed between two identical scaffolds concentration (Fig. 3a and b). For concentrations (≥60%) the viscosity

4580
B. Dorado et al. Ceramics International 49 (2023) 4578–4585

Fig. 4. Representation of the storage (G1), loss moduli (G2) and dynamic yield stress (τY) vs the solid content for the MG300 and SP100 inks (a); evolution of G1 as a
function of τY for both cordierite inks at the concentrations shown in a). The hatched zone designates the region of good printability.

observe that SP100 points seems slightly downshifted in relation to

MG300 for a given solid content (Fig. 4b). In fact, as τY remains rela­
tively lower for the SP100 at 65% of solid loading, a new ink with 67 wt
% of solids was prepared for SP100 that increased this parameter only
slightly, as can be seen in Fig. 4b, giving indication that the saturation
limit for the ink was reached. In the hatched region of Fig. 4b, G1 and τY
values are above 105 and 3 x 102 Pa, respectively, and the corresponding
concentrations produced printable inks for the present conditions of
absence of any additives. Accordingly, the 67% (SP100) and 65%
(MG300) inks were selected for subsequent studies. It seems reasonable
to assume that the rheological conditions producing the required gel
transition in the inks for robocasting are mostly determined by the
kaolin clay in the raw materials, which has demonstrated suitability for
DIW without the need for additives [25,26].
Final compositions contain cordierite as main phase, in particular
indialite, which is the hexagonal form of cordierite, with some presence
of protoenstatite (MgSiO3, high temperature polymorph) in the case of
MG300; as for the SP100, indialite is also the major phase, with minor
contents of mullite, spinel (MgAl2O4) and α-alumina (Fig. 5). Consid­
ering the composition in oxides of the original powders (see SI, Table S4)
Fig. 5. XRD patterns of 3D printed samples after sintering at 1250 ◦ C with the and the SiO2–Al2O3–MgO phase equilibrium diagram [27], a quantifi­
corresponding phase ascription (I, indialite; M, mullite; C, corundum; S, spinel; cation of the equilibrium phases at the sintering temperature was done,
E, protoenstatite). giving the following weight percentages: 82 cordierite, 16 proto­
enstatite, 2 silica for MG300, and 72 cordierite, 20 mullite, 8 silica for
curves approach; in fact, plots are very close for the 60 and 65% MG300 SP100, which seem a reasonable estimation in both materials. Silica is
inks (Fig. 3a), and they are also similar to the 65% SP100 suspension probably as glassy phase. A series of reactions occur during heating until
(Fig. 3b). During extrusion, the maximum shear rate occurs at the nozzle achieving the final cordierite products which entail the decomposition
walls, which can be estimated by the expression γ̇ = 4π Qr3 [24], being Q of kaolin, magnesite, gibbsite and talc, and subsequent reactions to form
the volumetric flow rate and r the nozzle radius. Considering Q = v A, cordierite and secondary phases (see XRD patterns at different temper­
where v is the printing speed and A is the printing nozzle area, the shear atures in the range RT-1250C in the SI, Fig. S5).
rate at the nozzle surface can be estimated as 131 s− 1 (compatible with During the whole process the structures maintain their original
an extrusion process). Hence, viscosities of the MG 300 and SP100 inks shapes, as can be seen in Fig. 2, although experiencing different linear
at 65% solid contents would have rim viscosities around 15–30 Pa s (of shrinkages (Table 1) along the horizontal x-y plane, ~7% (SP100) and
the same order as honey) during extrusion. ~12% (MG300), and the z-plane, ~20% (SP100) and ~25% (MG300),
As for G1 and G2 moduli, a plateau region is appreciated, indicative which could be attributed to certain weight slump effect but also to a
of a viscoelastic behavior, which progressively extends to higher shear possible accommodation of the kaolinite and talc clay particles during
stresses with the solid content of the inks (Fig. 3c and d). The plateau extrusion, which would be more significant in the SP100, explaining its
extension is minimum for the SP100 40% ink (Fig. 3d) and shows the lower τY; indeed, decreases in τY have been reported for Al2O3 inks
maximum extension for the MG300 65% ink (Fig. 3c). containing platelet-like particles, either Al2O3 or graphene, when the
Fig. 4a displays more clearly the variation of the G1 and G2 plateau platelet content increased, and also showing τY values below 6 × 102 Pa
values as a function of the solid content of MG300 and SP100 inks, and [23,28]. In fact, certain particle orientation along the extrusion axis is
with representation of the yield stress (τY) as well, estimated as the shear perceived in the green filaments particularly for SP100 composition (see
stress at the crossing point of both moduli. G1, G2 and τY show gradual SI, Fig. S6).
increases with the solid content in mutual inks although MG300 always The modeled rod distance, a, was reduced to 0.97 mm (SP100) and
displays higher values of these parameters up to around 65% solid 0.92 mm (MG300) for the long cylinders and prisms, and to 1.67 mm
contents, where certain convergence between corresponding data of (SP100) and 1.42 mm (MG300) for the short prisms (Fig. 2). In parallel,
both inks is observed. If we represent G1 and τY, regarded as two useful the filament diameter, φ, was reduced to 0.47 mm (SP100) and 0.43 mm
parameters for estimating the ink printability condition [25], we (MG300). The overall volumetric sintering shrinkage of 3D MG300
samples (39%) was higher than that of SP100 (28%) structures, also in

