Available online at www.sciencedirect.

com

Acta Materialia 57 (2009) 2919–2932

www.elsevier.com/locate/actamat

Designing highly toughened hybrid composites through

nature-inspired hierarchical complexity
M.E. Launey a, E. Munch a, D.H. Alsem a,b, H.B. Barth c,d, E. Saiz a,
A.P. Tomsia a, R.O. Ritchie a,d,*
a
Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA
b
National Center for Electron Microscopy, Lawrence Berkeley National Laboratory, CA 94720, USA
c
Experimental Systems Group, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA
d
Department of Materials Science & Engineering, University of California, Berkeley, CA 94720, USA

Received 12 February 2009; received in revised form 28 February 2009; accepted 1 March 2009
Available online 6 April 2009

Abstract

The notion of replicating the unique fracture resistance of natural composites in synthetic materials has generated much interest but
has yielded few real technological advances. Here we demonstrate how using ice-templated structures, the concept of hierarchical design
can be applied to conventional compounds such as alumina and poly(methyl methacrylate) (PMMA) to make bulk hybrid materials that
display exceptional toughness that can be nearly 300 times higher (in energy terms) than either of their constituents. These toughnesses
far surpass what can be expected from a simple ‘‘rule of mixtures”; for a !80% Al2O3–PMMA material, we achieve a KJc fracture tough-
ness above 30 MPa m1/2 at a tensile strength of !200 MPa. Indeed, in terms of specific strength and toughness, these properties for alu-
mina-based ceramics are at best comparable to those of metallic aluminum alloys. The approach is flexible and can be readily translated
to multiple material combinations.
Published by Elsevier Ltd. on behalf of Acta Materialia Inc.

Keywords: Ceramics; Composites; Toughness; Casting

1. Introduction atomic to the macro levels, and their interactions. Further-

more, it will be essential to develop new fabrication tech-
As existing structural materials reach their performance niques that are able to synthesize such materials in bulk
limits, one of the major scientific challenges for the 21st form.
century is the development of new stronger and tougher One approach in this quest for the design of new and
lightweight structural materials to support advances in superior structural materials is by mimicking the architec-
diverse strategic fields from building to transportation or ture of natural/biological materials and structures [1,2].
energy. This is a challenge that can only be met through Nature combines readily available compounds that typi-
an understanding of the relationships between materials cally exhibit poor macroscale mechanical properties, e.g.
architecture and mechanical response, spanning not simply brittle biological ceramics and compliant macromolecules,
microstructure but encompassing the influence of struc- to produce composites, such as bone, nacre or tooth den-
tural parameters acting at multiple length scales, from the tin, that are far stronger and tougher than could be
expected from the simple mixture of their individual com-
*
ponents [3,4]. The prospect of extending natural designs
Corresponding author. Address: Department of Materials Science &
Engineering, University of California, Berkeley, CA 94720, USA. Tel.: +1
to multiple material combinations is very appealing, but
510 486 5798; fax: +1 510 643 5792. an understanding of how natural structures achieve their
E-mail address: RORitchie@lbl.gov (R.O. Ritchie). unique mechanical response has only recently begun to

1359-6454/$36.00 Published by Elsevier Ltd. on behalf of Acta Materialia Inc.

doi:10.1016/j.actamat.2009.03.003
2920 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932

emerge. It is clear that natural materials [4–7] generate their ceramic) to exhibit further unique combinations of
fracture resistance or toughness primarily by extrinsic strength and toughness.
mechanisms [8] which ‘‘shield” an advancing crack from
the applied load. These mechanisms, which are quite differ- 2. Materials and methods
ent to those that toughen metals, for example, are simulta-
neously created at multiple dimensions, literally from 2.1. Processing
nano- to macro-scales. From a fracture mechanics perspec-
tive, they generate a characteristic crack resistance curve Lamellar alumina scaffolds were prepared by freeze-cast-
(R-curve) behavior where the fracture resistance actually ing water-based suspensions. The suspensions (solid content
increases with crack extension; in other words, these mate- 50 wt.%) were prepared by dispersing submicrometer
rials develop the majority of their toughening during crack alumina powders (SM8 high-purity submicrometer alumina
growth and not during crack initiation [9]. powder, Baikowski) in deionized water with alumina balls.
A major difference between natural and synthetic mate- An ammonium polymethacrylate anionic dispersant (Dar-
rials is clearly in the way that they are made. Nature grows van CN, R.T. Vanderbilt Co., Norwalk, CT, 1.4 wt.% of
complex structures from the ‘‘bottom up” using the princi- the powder) and an organic binder (polyvinylalcohol,
ples of biologically controlled self-assembly [3,4]. However, 1.4 wt.% of the powder) were added to achieve a good
this approach is generally not feasible for making synthetic powder dispersion and ensure the integrity of the ceramic
materials in practical sizes with severe time constraints. structure after eliminating the water. Sucrose (b-D-fructofur-
Although there are several outstanding examples of mate- anosyl a-D-glucopyranoside, 4 wt.% of the water content)
rials fabricated in the laboratory by applying physicochem- was also added to provide a binder and to control the struc-
ical principles taken from the growth of natural ture of the growing ice crystals, and to ensure the formation
composites, these are invariably limited to thin films or of fine ceramic lamellae with a characteristic microscopic
microscopic samples [10–14]. Largely because of this, the roughness [18]. The suspensions were placed in cylindrical
field of biomimicry has generated few practical new mate- Teflon molds (25–50 mm diameter, 25–50 mm tall) sitting
rials. The key challenge here is to find ‘‘top-down” fabrica- on a copper cold finger. The temperature of this cold finger
tion approaches that are able to mimic natural structural was monitored and manipulated by combining a cold bath
designs over as many length scales as possible. with ring heaters (Fig. 1). In order to obtain parallel ceramic
In this work, freeze casting (also known as ice-templat- lamellae aligned over macroscopic dimensions (centimeters),
ing) is used to fabricate bulk hybrid materials with unique the copper surface where the ice nucleates was scratched uni-
hierarchical structures modeled on the structure of nacre directionally with silicon carbide paper (grade P400) and
(abalone shell). This technique uses the freezing of cera- vibration was applied to the mold during freezing [18]. The
mic suspensions to form porous or dense ceramics with cold finger was cooled to "80 !C with the cooling rate vary-
complex shapes and architectures. In particular, when ing between 1 and 20 !C min"1. The frozen samples were
freezing of the suspensions is carried out directionally then freeze-dried to sublimate the water and subsequently
using a temperature gradient, it is possible to fabricate fired in air at 400 !C for 2 h to remove the organics and after-
lamellar scaffolds whose microstructure resembles that of wards sintered at 1500 !C for 2 h to densify the ceramic
the inorganic component of nacre at multiple length scales lamellae. This resulted in a bulk scaffold, several centimeters
[15–17]. Here we show how this technology can be used to in size (Fig. 2), formed by macroscopically oriented dense
build nacre-like composites by filling the lamellar alumina lamellae.
scaffolds with a second, relatively soft, organic phase The ceramic scaffolds with finer lamellae (!5 lm wide,
(poly(methyl methacrylate), PMMA), which acts as the produced with a cold finger cooling rate of 20 !C min"1)
non-load-bearing ”lubricant” phase. The combination of were used as the base for the fabrication of ceramic
alumina and PMMA allows the fast, systematic explora- ‘‘brick-and-mortar” structures with !20 vol.% porosity.
tion of the effect of structural parameters at multiple The scaffolds were infiltrated with a short-chain polymer
length scales in the mechanical response. The final mate- with a low melting point (paraffin wax) that acted as a sac-
rials can have ceramic contents up to !80 vol.% and, like rificial phase. The infiltrated scaffolds were uniaxially
their natural counterparts, exhibit rising R-curve behavior pressed at 80 !C, just above the melting point of the
with fracture toughnesses up to 300 times higher (in organic phase, at !50–100 MPa. While liquid, the polymer
energy terms) than their main constituent, Al2O3. In this flows and allows the densification of the ceramic structure,
regard, they represent the toughest ceramic-based materi- and after cooling it holds the bricks together for transfer
als ever made. Such unprecedented mechanical behavior is into an air furnace. A thermal treatment in air was applied
interpreted in terms of the prevalent mechanisms of to remove the organic phase from the structure (2 h at
microstructural damage and toughening, incorporating 400 !C) and to promote densification (2 h at 1500 !C).
the concept of the ‘‘lubricant phase”. Our intent is to Finally the samples were pressed isostatically at 1.4 GPa
incorporate the design lessons and fabrication technolo- followed by a new sintering step (2 h at 1500 !C) to pro-
gies from this work in the development of novel hybrid mote the formation of inorganic (ceramic) bridges between
materials that will combine diverse phases (e.g. metal– the bricks.
 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932 2921