4581
B. Dorado et al. Ceramics International 49 (2023) 4578–4585

Table 2
Density (ρg, ρapp, ρtrue) and porosity values (Pmacro, Pstruts, Ptotal) of the 3D samples.
sample ρg (g cm− 3) ρapp (g cm− 3) ρtrue (g cm− 3) Pmacro (%) Pstruts (%) P total (%)

3dSP100-sp 0.65 ± 0.02 2.46 ± 0.02 1.45 ± 0.02 54 ±2 45 ± 3 76 ± 3

3dSP100-lp 0.81 ± 0.02 2.39 ± 0.02 1.44 ± 0.03 38 ±3 49 ± 2 69 ± 3
3dSP100-c 0.85 ± 0.04 2.37 ± 0.02 1.48 ± 0.08 18 ±3 51 ± 3 67 ± 3
3dMG300-sp 0.88 ± 0.01 2.46 ± 0.02 1.60 ± 0.01 53 ±2 36 ± 3 66 ± 3
3dMG300-lp 1.02 ± 0.02 2.36 ± 0.03 1.65 ± 0.02 38 ±3 37 ± 3 61 ± 3
3dMG300-c 1.07 ± 0.02 2.35 ± 0.03 1.62 ± 0.03 21 ±4 49 ± 3 59 ± 3

Fig. 6. Representative microstructures of the SP100 (a–d) and MG300 (e–h) cordierite samples on the filament surface (a,b.e,f) and the cross sections (c,d,g,h).