Fig. 1. (a) The freeze-casting set-up showing how the ceramic suspension is placed on top of a cold finger that is cooled at a constant rate (bottom line in
the graph) in order to form a temperature gradient that promotes the formation of lamellar ice. The ice expels the ceramic particles as it grows to leave a
layered ceramic structure after water removal. (b) The balance between imposed (parallel to the temperature gradient) and crystallographic preferred
growth directions results in a characteristic dendritic roughness in one of the sides of the growing ice crystals. This roughness is replicated in the surface of
the ceramic lamellae.

Fig. 2. (a) Porous scaffolds of practical dimensions obtained by freeze-casting of ceramic suspensions. The control of the processing conditions leads to
macroscopic samples that exhibit (b) lamellar structures oriented over several centimeters followed by polymer infiltration (the lighter phase is the ceramic;
the dark phase is the polymer). k represents the microstructural wavelength, and d the thickness of the lamellae. (c) The process results in the formation of
ceramic bridges between lamellae due to the trapping of ceramic particles by the growing ice. (d) By using sucrose as an additive to the freeze-casting slurry
the growing ice crystals develop a characteristic surface topography that translates into a microscopic roughness in the ceramic walls.
2922 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932

The scaffolds were infiltrated with a second polymeric methacrylate initiated by AIBN (0.5 wt.% AIBN in MMA)
phase by a free radical polymerization in bulk of methyl- for 8 h at 50 !C in nitrogen. The use of identical polymer-
methacrylate (MMA) initiated by 2,20 -azobisisobutyronit- ization conditions led to polymers with identical molecular
rile (AIBN, 0.5 wt.% in MMA) at 70 !C for 12 h. The weights.
infiltrated samples were annealed in an oven at 150 !C To compare the behavior of our hybrid composites with
for 2 h in gaseous nitrogen to complete the polymerization that of nacre, small beams were prepared from the nacre-
reaction and slowly cooled (!1 !C min"1) to room temper- ous layer of a red abalone shell (Haliotus rufescens). Raw
ature to minimize thermally induced residual stresses. samples were first cut from the shell using a low-speed dia-
In order to enhance adhesion at the organic/inorganic mond saw. The calcitic layer was ground away to leave the
interface, 3-(trimethoxysilyl)propyl methacrylate (c-MPS) nacre and the samples were then lapped on both faces.
was grafted on the surface of selected scaffolds prior to Finally, thin samples were sliced to obtain the final beams
the polymer infiltration [19]. The scaffolds were first (20 mm long, 2.3–2.4 mm wide and 1.2–1.4 mm thick).
immersed in a Piranha solution (1:1 H2O2:30 vol.% Samples were kept immersed in water so that they
H2SO4) for 10–15 min in order to promote the formation remained hydrated until testing.
of hydroxyl groups in the internal surfaces [20]. After care- Rectangular beams for bending tests with the tensile
fully rinsing with deionized water, the scaffold was faces parallel to the ceramic layers were sectioned from
immersed in a solution of the grafting agent (c-MPS), in the infiltrated scaffolds by using a water-cooled, low-speed
acetone (50 vol.% acetone, 50 vol.% c-MPS) for 6–12 h. diamond saw. The final specimens were 18–28 mm long,
The objective was to form a covalent bond between the 1.3–1.5 mm thick (B) and 3.0–3.2 mm wide (W). Plane-
c-MPS molecules and the hydroxyl groups on the ceramic strain fracture toughness, KIC, and R-curve measurements
surface. The c-MPS acts as a coupling agent between the were performed on single-edge notched bend (SE(B)) spec-
PMMA and the alumina, yielding stronger adhesion at imens. An initial notch was applied with a low-speed dia-
the organic/inorganic interface after infiltration. mond saw and was subsequently sharpened by repeatedly
The infiltration of the grafted scaffolds was performed sliding a razor blade over the saw-cut notch using a cus-
using two polymerization steps. The first step was a free rad- tom-made rig, while continually irrigating with 1 lm dia-
ical polymerization of MMA initiated by AIBN run in a sol- mond slurry. The final micro-notches had a root radius
vent (toluene, 2:1 toluene:MMA by weight) with a relatively of !3–5 lm. Sharp cracks with initial crack lengths, a, of
high initiator concentration (1 wt.% AIBN in MMA, 1 h at !1.6–1.7 lm were generated in general accordance with
70 !C in N2). The solvent helps to minimize the self-acceler- ASTM standards [21]. The orientation of the notch was
ation phenomenon inherent in the free radical polymeriza- such that the nominal crack-growth direction was perpen-
tion reaction and allows an increase in the amount of dicular to the ceramic layers.
activator. The high concentration of activator maximizes Alumina SE(B) specimens (24 mm long, 6.0 mm wide and
the chances for radicals to develop on c-MPS molecules 2.4–2.5 mm thick) were fatigue pre-cracked using a half-chev-
bonded to the ceramic surface, leading to chemically grafted ron starter notch (in addition to a Vickers hardness indent
PMMA chains. After evaporation of the solvent, the infiltra- placed at the notch tip) in order to facilitate crack initiation.
tion was performed by the bulk free radical polymerization SE(B) specimens of PMMA (22 mm long, 2.7–2.8 mm wide
previously described for a non-grafted material. and 1.2–1.4 mm thick) and hydrated nacre were micro-
The mechanical response of the hierarchical materials notched. Prior to testing, all specimens were polished on both
was compared to that of homogeneous particulate faces down to a 1 lm finish using diamond suspensions.
PMMA/Al2O3 nanocomposites, as well as dense PMMA
and alumina. The nanocomposites were prepared by press- 2.2. Mechanical characterization
ing uniaxially (20 MPa) the same Al2O3 powders used in
the preparation of freeze-cast scaffolds to form pellets 2.2.1. Strength and KIC measurements
(diameter 30 mm, height 15 mm) which were then sintered Three-point bend tests were performed to generate quan-
in air at 1100 !C for 2 h. The resulting porous samples (20– titative stress–strain curves and measure the plane-strain
45 vol.% porosity) were infiltrated with PMMA following fracture toughness, KIC. The strength tests used a support
the procedure described above for non-grafted materials. span S of 12.5 mm and a displacement rate of 1 lm s"1 in
Dense alumina samples were obtained by slip casting a general accordance with ASTM standards [22]. The tough-
water-based ceramic suspension (60 wt.% of alumina pow- ness measurements were performed on SE(B) specimens,
der in water). The suspensions for slip casting and freeze also using a support span of 12.5 mm. Values were deter-
casting used the same starting powders and similar solid mined by monotonically loading the specimens to failure at
contents with identical sintering cycles. Slip casting was a constant displacement rate of 1 lm s"1. All toughness tests
selected to prepare the reference ceramic materials because satisfied the plane-strain and small-scale yielding require-
the ceramic forms through the packing of powders from a ments for valid KIC measurement, as per ASTM standards
liquid suspension, as in freeze casting. Pure PMMA sam- [21]. All values presented represent at least an average of
ples were obtained by the polymerization in bulk of methyl- three measurements per configuration (n P 3).
 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932 2923

2.2.2. Fracture toughness J–R curve measurements take in account the fact that the crack is advancing and
Nonlinear-elastic fracture mechanics methods were used that the ligament b is continuously decreasing, an incre-
to evaluate the fracture toughness resistance in terms of the mental definition of Eq. (5) was used [21]:
J-integral as a function of crack extension, Da, under a # ! "! "$# $
monotonically increasing driving force. R-curves were mea- 1:9 AplðiÞ " Aplði"1Þ aðiÞ " aði"1Þ
J plðiÞ ¼ J plði"1Þ þ 1" :
sured on SE(B) specimens in situ in a scanning electron bði"1Þ B bði"1Þ
microscope using a three-point bending stage, again with ð6Þ
n P 3. Tests were conducted in three-point bending with
a span (S = 12.5 mm) to width (W ! 3.0–3.2 mm) ratio of Provided all validity criteria [21] were met, K-based
!4, in accordance with ASTM standard E1820 [21]. The fracture toughness values were back-calculated from
crosshead displacement was measured with a linear vari- the J measurements using the standard J–K equivalence
able-displacement transducer, while the load was recorded for nominally mode I fracture, specifically that
using a 150 N load cell. Prior to in situ testing, specimens KJ = (J.E0 )½.
were coated with a !20 nm thin layer of gold. R-curve
determination was limited to small-scale bridging condi- 2.3. Microstructural characterization
tions, where the size of the zone of crack bridges behind
the crack tip remained small compared to the in-plane test Microstructures were characterized using optical and
specimen dimensions. The use of nonlinear elastic fracture electron microscopy. Image analysis and thermogravimetry
mechanics with the J-integral as the driving force for crack were used to calculate the relative amounts of organic and
initiation and growth was employed to capture the contri- inorganic phases. The density of the porous scaffolds was
bution from inelastic deformation in the evaluation of the also measured by the Archimedes method. Total parame-
toughness of the hybrid composites. ters micro X-ray computed tomography (lXCT) was
The stress intensity at each measured crack length was employed to visualize in three dimensions the crack path
calculated by measuring the nonlinear strain-energy release and distribution of micro-damage after R-curve testing.
rate, J, where J is defined as the rate of change in potential The lXCT setup at the Lawrence Berkeley National Lab-
energy under nonlinear-elastic conditions for a unit oratory’s Advanced Light Source is similar to standard
increase in crack area [21]. J was calculated from the tomography procedures [23]; samples are rotated in a
applied load and instantaneous crack length according to monochromatic X-ray beam and the transmitted X-rays
ASTM standards [21], and was decomposed into its elastic imaged via a scintillator, magnifying lens and a digital cam-
and plastic contributions: era to give an effective voxel size in the reconstructed three-
dimensional (3-D) image of 4.45 lm. The samples were
J ¼ J el þ J pl : ð1Þ scanned in absorption mode and the reconstructed images
The elastic contribution Jel is based on linear-elastic were obtained using a filtered back-projection algorithm.
fracture mechanics: In absorption mode the grayscale values of the recon-
structed image are representative of the absorption coeffi-
K 2I cient. To maximize the signal-to-noise ratio, the energy
J el ¼ ; ð2Þ
E0 was selected at 15 keV; this optimizes the interaction
where KI is the mode I stress-intensity factor, and E’ = E in between the X-rays and the sample. Two-dimensional
plane stress and E/(1 " m2) in plane strain (m is Poisson’s images were taken every quarter of a degree between 0!
ratio). For SE(B) specimens, KI is given in terms of the and 180!. The data sets were reconstructed using the soft-
applied load P by: ware Octopus [24] and the 3-D visualization was performed
using AvizoTM software [25].
PS
KI ¼ f ða=W Þ; ð3Þ
BW 3=2
3. Results
where f is a non-dimensional function of the crack length a:
f ða=W Þ 3.1. Microstructural features
1=2 2
3ða=W Þ ½1:99 " ða=W Þð1 " a=W Þð2:15 " 3:93ða=W Þ þ 2:7ða=W Þ Þ(
¼ : 3.1.1. Lamellar composites
2ð1 þ 2a=W Þð1 " a=W Þ3=2
The controlled directional freezing of water-based cera-
ð4Þ mic suspensions was used to create porous ceramic scaf-
Using the load-line displacements, the plastic compo- folds, several centimeters in size (Fig. 2a). By cooling the
nent Jpl for a stationary crack in bending is given by: cold finger at a constant rate starting from room tempera-
ture it was possible to promote the directional growth of
1:9Apl
J pl ¼ ; ð5Þ lamellar ice. The ice expels the ceramic particles as it grows;
Bb subsequent water sublimation and sintering results in a
where Apl is the plastic area under force vs. displacement scaffold formed by ceramic layers of homogeneous thick-
curve, and b is the uncracked ligament length (W–a). To ness running continuously from the bottom to the top of
2924 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932