Fig. 7. Compression strength (σ) and elastic modulus (E) data for the different
sintered 3D structures of SP100 and MG300 materials. Image shows the 3D
sample types used for testing.
Fig. 8. Comparison of strength results with published data for porous cordierite
materials including 3D printed (DIW and DLP), replica and foaming methods as
consonance with the higher mass loss of the first (40% against 30% of a function of the geometric density (ρg). Data for dense cordierite are also
SP100) associated to the plus of the magnesite decomposition (See included. Straight lines represent prediction for open cells foams [32] and the
corresponding thermograms of Fig. S7 in the SI). best fit to present data (red squares). (For interpretation of the references to
In Table 2, ρg, ρapp, and ρtrue for the different samples, together with colour in this figure legend, the reader is referred to the Web version of
the open macroporosity of the pattern (Pmacro), the porosity of the struts this article.)
(Pstrut, referred just to the solid volume) and the total porosity (Ptotal) of
the 3D samples, are summarized. The strut porosity is ~12% higher on the MG300 when comparing, of course, the same type of model struc­
average for the SP100 (~ 48%) than for the MG300 (~36%) samples, ture, being the short prism structures those with the highest macro-
thus, ρtrue is higher for the MG300 material, about 1.7 vs 1.5 g cm− 3 of porosity attributable to the higher spacing (a) between rods. As for the
the SP100 material, in agreement with the higher contraction during pores in the solid struts, SP100 sample clearly displays larger pores than
sintering and the slightly higher green density of MG300 structures the MG300 as we can perceive by comparing the microstructures of the
(1.01 vs. 0.93 g cm− 3). The SP100 structures have more porosity than rod surface (Fig. 6a–b, e-f)) and inner (Fig. 6c–d, g-h) of both materials.

4582
B. Dorado et al. Ceramics International 49 (2023) 4578–4585

Fig. 9. Calculation of the CTE for 3dSP100-c and 3dMG300-c samples along the z-axis (a), and along x- and z-directions of 3dSP100-lp sample (b).

Regarding the strength of the samples displayed in Fig. 7, few as­ increases to a maximum of 3.4 × 10− 6 K− 1. These values are relatively
sertions can be drawn. Firstly, σ values are always higher for MG300 low, and of the same order of other bulk cordierites from raw materials
regardless of the type of structure, and secondly, the top σ data for each sources, but higher than CTE of pure cordierite bodies (<2 × 10− 6 K-1)
cordierite batch, 24 MPa (MG300) and 14 MPa (SP100), correspond to [9] owing to the presence of mullite (SP100) and protoenstatite
the cylinder shape, which is also the less porous (Table 2). The elastic (MG300), both ceramics with higher CTE, ~5 × 10− 6 K-1 [38] and ~12
modulus of the structures shows a similar behavior in both cordierites, × 10− 6 K-1 [39], respectively. Further reduction of CTE may be achieved
being the shorter structures the more deformable, whereas the tall using a higher sintering temperatures but below the respective melting
prisms and cylinders exhibit larger E, and if comparing both materials, points, which are 1340 ◦ C (MG300) and 1460 ◦ C (SP100), see SI Fig. S.8,
the 3D MG300 always achieves higher E (Fig. 7). increasing sintering times and reducing the impurities content to gain
The σ data can be represented as a function of ρg of the present crystallinity [11,31].
robocast samples, and also compared with available data for various
cordierite materials including cellular (foams and closed pore cells), 3D
printed (DIW and DLP methods) and dense bodies (printed and pressed Table 3
pellets) (Fig. 8). Comparing the different porous cordierite materials it Data for thermal conductivity of the 3D samples (TPS) and for bulk pellets (laser
flash technique). Ptotal, Pstrut of the printed samples, and Ptotal of the bulk samples
seems evident that strength of foams [29,30] are one order of magnitude
are included.
lower than closed cell cordierite [10,31] and printed honeycombs [16]
of similar density, whereas present logpile DIW structures approach σ Samples Thermal conductivity, KT Ptotal Pstrut