the sample (Fig. 2c). The layers are dense and aligned over 3.1.2. Brick-and-mortar composites
macroscopic dimensions (centimeters), and are connected Hot pressing of the lamellar material and subsequent
by thin (61 lm) ceramic bridges (Fig. 2c), with a grain size sintering at 1500 !C resulted in brick-and-mortar architec-
of the order of 1 lm. tures with large ceramic contents of up to 80 vol.%
A key characteristic of the freeze-casting process is the (Fig. 3a). The fine-grained (1 lm) alumina ‘‘bricks” were
ability to manipulate the materials structure at multiple 5–10 lm wide and 20–100 lm long. After infiltrating these
dimensions. The microstructural wavelength, k, and the structures with PMMA, the polymer layers had an overall
lamellae thickness, d (Fig. 2), are primarily a function of thickness of !1–2 lm; however, there were large regions in
the ice-front velocity (controlled by the cooling rate of which the alumina bricks were separated by submicrometer
the cold finger) and to a lesser extent of the ceramic particle polymer films (Fig. 3b). Another important feature of these
size [26,27]. The faster the speed of the freezing front (the brick-and-mortar architectures was the formation of a high
faster the cooling rate of the finger), the finer the micro- density of ceramic bridges between bricks during sintering.
structural wavelength and the lamellae thickness. As our These bridges (typically of submicrometer width) provide
objective was to mimic the nacre structure and fabricate strong and stiff links between the ceramic layers (Fig. 3b
samples with the larger ceramic contents, we maximized and c). The brick-and-mortar materials mimic the structure
the solid content of the suspensions. The maximum alu- of nacre at several length scales, in that they display larger
mina concentrations that could be used to consistently fab- ceramic contents, thin polymer layers, a characteristic
ricate lamellar materials with macroscopically oriented interfacial roughness and a relatively large density of these
ceramic layers were of the order of 50 wt.% (!20 vol.%). thin inorganic bridges. However, they are still an order of
The porosity of the resulting scaffolds after sintering was magnitude coarser than the natural material.
!64 vol.% and the lamellae thickness, d, varied between
!5 and 20 lm for cold finger cooling rates between 20 3.2. Stress–strain curve behavior
and 1 !C min"1, respectively [26]. Faster ice growth can
be achieved by reducing the temperature gradient during Bending stress–strain curves for the Al2O3–PMMA
freezing (e.g. by using a top cold finger); however, this hybrid materials replicate qualitatively those of nacre by
can result in a tilted lamellae growth as the ice growth showing >1% inelastic deformation prior to failure
direction depends on the balance between the preferred (Fig. 4a). The ultimate flexural strengths are summarized
crystallographic orientation and that which is forced by in Table 1. The strengths in the direction perpendicular
the temperature gradients [26]. Eventually, the lamellae to the ceramic layers are comparable to that of pure
thickness is limited by the alumina particle size (!1 lm Al2O3, with average values of 116 and 210 MPa for the fin-
after sintering). Subsequent infiltration with PMMA est lamellae and brick-and-mortar structures, respectively
resulted in a composite with a uniform two-phase lamellar (Table 1 and Fig. 4b and d). By comparison, the flexural
microstructure (Fig. 2b) with a !36 vol.% of ceramic phase strengths of pure Al2O3 and PMMA were found to be
as determined by both thermogravimetry and image 325 ± 20 and 78 ± 5 MPa, respectively, while the fracture
analysis. strength of hydrated nacre is 135 ± 4 MPa, consistent with
During the growth of ice, the balance between imposed previously reported data [28–32]. Decreasing the thickness
(parallel to the temperature gradient) and crystallographic of the ceramic layers had a slight effect on the flexural
preferred growth directions result in a slight tilting of the strengths (Fig. 4b). Furthermore, whereas increasing adhe-
ice lamellae and the formation of dendrites on one of the sion at the Al2O3–PMMA interface though chemical graft-
sides (Fig. 1). After removing the water, these dendrites ing resulted in a relatively small increase in the strength of
are reflected in a characteristic roughness on one of the lamellar structures, it had a very large effect (more than
sides of the ceramic layers while the other remains very 80% increase) on the strength of the brick-and-mortar
smooth at the microscopic level (Fig. 1). This roughness structures (Fig. 4).
plays a critical role in controlling shear at the organic/
inorganic interface during loading and can be manipu- 3.3. Plane-strain fracture toughness, KIC
lated using different additives in the suspension. The addi-
tives determine the shape of the growing ice crystals by The KIC fracture toughness values for the hybrid mate-
modifying the interfacial tensions, the degree of superco- rials, which represent a toughness for crack initiation, KIC,
oling ahead of the ice front and the viscosity and the were almost double of what could be expected from the
phase diagram of the solvent (promoting, for example, simple ‘‘rule of mixtures” of the properties of the two con-
the formation of fugitive phases that are eliminated dur- stituents (Fig. 4c and d and Table 1). Refinement of the
ing sintering) [18]. In the present study, the use of sucrose lamellae size slightly improved the initiation toughness of
resulted in the formation of a characteristic microscopic lamellar structures (Fig. 4c). As for the strength properties,
roughness characterized by relatively smooth dendritic- grafting of the Al2O3–PMMA interface also resulted in a
like features, 1–5 lm high, running in the direction of small increase of KIC for the lamellar structures (Fig. 4c),
solidification (Fig. 2d) on one of the lamellae sides but in a significant enhancement in toughness (almost
[18,26]. 70%) for brick-and-mortar structures (Fig. 4d).
 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932 2925