data of closed cell cordierites. In Fig. 8 are also included two straight (Wm− 1K− 1) (%) (%)
solid lines, one of them representing the Gibson and Ashby theoretical 3dSP100-K 0.39 ± 0.03 67 54
prediction for brittle open cell foams σ/σs ~ (ρ/ρs)3/2 [32] (σs, ρs rep­ 3dMG300-K 0.69 ± 0.01 56 40
resenting values of the strut dense material). The second line with a SP100-dense 1.67 ± 0.08 20 –
higher exponent, σ/σs ~ (ρ/ρs)1.9, seems to fit better the present DIW SP100-pg 0.67 ± 0.05 48 –
MG300-dense 1.78 ± 0.09 28 –
and closed cell cordierite data. The σ of 25 MPa for one of the DIW MG300-pg 0.73 ± 0.05 46 –
honeycombs seems quite above the other porous cordierites probably
due to hexagonal infill pattern [16], which has proven its strengthening
benefit [33]. Data for dense materials include two DLP cordierites [34]
and a sol-gel processed cordierite [35]. Although densities of “solid”
cordierites are relatively close, the strength of the pressed pellet is above
one order of magnitude higher than the solid cordierite by DLP. Fig. 8
evidences the versatility of printing methods that allows a wide range of
densities and robustness depending of the choice of CAD model and
printing method, thus being complementary and not excluding.
It can be argued that these 3D structures have been tested in the
direction perpendicular to the printing plane, and anisotropy may arise
if tested in the other two perpendicular directions. As the present
structures have different dimensions in the plane and length directions,
compression testing methods in the other planes could be misleading,
and particularly for the cylinder shapes. Nevertheless, if we consider as
source of the possible anisotropy only the logpile layout, leaving outside
the existence of frames and the possible orientation effects of the par­
ticles in the rods, it can be said that the compression resistance would be
quite similar for the 3 different orientations (in-plane and 2 cross-plane)
according to the results of Miranda et al. for cubic logpile specimens as
shown by finite element methods [36]. Differences in strength with the
printing direction have been reported for space filling ceramic bulk parts
mainly attributed to defects associated to entrapped air among the fil­ Fig. 10. Thermal conductivity at 25 ◦ C against total porosity of present
aments [37]. cordierite materials (3D printed and pressed pellets with and without porogens)
Measurements of the CTE gave very similar values for both cordierite compared to representative data from the literature including a dense
cylindrical printed structures (Fig. 9) showing a region (RT -200 ◦ C) of cordierite-mullite sample [41], conventional and sol-gel (Ptotal = 0.2) processed
lower CTE (~2 × 10− 6 K− 1) whereas in the range 200–800 ◦ C the CTE materials [35], extruded cordierite samples [40], a closed pore cellular sample
[10] and a foam [31].

4583
B. Dorado et al. Ceramics International 49 (2023) 4578–4585

To verify possible orientation effects on the CTE of the 3D structures, through the FPI contract Ref. PRE2019-091429 (2019 call).
thermal expansion measurements were done along the printing plane
and in the z-direction for the 3 dSP100-lp sample as an example. CTE Appendix A. Supplementary data
results plotted in Fig. 9b indicate that the differences of CTE for both
directions are <10%, 3.2 × 10− 6 K− 1for the z direction vs. 2.9 × 10− 6 K-1 Supplementary data to this article can be found online at https://doi.
for the printing plane, which do not seem of significance. org/10.1016/j.ceramint.2022.09.343.
Data for KT are shown in Table 3 measured for the same type of
logtype specimens in both cordierite samples, demonstrating larger References
values of these parameters for the MG300 specimen in agreement with
its lower total porosity (Table 3). These data are compared with KT of [1] S. Hajimirzaee, A.M. Doyle, 3D printed catalytic converters with enhanced activity
for low-temperature methane oxidation in dual-fuel engines, Fuel 274 (2020),
same cordierites (SP100, MG300) but measured in dense and porous 117848.
samples (Table 3). The bulk samples obviously have higher KT than the [2] Z. Lei, H. Gao, X. Chang, L. Zhang, X. Wen, Y. Wang, An application of green
corresponding 3D specimens as they display lower porosity, and alike surfactant synergistically metal supported cordierite catalyst in denitration of
Selective Catalytic Oxidation, J. Clean. Prod. 249 (2020), 119307.
for the porogen samples as their porosities are lower and closed-type. It [3] W.M. Carty, P.W. Lednor, Monolithic ceramics and heterogeneous catalysts:
is illustrative to represent the KT of all these samples as a function of honeycombs and foams, Curr. Opin. Solid State Mater. Sci. 1 (1996) 88–95.
their corresponding porosity, including other published data as well, in [4] W. Haase, L. Sedda, New ideas and proven solutions for kiln furniture made from
cordierite and oxide ceramics, CFI Ceram. Forum Int. 76 (1999) 9–14.
particular, extruded [40] and closed cell cordierites [10], a foam [31], [5] E. García, P. Miranzo, M.I. Osendi, Alterations in cordierite based burners
conventional and sol-gel bulk cordierites [35], and a dense cordier­ subjected to radiant mode ageing conditions, J. Eur. Ceram. Soc. 23 (2003)
ite/mullite composite [41], measured either by TPS or laser flash 3097–3103.
[6] M.A. Camerucci, G. Urretavizcaya, M.S. Castro, A.L. Cavalieri, Electrical properties
method (Fig. 10). All data show an exponential-type decay with the
and thermal expansion of cordierite and cordierite-mullite materials, J. Eur.
porosity, in particular, KT of the SP100 and MG300 bulk cordierite Ceram. Soc. 21 (2001) 2917–2923.
samples seem to fit reasonably the exponential expression KT= K0 e{− 1.5 [7] C. Lee, I.S. Kim, B. Lee, Slow crack growth in a cordierite-based glass–ceramic
x P/(1− P)}
for spheroidal inclusions [42], taken the thermal conductivity missile radome due to stress corrosion, Eng. Fail. Anal. 93 (2018) 76–86.
[8] Z. Li, J. Wu, L. Song, Y. Huang, Effect of composition on sinter-crystallization and
of the dense material as K0 = 2.6 Wm− 1K− 1. Materials with porosity properties of low temperature co-fired α-cordierite glass–ceramics, J. Eur. Ceram.
levels >50%, basically the cellular and printed materials, present KT Soc. 34 (2014) 3981–3991.
data higher than the predicted by the above expression. [9] Z. Chen, Z. Xu, F. Cui, J. Zhang, X. Sun, Y. Shang, R. Guo, N. Liu, S. Cai, C. Zheng,
Direct ink writing of cordierite ceramics with low thermal expansion coefficient,
J. Eur. Ceram. Soc. 42 (2022) 1685–1693.
4. Conclusions [10] X. Luan, J. Li, Z. Wang, W. Feng, K. Huang, S. Liu, Hierarchically cell-window
structured porous cordierite prepared by particle-stabilized emulsions using potato
starch as a modifier, J. Eur. Ceram. Soc. 41 (2021) 4369–4380.
3D cordierite structures are directly printed by robocasting from [11] J.R. González-Velasco, M.A. Gutiérrez-Ortiz, R. Ferret, A. Aranzabal, J.A. Botas,
water-based inks of different raw material reactive mixtures without Synthesis of cordierite monolithic honeycomb by solid state reaction of precursor
additives. Different logpile type models are printed with compression oxides, J. Mat. Sci. 34 (1999) 1999–2002.
[12] A. Chowdhury, S. Maitra, S. Das, A. Sen, G.K. Samanta, P. Datta Synthesis,
strength values in the 10–24 MPa range that decrease linearly with the Properties and applications of cordierite ceramics, Part 1, Interceram 56 (2007)