shown in Fig. 5b. According to active ASTM standards

[21], the maximum J capacity for a specimen is given by
the smaller of Jmax = bry/20 or Bry/20; similarly, the max-
imum crack extension capacity is given by: Damax = 0.25b.
The J–R curve is therefore defined as the data in a region
bounded by these Jmax and Damax limits; for example, the
maximum measurement capacity of the brick-and-mortar
specimens were Jmax = 15 kJ m"2 (with B = 1.5 mm) and
Damax = 0.4 mm (with b = 1.6 mm), as illustrated in
Fig. 5b. The ASTM standards provide a somewhat arbi-
trary conservative limit. However, we can define a critical
toughness, Jc, using a physical criterion as the maximum
valid J-integral before the measurement become geometry
dependent due to large-scale bridging (Fig. 5b). This crite-
ria yields J-integral values that are slightly above the max-
imum crack extension, Damax, but have a clear physical
meaning. Most importantly, the conditions for J-domi-
nance were met in all the tests such that b, B P 25 (Jc/
ry). This latter criterion ensures that the J-integral values
represent valid fracture toughness values.
The R-curves can also be expressed with a K-based
description (termed KJ), using the nominal mode I J–K
equivalence KJ = (J ) E0 )½, with E0 = E/(1 " m2) in plane
strain (m is Poisson’s ratio). These calculations require
knowledge of the Young’s modulus. The structure of the
lamellar and brick-and-mortar hybrid composites is clearly
directional; properties are orthotropic, with values of E
lying between the upper and lower bound of the ‘‘rule of
mixtures”, as defined by the Reuss and Voigt models. Mea-
surements using three-point bend tests showed that the
elastic modulus values fell approximately at the mid-point
between these upper and lower bounds. Using E values for
PMMA of 3 GPa and for Al2O3 of 300 GPa, the elastic
moduli for the lamellar and brick-and-mortar hybrid struc-
tures at 35 and !80 vol.% ceramic content were calculated
to be 57 and 115 GPa, respectively, in the direction perpen-
dicular to the ceramic layers. These values were confirmed
experimentally by contact ultrasonic measurements using a
pulse–echo overlap technique with MHz shear and longitu-
dinal piezoelectric transducers, and a pulser/receiver to
measure the shear and longitudinal wave speeds at room
temperature. Note that a relatively small error in the value
of E has only a very limited impact on the computed KJ
toughness; specifically, a ± 20% error on E results in just
a ± 1% error in KJ.
R-curves are shown in Fig. 6 in terms of the equivalent
stress intensity, KJ, as a function of crack extension, Da, for
Fig. 3. (a) Brick-and-mortar architectures with large ceramic contents up
the lamellar (Fig. 6a) and brick-and-mortar (Fig. 6b)
to 80 vol.% are prepared through pressing of the lamellar materials and
subsequent sintering. (b) These structures display large areas with hybrid composites. It is apparent that both natural (nacre)
submicrometer polymer films between alumina blocks. (c) Ceramic bridges and synthetic composites exhibit marked rising R-curve
between bricks are formed during the second sintering steps. behavior indicative of extensive extrinsic toughening. The
degree of toughening in the hybrid composites actually sur-
3.4. Resistance-curve behavior passes that of nacre (the often cited ‘‘gold standard” in bio-
mimetic design). The grafted Al2O3–PMMA lamellar and
The load-line displacement data (e.g. Fig. 5a) were ana- brick-and-mortar structures reach (steady-state) fracture
lyzed to evaluate the JR(Da) resistance curves. Typical toughnesses KJc of up to 15 MPa m1/2 (Jc ! 5 kJ m"2)
results for the grafted brick-and-mortar structures are and 30 MPa m1/2 (Jc ! 8 kJ m"2), respectively (Table 1).
2926 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932

Fig. 4. (a) Bending stress–strain curves for the Al2O3–PMMA hybrid materials mimic those of nacre, and show >1% inelastic deformation prior to failure.
The curves correspond to non-grafted lamellar hybrid composite and hydrated nacre (abalone shell). The role of the interfacial chemistry (grafting) in
raising the interface energy is illustrated for both the lamellae and brick-and-mortar structures. Grafting increases both the (b) strength and (c) crack-
initiation toughness, KIC, of the lamellar composites. (d) This improvement is much more pronounced in the brick-and-mortar structure.

Table 1 (1 lm) alumina, these homogeneous Al2O3–PMMA com-

Mechanical properties of the hybrid composites. The properties listed are posites display minimal extrinsic toughening and conse-
crack-initiation fracture toughness (KIC), flexural strength (rf) and crack- quently no rising R-curve behavior with very low fracture
growth fracture toughness at instability (KJc).
toughness below 2 MPa m1/2.
Structure/ Interfacial Lamellae/ KIC rf KJc As for the strength and crack-initiation toughness prop-
ceramic chemistry brick (MPa m1/2) (MPa) (MPa m1/2)
erties, grafting only offers a marginal increase in KJc for the
content thickness
(lm) lamellar materials (Fig. 6a), but a significant enhancement
(a factor of 2 higher) for the brick-and-mortar structures
Lamellar Non-grafted 20 3.2 ± 0.3 90 ± 7 12 ± 2
36 vol.% 10 3.7 ± 0.1 100 ± 9 (Fig. 6b). Refinement of the lamellae width did not affect
5 3.8 ± 0.2 104 ± 19 the R-curve behavior of the lamellar composites and
Grafted 20 3.6 ± 0.1 112 ± 5 16 ± 2 yielded similar KJc values (Fig. 6a).
10 3.9 ± 0.1 113 ± 7
5 4.2 ± 0.3 116 ± 10
3.5. Crack-growth observations
Brick-and- Non-grafted 5 3.1 ± 0.1 115 ± 6 15 ± 2
mortar Grafted 5.1 ± 0.6 210 ± 13 32 ± 5 Stable crack growth during the R-curve measurements
80 vol.%
provides an opportunity to observe directly the toughening
mechanisms and their relationship to the characteristic fea-
The latter is an outstandingly high toughness for a material tures of the microstructure. By performing these tests
consisting of 80% alumina with strength above 200 MPa; it in situ in the scanning electron microscope, it is possible
is far in excess of the toughness of either constituent, and to measure the R-curve while simultaneously monitoring
we believe that it is probably the toughest ceramic material the evolution of damage mechanisms ahead of the crack
reported to date. Indeed, the KJc toughness of the grafted tip and the extrinsic shielding mechanisms acting in the
brick-and-mortar hybrid materials is an order of magni- wake. Figs. 7 and 8 show typical examples of crack growth
tude higher than the standard hot-pressed homogeneous in the lamellar and brick-and-mortar architectures, respec-
nanocomposites consisting of 500 nm Al2O3 particles dis- tively. Several mechanisms are clearly visible: (i) extensive
persed in 55 vol.% PMMA. Similar to the fine-grained inelastic deformation resulting in a wide (millimeter-scale)
 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932 2927

Fig. 5. (a) Typical load vs. load-line displacement curve, and (b) JR(Da)–R curve for stable crack extension in the grafted brick-and-mortar composite. The
validity of the data points is defined by the measurement capacity of each specimen in accordance with ASTM standards [21]. The critical toughness, Jc,
here was defined as the maximum J-integral before the test data becomes geometry-dependent due to large-scale bridging (i.e. crack-wake bridging zones
are no longer small compared to the in-plane specimen dimensions).

Fig. 6. Crack-resistance curves showing resistance to fracture in terms of the stress intensity, KJ, as a function of crack extension, Da, for the (a) lamellar,
and (b) brick-and-mortar hybrid Al2O3–PMMA composites. The fracture toughness, KJ, is back-calculated from the J measurements, shown in Fig. 4,
using the J–K equivalence for mode I fracture (see text). R-curve data for natural nacre (abalone shell) are adapted from Ref. [6].

distribution of damage around the crack tip (Figs. 7a and of polymer tearing and stretching over micrometer dimen-
9)—this damage principally takes the form of microme- sions, and ‘‘pull out” via frictional sliding between ceramics
ter-scale cracking of ceramic layers outside of the main bricks (Fig. 8).
crack path (Figs. 7a and 9); (ii) inelastic deformation within Post-mortem examination of the fracture surfaces of the
the polymer in the form of micrometer-scale tearing (Figs. lamellar and brick-and-mortar composites after R-curve
7 and 8); and (iii) elongated voids within the polymeric testing illustrates the effect of chemical grafting on the
interlayers of lamellar materials, many of which result from degree of delamination at the organic–inorganic interface
the processing, but which appear to promote local plastic during fracture (Fig. 10a and c). Better adhesion due to
flow and grow during loading (Fig. 7a). These brittle and chemical grafting limits delamination and the resulting
ductile damage processes appear to be quite distinct with fracture surfaces are comparatively much flatter (Fig. 10b
little evidence of interaction. and d).
The presence of such damage initiated ahead of (and dis-
tinct from) the main growing crack leads to the formation 4. Discussion
of uncracked-ligament bridging [33,34] in the wake of the
crack tip. These elastic bridges, which are quite distinct Both lamellar and brick-and-mortar materials are strong
from the pre-existing stiff ceramic bridges between layers and tough and display >1% inelastic strains when loaded in
and bricks, are formed in situ before the main growing tension (Fig. 4a). The bending strengths are comparable to
crack links to the microcracks and voids; the bridges effec- that of the ceramic, and the initiation toughnesses are well
tively span the crack and carry load that would otherwise above that of the alumina or PMMA. The refinement in
be used to promote macroscopic crack advance.1 In addi- lamellar spacing slightly improved both the strength and
tion, the brick-and-mortar materials show clear evidence KIC fracture toughness (Fig. 4b and c). This implies that
flaws in the material control the strength and subsequently
1
Instead of crack extension being solely associated with the main crack the initiation toughness, with the thicker lamellae repre-
tip growing forward, crack advance also occurs by microcracks (or other senting a larger sampling volume and a higher probability
damage) initiated ahead of the main crack tip linking back to the tip. of finding a large flaw. However, what distinguishes the
2928 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932

ing R-curve behavior indicative of significant extrinsic

toughening. While single-value linear-elastic parameters
based on crack initiation, such as KIC, have traditionally
been used to quantify toughness, they cannot capture, or
even represent, the multiple length-scale toughening acting
in these materials [9]. Evaluating their toughness requires
nonlinear-elastic fracture mechanics to characterize the
contributions from inelastic deformation, and more impor-
tantly, R-curve analysis to characterize the toughness asso-
ciated with crack growth, both of which are afforded by the
use of the J–R curve.
The in situ analysis of crack growth clearly shows how
control of the structural architecture at several length
scales permits the generation of multiple toughening mech-
anisms operating over a large range of dimensions; this cre-
ates extensive crack-tip shielding and unprecedented
fracture resistance for a ceramic-based material. Observed
toughening mechanisms include uncracked-ligament bridg-
ing in the crack wake (Figs. 7 and 9), grain bridging/pull-
out, microcracking, crack deflection/delamination, void
growth and the fracture of (pre-existing) ceramic bridges
at the micrometer scale, and inelastic deformation of the
polymer at submicrometer scales and higher; moreover,
the extent of the damage is not localized but rather shows
a wide distribution over millimeter dimensions. The elon-
gated voids in the polymeric layer can act as very effective
local crack arresters, in accord with the well-known
Cook–Gordon mechanism [35] in laminates, and the
energy dissipation to break the resulting crack bridges
leads to ductile-phase toughening [36]. These mechanisms
have clear parallels in natural structures such as cortical
bone and tooth dentin [7,37], not to mention rocks as well
as polymer and metal–matrix composites [33].
As in the lamellar materials, the brick-and-mortar
microstructures exhibit extensive elastic bridging; however,
unlike the lamellar materials there is little or no fracture of
the ceramic bricks, rather there is some degree of ‘‘pull-
out” between the bricks which in many respects replicates
at microscopic dimensions the nanoscale phenomena
observed in nacre (Fig. 8) [38]. This form of toughening
is particularly potent and the prime source of the fracture
resistance of most monolithic structural ceramics with engi-
neered grain boundaries, i.e. coarse-grained Al2O3, Si3N4
and SiC [39–46]. Nevertheless, the contribution from bridg-
ing alone does not account for the unusually high tough-
ness of the brick-and-mortar structure. There is a key
distinction between the lamellar and brick-and-mortar
Fig. 7. Scanning electron micrographs taken during in situ R-curve materials that leads to critical differences in the nature
measurements of the toughness of a lamellar Al2O3–PMMA structure.
and relative contribution of the different toughening mech-
The image taken during stable crack propagation shows the toughening
mechanisms acting at multiple length scales: (a) wide distribution of anisms. In lamellar structures, the relatively thick polymer
damage over millimeter dimensions in the form of contained microcrack- layer still plays a structural, load-bearing role. Conversely,
ing within the ceramic and the voids in the polymer layers, (b) ductile the brick-and-mortar structure has a high ceramic content
uncracked-ligament bridging, and (c) inelastic deformation within the and the thin organic phase is not load bearing, akin to
polymer in the form of micrometer-scale tearing.
many hard mineralized natural composites; rather it acts
as the ‘‘lubricant phase” and controls the sliding of the
freeze-cast materials is that, like natural composites, they load-bearing ceramic blocks (Fig. 8). In this respect, the
clearly exhibit extensive subcritical crack growth during ris- lubricant phase plays a role similar to dislocations in met-
 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932 2929

Fig. 8. Similarities between the toughening mechanisms acting in artificial (Al2O3–PMMA) and natural nacre. Scanning electron micrographs taken
during an in situ R-curve measurement of (a, c and e) a grafted brick-and-mortar structure and (b, d and f) of hydrated nacre (abalone shell) show similar
mechanisms although the nacre clearly has a finer structure. The image taken during stable crack propagation shows the toughening mechanisms acting at
multiple length scales: (a) ‘‘pull out” mechanisms similarly to that observed in (b) nacre, (c) polymer tearing and stretching over micrometer dimensions as
also observed in the (d) organic phase of nacre, and (e) frictional sliding resisted by the interface roughness of the ceramic bricks as observed in (f) nacre.
Note that the thin bright lines between the sliding grains in (e) results from the deformation of the gold coating during sliding; indeed, all the mechanisms
described above are rendered visible by electrical charging in the SEM resulting from the deformation of the gold coating during deformation.