total porosity of the structures, analogously to reported results for 18–22.
diverse cordierite materials. The coefficient of thermal expansion of the [13] R.N. Dasa, C.D. Madhusoodana, K. Okada, Rheological studies on cordierite
honeycomb extrusion, J. Eur. Ceram. Soc. 22 (2002) 2893–2900.
structured samples also shows normal values (2–3 x 10− 6 K− 1) for bulk
[14] Z. Chen, J. Li, C. Liu, Y. Liu, J. Zhu, C. Lao, Preparation of high solid loading and
cordierite materials processed from raw sources with no evidence of low viscosity ceramic slurries for photopolymerization-based 3D printing, Ceram.
anisotropy regarding the printing plane. The thermal conductivity of the Int. 45 (2019) 11549–11557.
reticulated materials is particularly low for the 3D SP100 printed [15] Y. Pan, P. Zhu, R. Wang, Z. Si, B. Li, Y. Yao, Direct ink writing of porous cordierite
honeycomb ceramic, Ceram. Int. 45 (2019) 15230–15236.
cordierite sample (~0.4 Wm− 1K− 1) owing to its inner porosity and the [16] S. Mamatha, P. Biswas, D. Das, R. Johnson, 3D printing of cordierite honeycomb
3D design. Compression strength of present 3D printed cordierites show structures and evaluation of compressive strength under quasi-static condition, Int.
a potential dependence with density of the structures, σ/σs ~ (ρ/ρs)1.9 , J. Appl. Ceram. Technol. 17 (2020) 211–216.
[17] M.A.S.R. Saadi, A. Maguire, N.T. Pottackal, M.S.H. Thakur, M.M. Ikram, A.J. Hart,
whereas KT reasonably follows the predicted exponential decay with P.M. Ajayan, M.M. Rahman, Direct ink writing: a 3D printing technology for
porosity, KT= K0 e {− 1.5 P/(1− P)}. AM methods allow choosing of the inner diverse materials, Adv. Mat. 34 (2022), 2108855.
pattern adjusted to the intended application of the porous cordierite [18] E. Peng, D. Zhang, J. Ding, Ceramic robocasting, Recent achievements, potential,
and future developments, Adv. Mat. 30 (2018), 1802404.
material, while the structure is maintained during the reaction/sintering [19] M. Rafiee, R.D. Farahani, D. Therriault, Multi-material 3D and 4D printing: a
processes, and appropriate mechanical resistance as well as CTE and KT survey, Adv. Sci. 7 (2020), 1902307.
values are achieved. Therefore, honeycomb type cordierite structures [20] J.A. Lewis, Direct ink writing of 3D functional materials, Adv. Funct. Mater. 16
(2006) 2193–2204.
can be directly printed by robocasting from raw materials mixtures [21] Y. Lakhdar, C. Tuck, J. Binner, A. Terry, R. Goodridge, Additive manufacturing of
containing kaolin without any printing additive, thus the present pro­ advanced ceramic materials, Prog. Mater. Sci. 116 (2021), 100736.
cess avoids the need for previously reacted low cost raw materials [22] A.J. Allen, I. Levin, S.E. Witt, Materials research & measurement needs for ceramic
additive manufacturing, J. Am. Ceram. Soc. 103 (2020) 6055–6069.
mixtures providing similar properties.
[23] L. Moreno-Sanabria, C. Ramírez, M. I. Osendi, M. Belmonte, P. Miranzo. Enhanced
thermal and mechanical properties of 3D printed highly porous structures based on
Declaration of competing interest γ-Al2O3 by adding graphene nanoplatelets, Adv. Mat. Technol. 2022, 2101455.
[24] A. M’Barki, L. Bocquet, A. Stevenson, Linking rheology and printability for dense
and strong ceramics by direct ink writing, Sci. Rep. 7 (2017) 6017.
The authors declare that they have no known competing financial [25] S.S.L. Chan, R.M. Pennings, L. Edwards, G.V. Franks, 3D printing of clay for
interests or personal relationships that could have appeared to influence decorative architectural applications: effect of solids volume fraction on rheology
the work reported in this paper. and printability, Addit. Manuf. 35 (2020), 101335.
[26] C. F. Revelo, H. A. Colorado. 3D printing of kaolinite clay ceramics using the direct
ink writing (DIW) technique. Ceram. Inter. 44 (2918) 5673-5682.
Acknowledgements [27] E.F. Osborn, A. Muan, Phase Equilibrium Diagrams of Oxide Systems, Published by
the American Ceramic Society and the Edward Orton Jr, Ceramic Foundation,
Columbus, OH, 1960, p. 246.
This work was supported by the Spanish Government through pro­ [28] M. Lorenz, B. Dietemann, L. Wahl, C. Bierwisch, T. Kraft, H. Kruggel-Emden,
jects RTI2018-095052-B-I00 (MICINN/AEI/FEDER, UE) and EIN2020- N. Travitzky, Influence of platelet content on the fabrication of colloidal gels for
112153 (MCINN/AEI/10.13039/501100011033), the latter also sup­ robocasting: experimental analysis and numerical simulation, J. Eur. Ceram. Soc.
40 (2020) 811–825.
ported by the European Union through “Next Generation EU/PRTR”. L.
M.S. acknowledges the financial support from MICINN/AEI/FEDER