als or microcracking in rocks in that it acts to relieve high on the brick-and-mortar structures; the stronger grafted
stresses in the structure, thereby reducing the risk of frac- interface enables the organic layer to act as a more effective
ture. As a result chemical grafting has a much larger effect viscoelastic glue, i.e. permitting yet limiting the extensive of
2930 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932

days of research on whiskers [49]. However, from the per-

spective of toughening, the use of the other larger length
scales should not be ignored; nanoscale reinforcements
alone are not the best direction to take [50,51]. The prime
extrinsic toughening mechanisms, namely crack deflection
and particularly crack bridging, are promoted by increas-
ing, not decreasing, reinforcement dimensions [52], as is
illustrated in the present work by the lack of toughening
observed in the Al2O3–PMMA homogeneous nanocom-
posites (Fig. 11).
Large reinforcements in the form of fibers or particles
have been successfully used to increase fracture resistance
[57]. Crack bridging is again the most prominent toughen-
ing mechanism, particularly in ceramic–matrix composites.
This mechanism is enhanced by utilizing fibers with weak
fiber–matrix bonding; when the matrix fails, the fibers are
left intact spanning the crack wake and can act as bridges
to inhibit crack opening [52–55]. With continuous-fiber
reinforced polymer composites, i.e. carbon-fiber-reinforced
polymers, high volume fractions of graphite fibers with
strong matrix–fiber interfaces are added for strength and
stiffness. In all these structures, it is generally difficult to
Fig. 9. Synchrotron X-ray tomography image of crack propagation from
a sharpened notch (on the left-hand side) during an R-curve test of a
achieve both high strength and high toughness; the trade-
lamellar Al2O3–PMMA composite, showing the wide distribution of off comes in part from the fact that both the matrix and
damage (in yellow) spread over several millimeters surrounding the the reinforcement play a load-bearing role and weakening
deflected crack path (in blue). Crack propagation is nominally from left to the interfaces to promote crack path deflection also weak-
right. (For interpretation of color mentioned in this figure the reader is ens the structure. With the current naturally inspired com-
referred to the web version of the article.)
posites, freeze-cast lamellar materials exhibit layer
thicknesses that can be almost an order of magnitude thin-
interfacial sliding, as in nacre (Fig. 8). In contrast, grafting ner than those of soft–hard composites prepared by con-
has a very limited role in lamellar materials where the poly- ventional techniques [56–60]. As a result they can raise
mer has a structural function. The roughness of the ceramic the degree of extrinsic toughening to values comparable,
interfaces and the presence of stiff (pre-existing) ceramic or even superior, to those of fiber-reinforced composites
bridges between grains with micrometer and submicrome- without any loss in strength. However, it is the brick-
ter dimensions (Figs. 2 and 3) also contribute to the con- and-mortar Al2O3–PMMA structures, where the soft
trolled sliding (‘‘sliding interference”) between the rough polymeric phase is no longer structural but serves as a
ceramic layers (Fig. 8), thereby enhancing the toughness stress-relieving lubricant, that provide the optimum combi-
through extremely efficient energy dissipation. The result nation of strength and toughness; these structures best
is a synthetic material which, like nacre, is far tougher that illustrate the unique advantages and opportunities enabled
what could be expected from the simple mixture of its con- by biomimetic design.
stituents (Fig. 11).
Two key design concepts emerge from the examination 5. Summary and conclusions
of these biomimetic, freeze-cast hybrid structures: the need
of multiscale hierarchical architectures and the use of soft Using the inspiration of natural materials, specifically
(non-load-bearing) phases that can act as a ‘‘lubricant”. nacre, we have developed a suite of ceramic-based hybrid
These concepts are quite different to those employed in materials with unprecedented combinations of strength
other recent studies to develop improved structural com- and toughness, which when normalized with respect to
posites, which of late have largely focused on the use of the density is comparable to those of metallic aluminum
nanoscale reinforcements such as carbon nanotubes (with alloys [16]. This has been achieved through careful manip-
their near-ideal strength and extreme stiffness), platelets ulation of structural architecture to replicate many of the
or nanofibers to increase toughness [47]. Although it has features underlying the unique properties of nacre, in par-
been claimed that materials reinforced at the nanoscale ticular its characteristic R-curve behavior which derives
would have exceptional properties, results to date have from a confluence of toughening mechanisms that emulate
been disappointing [48]. If the composite is to be used for those of nacre over multiple length scales. Like in nacre,
a small-volume structure, clearly the reinforcements must our best synthetic materials reflect the natural concept
also be small; moreover, small-volume reinforcements tend of a hard ceramic phase providing for material strength,
to be much stronger, as has been known since the early separated by a softer ‘‘lubricant” phase to relieve high
 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932 2931

Fig. 10. Scanning electron micrographs of the fracture surfaces of (a and b) lamellar and (c and d) brick-and-mortar Al2O3–PMMA composites. (a and c)
Non-grafted interfaces (which lead to weaker boundaries) show extensive interface delamination during fracture, while (b and d) chemical grafting results
in better adhesion (stronger boundaries) and fracture surfaces that are comparatively much flatter.

Fig. 11. J-based measurements of the fracture toughness for of the lamella
and brick-and-mortar Al2O3–PMMA composites, showing their excep-
tional fracture toughness as compared to that of their constituent phases.
The toughness of the pure components, along with that from the rule of
mixtures, and KJc measurements for Al2O3–PMMA homogeneous nano- Fig. 12. Back-scattered electron micrograph of a ceramic–metal compos-
composites, are also shown. ite, here Al2O3/Al–Si, with a lamellar microstructure containing 36 vol.%
of ceramic. The bright phase is Al–Si (d * 20 lm), and the dark phase is
stresses in order to enhance toughness. The result is the Al2O3 (d * 10 lm). Some of the observed toughening mechanisms, i.e.
highest toughness ceramic reported to date, with a ductile uncracked-ligament bridging and tearing, are similar to those
strength of 200 MPa and a fracture toughness that is observed in the polymer–ceramic hybrid composites.
2932 M.E. Launey et al. / Acta Materialia 57 (2009) 2919–2932