4584
B. Dorado et al. Ceramics International 49 (2023) 4578–4585

[29] F.A. Costa Oliveira, S. Dias, M.F. Vaz, J.C. Fernandes, Behaviour of open-cell [36] P. Miranda, A. Pajares, F. Guiberteau, Finite element modeling as a tool for
cordierite foams under compression, J. Eur. Ceram. Soc. 26 (2006) 179–186. predicting the fracture behavior of robocast scaffolds, Acta Biomater. 4 (2008)
[30] S. Lee, J.H. Ha, J. Lee, I.H. Song, Enhanced mechanical strength of talc-containing 1715–1724.
porous kaolin prepared by a replica method, J. Korean Ceram. Soc. 58 (2021) [37] E. Feilden, E. García-Tuñón Blanca, F. Giuliani, E. Saiz, L. Vandeperre, Robocasting
123–133. of structural ceramic parts with hydrogel inks, J. Eur. Ceram. Soc. 36 (2016)
[31] H. Li, C. Lia, L. Wu, H. Wang, J. Li, M. Fu, C.A. Wang, In-situ synthesis and 2525–2533.
properties of porous cordierite ceramics with adjustable pore structure, Ceram. Int. [38] H. Schneider, E. Eberhard, Thermal expansion of mullite, J. Am. Ceram. Soc. 73
46 (2020) 14808–14815. (1990) 2073–2076.
[32] L.G. Gibson, M.F. Ashby, Cellular solids, in: Cambridge Solid State Science Series, [39] J.F. Sarver, F.A. Hummel, Stability relations of magnesium metasilicate
second ed., Cambridge University Press, 1997, pp. 175–234 (Chapter 5). Polymorphs, J. Am. Ceram. Soc. 45 (1962) 152–156.
[33] J. Maurath, N. Willenbacher, 3D printing of open-porous cellular ceramics with [40] E. García, M.I. Osendi, P. Miranzo, Thermal diffusivity of porous cordierite ceramic
high specific strength, J. Eur. Ceram. Soc. 37 (2017) 4833–4842. burners, J. Appl. Phys. 92 (2002) 2346–2349.
[34] Z. Chen, C. Liu, J. Li, J. Zhu, Y. Liu, C. Lao, J. Feng, M. Jiang, C. Liu, P. Wang, Y. Li, [41] X. Xu, Y. Zhang, Ji Wu, C. Hu, Z. Tang, Preparation and performance study of
Mechanical properties and microstructures of 3D printed bulk cordierite parts, cordierite/mullite composite ceramics for solar thermal energy storage, Int. J.
Ceram. Int. 45 (2019) 19257–19267. Appl. Ceram. Technol. 14 (2017) 162–172.
[35] A. Yamuna, R. Johnson, Y.R. Mahajan, M. Lalithambika, Kaolin-based cordierite [42] W. Pabst, E. Gregorová, Conductivity of porous materials with spheroidal pores,
for pollution control, J. Eur. Ceram. Soc. 24 (2004) 65–73. J. Eur. Ceram. Soc. 34 (2014) 2757–2766.

4585