!300 times (in J-terms) larger than the main constituent, [18] Munch E, Saiz E, Tomsia AP, Deville S. J Am Ceram Soc 2009, in
Al2O3, and exceeds values of 30 MPa m1/2 (Jc ! press.
[19] Abboud M, Turner M, Duguet E, Fontanille M. J Mater Chem
8000 J m"2). 1997;7:1527.
Future work is being directed towards the formation of [20] Fakunle ES, Aguilar ZP, Shultz JL, Toland AD, Fritsch I. Langmuir
hybrid materials with much higher inorganic contents, the 2006;22:10844.
manipulation of the properties of the soft lubricating phase [21] ASTM E1820-06. Annual book of ASTM standards, vol. 03.01:
and extending this concept to other material combinations. Metals—Mechanical testing; Elevated and low-temperature tests;
Metallography. West Conshohocken, PA: ASTM International;
In particular, we are attempting to process ceramic scaf- 2006.
folds infiltrated with higher melting point metals with the [22] ASTM D790-03. Annual book of ASTM standards, vol. 08.01:
objective of developing strong and tough ceramic-based Plastics (I). West Conshohocken, PA: ASTM International; 2003.
materials that can operate at elevated temperatures. Preli- [23] Kinney JH, Nichols MC. Annu Rev Mater Sci 1992;22:121.
minary investigations on lamellar ceramic/metal materials, [24] Vlassenbroeck J, Dierick M, Masschaele B, Cnudde V, Hoorebeke L,
Jacobs P. Nucl Instrum Methods Phys Res Sect A 2007;580:442.
i.e. Al2O3/Al–Si (Fig. 12), show promising properties, spe- [25] Mercury CS. AvizoTM 3D visualization framework. Chelmsford, MA.
cifically that their strength and toughness significantly [26] Deville S, Saiz E, Tomsia AP. Acta Mater 2007;55:1965.
exceeds the rule of mixtures of the properties of the two [27] Deville S. Adv Eng Mat 2008;10:155.
constituent phases. [28] Currey JD. Proc R Soc London Ser B 1977;196:443.
[29] Jackson AP, Vincent JFV, Turner RM. Proc R Soc Lond Ser B
1988;234:415.
Acknowledgments [30] Sarikaya M. Microsc Res Tech 1994;27:360.
[31] Menig R, Meyers MH, Meyers MA, Vecchio KS. Acta Mater
This work was supported by the Director, Office of Sci- 2000;48:2383.
ence, Office of Basic Energy Sciences, Division of Materials [32] Wang RZ, Suo Z, Evans AG, Yao N, Aksay IA. J Mater Res
2001;16:2485.
Sciences and Engineering, of the US Department of Energy [33] Shang JK, Ritchie RO. Metall Trans A 1989;20:897.
under Contract No. DE-AC02-05CH11231. We acknowl- [34] Koester KJ, Ager JW, Ritchie RO. Nat Mater 2008;7:672.
edge the assistance of the X-ray microtomography beam- [35] Cook J, Evans CC, Gordon JE, Marsh DM. Proc R Soc Lond Ser A
line (8.3.2) at the Advanced Light Source (ALS) at 1964;282:508.
Lawrence Berkeley National Laboratory. The authors also [36] Sigl LS, Mataga PA, Dalgleish BJ, McMeeking RM, Evans AG. Acta
Metall 1988;36:945.
thank Prof. Daryl Chrzan and Mark Jhon for many helpful [37] Nalla RK, Kinney JH, Ritchie RO. Biomaterials 2003;24:3955.
discussions, and Dr. Mary Laura Lind and Prof. William [38] Lin AYM, Chen PY, Meyers MA. Acta Biomater 2008;4:131.
L. Johnson at the California Institute of Technology for [39] Becher PF, Sun EY, Plucknett KP, Alexander KB, Hsueh CH, Lin
the elasticity measurements. HT, et al. J Am Ceram Soc 1998;81:2821.
[40] Cao JJ, MoberlyChan WJ, DeJonghe LC, Gilbert CJ, Ritchie RO. J
Am Ceram Soc 1996;79:461.
References [41] Kruzic JJ, Cannon RM, Ritchie RO. J Am Ceram Soc 2004;87:93.
[42] Kruzic JJ, Cannon RM, Ager JW, Ritchie RO. Acta Mater
[1] Mayer G. Science 2005;310:1144. 2005;53:2595.
[2] Ortiz C, Boyce MC. Science 2008;319:1053. [43] Lange FF. J Am Ceram Soc 1973;56:518.
[3] Fratzl P, Weinkamer R. Prog Mater Sci 2007;52:1263. [44] Mai YW, Lawn BR. J Am Ceram Soc 1987;70:289.
[4] Meyers MA, Chen PY, Lin AYM, Seki Y. Prog Mater Sci 2008;53:1. [45] Padture NP, Lawn BR. J Am Ceram Soc 1994;77:2518.
[5] Aizenberg J, Weaver JC, Thanawala MS, Sundar VC, Morse DE, [46] Swanson PL, Fairbanks CJ, Lawn BR, Mai YW, Hockey BJ. J Am
Fratzl P. Science 2005;309:275. Ceram Soc 1987;70:279.
[6] Barthelat F, Espinosa HD. Exp Mech 2007;47:311. [47] Ajayan PM, Tour JM. Nature 2007;447:1066.
[7] Nalla RK, Kruzic JJ, Kinney JH, Ritchie RO. Biomaterials [48] Ritchie RO, Dzenis Y. Science 2008;320:448.
2005;26:217. [49] Kelly A. Strong solids. Oxford: Clarendon Press; 1966.
[8] Ritchie RO. Mater Sci Eng A 1988;103:15. [50] Cox B, Yang QD. Science 2006;314:1102.
[9] Launey ME, Ritchie RO. Adv Mater 2009, in press. doi:10.1002/ [51] Marshall DB, Cox BN. Ann Rev Mater Res 2008;38:425.
adma.200803322. [52] Evans AG. J Am Ceram Soc 1990;73:187.
[10] Hartgerink JD, Beniash E, Stupp SI. Science 2001;294:1684. [53] Ishikawa T, Kajii S, Matsunaga K, Hogami T, Kohtoku Y,
[11] Aizenberg J. Adv Mater 2004;16:1295. Nagasawa T. Science 1998;282:1295.
[12] Tang ZY, Kotov NA, Magonov S, Ozturk B. Nat Mater 2003;2:413. [54] Naslain R, Dugne O, Guette A, Sevely J, Brosse CR, Rocher JP, et al.
[13] Sellinger A, Weiss PM, Nguyen A, Lu YF, Assink RA, Gong WL, J Am Ceram Soc 1991;74:2482.
et al. Nature 1998;394:256. [55] Davis JB, Marshall DB, Morgan PED. J Eur Ceram Soc 2000;20:583.
[14] Podsiadlo P, Kaushik AK, Arruda EM, Waas AM, Shim BS, Xu JD, [56] Clegg WJ, Kendall K, Alford NM, Button TW, Birchall JD. Nature
et al. Science 2007;318:80. 1990;347:455.
[15] Deville S, Saiz E, Nalla RK, Tomsia AP. Science 2006;311:515. [57] Marshall DB, Ratto JJ, Lange FF. J Am Ceram Soc 1991;74:2979.
[16] Munch E, Launey ME, Alsem DH, Saiz E, Tomsia AP, Ritchie RO. [58] Saunders RA, Lekakou C, Bader MG. Compos Sci Technol
Science 2008;322:1516. 1999;59:983.
[17] Fukasawa T, Ando M, Ohji T, Kanzaki S. J Am Ceram Soc [59] Hwu KL, Derby B. Acta Mater 1999;47:545.
2001;84:230. [60] Mekky W, Nicholson PS. Compos Pt B—Eng 2007;38:35.