See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225952661

Calculations of theoretical strength: State of the art and history

Article  in  Journal of Computer-Aided Materials Design · March 2004

DOI: 10.1007/s10820-004-4567-2

CITATIONS READS

97 1,802

4 authors, including:

Jaroslav Pokluda Pavel Šandera

Brno University of Technology Brno University of Technology
183 PUBLICATIONS   1,613 CITATIONS    54 PUBLICATIONS   620 CITATIONS   

SEE PROFILE SEE PROFILE

Mojmir Sob
Masaryk University
253 PUBLICATIONS   4,377 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Antiphase boundaries (APBs) in magnetic materials View project

I do not rember now the precise title of the project, I can complete it tomorrow View project

All content following this page was uploaded by Mojmir Sob on 09 December 2018.

The user has requested enhancement of the downloaded file.

 Journal of Computer-Aided Materials Design, 11: 1–28, 2004. © Springer 2005
DOI 10.1007/s10820-004-4567-2

Calculations of theoretical strength: State of the art and history

J. POKLUDAa,∗ , M. ČERNÝa , P. ŠANDERAa and M. ŠOBb,c

a
Institute of Engineering Physics, Brno University of Technology, Technická 2, CZ-61669 Brno, Czech
Republic; b Department of Theoretical and Physical Chemistry, Faculty of Science, Masaryk University,
CZ-61137 Brno, Czech Republic; c Institute of Physics of Materials, Czech Academy of Sciences,
Žižkova 22, CZ-61669 Brno, Czech Republic

Received 21 June 2004; Accepted 28 June 2004

Abstract. Current state and historical evolution of theoretical strength calculations is presented as a
brief overview completed by a database of selected theoretical and experimental results. Principles of
a sophisticated analysis of mechanical stability of crystals are elucidated by means of a schematic
example. Stability conditions and Jacobian matrixes are presented for selected crystalline symmetries
and deformation paths. The importance of this analysis for understanding micromechanics of fracture
is shown against the background of the influence of crystal defects. Differences between theoretical
and experimental theoretical strength (TS) values are discussed and some challenging tasks are out-
lined for the near future.
Keywords: elastic stability, theoretical strength

1. Introduction

The strength of any solid has an upper limit called the theoretical (ideal) strength
(TS). This value corresponds to the failure of an infinite perfect single crystal loaded
in a defined mode. The strength of engineering materials is usually controlled by
nucleation and motion of dislocations or microcracks. If such defects were not pres-
ent, the material will only fail when the TS is reached. Until recently loads of this
magnitude were approached in studies of the mechanical behavior of whiskers of very
pure metals and semiconductors [1–3]. Starting from the beginning of the last cen-
tury, there is a more or less continuous effort expended in order to obtain theoretical
and experimental data concerning TS of various solids. The TS values set an upper
limit to the envelope of attainable stresses and its knowledge enables us to assess the
gap remaining to upper strength values of advanced engineering materials in each
period of time. However, this is by far not the only reason for the TS investigation.
From the theoretical point of view, the TS plays a decisive role in the fundamental
theory of fracture. For example, the stress necessary for nucleation of dislocation
loop can be identified with the shear TS value and the local stress for nucleation
of a cleavage crack should overcome the tensile TS value [4–7]. The ratio of these
values expresses a tendency of the crystal matrix to become brittle or ductile [8–
10]. The TS values may be also used in the construction or checking of semi-empiri-
cal interatomic potentials. From the practical point of view, the shear TS appears to
control both the onset of fracture and the dislocation nucleation in defect-free thin
∗
To whom correspondence should be addressed.
2 J. Pokluda et al.

films and, in particular, in nano-structured materials that are currently being devel-
oped. This has been confirmed most eloquently by nanoindentation experiments (see
e.g. Refs. 11–14) which suggest that the onset of yielding at the nanoscale is con-
trolled by homogeneous nucleation of dislocations in a small, dislocation free, volume
under the nanoindenter where stresses approach the TS. The perfect single crystal-
line wires (whiskers) are used as reinforcements in advanced composite materials and
large metallic and ceramic single crystals start to be important in special engineering
components, e.g. in turbine blades [15].
The aim of this paper is to give a brief picture of both the historical development
and, particularly, the current state in the field of TS investigation. At first, some basic
theoretical considerations are introduced defining generally the present state of the
art in TS calculations at the zero Kelvin temperature (T = 0 K), followed by a brief
reminiscence on the historical evolution of TS calculation methods. The influence of
lattice defects and temperature is mentioned in connection with the problem of exper-
imental investigation of TS. Theoretical data on TS of pure elements and of selected
compounds are compared to available experimental values. Finally, some suggestions
are outlined concerning the future of TS calculations.

2. Remarks to the current state of the theoretical strength analysis

When considering a perfect crystal deforming homogeneously in a nonlinear elastic

way [16], the total internal energy of the crystal can be expanded as

1


E = E0 + V σi ηi + V Cij ηi ηj + O(η3 ), i, j = 1, 2, . . . , 6, (1)
i
2 i j

where σi are components of the stress tensor, ηi are components of the Lagrangian
strain tensor (see Appendix A), Cij are the second order elastic moduli and V is the
crystal volume. Here, the simple Voigt notation can be used since the tensors σij , ηij
and Cij kl (i, j, k, l = 1, 2, 3) are symmetric with respect to (ij ) ↔ (j i) interchange. By
setting E0 =0 and using the internal energy per unit volume Eu = E/V , one can write
∂ 2 Eu
Cij = , (2)
∂ηi ∂ηj
∂Eu
σi = . (3)
∂ηi
As it will be discussed further, the energy vs. strain dependence can be determined
using empirical inter-atomic potentials or calculating fully the electronic structure and
related cohesion of the crystal. For every particular loading, the stress state is charac-
terized by six stress tensor components. Consequently, an infinite number of “theoret-
ical strengths” exists for a given crystal. For practical reasons the TS was evaluated
only for several special cases of loading each defined by a single value of the stress
tensor component. Specifically, the uniaxial tensions and compressions along various
crystallographic directions, for the isotropic hydrostatic tension and for pure shear in
certain planes and directions. The respective TS values denoted here as σiut , σiuc , σiht ,
σihc and τis (u denoting the uniaxial, h the hydrostatic, and s the shear deformation,
respectively, t stands for tension, c for compression) cover, to a reasonable degree,
 Calculations of theoretical strength 3

the most important cases occurring in engineering practice. When no other instability
occurs in the material, such as unstable phonon modes, phase transformations, elastic
instabilities etc. the TS corresponds to the first point of inflexion on the energy vs.
strain curve. For a quasi-stationary stressed system, a general stability condition leads
to the requirement that the free energy (and at T = 0 also the total internal energy)
be minimum in subsequent constant stress ensembles in accordance with the second
law of thermodynamics [17–21]. If the solid is strained infinitesimally from the refer-
ence state associated with the stress σij (in the standard notation) by a strain tensor
εij , the related Cauchy (true) stress tij can be written as
tij = σij + Bij kl εkl ,
where
 
Bij kl = Cij kl + 21 δik σj l + δj k σil + δil σj k + δj l σik − 2δkl σij (4)
is the elastic stiffness matrix (i, j, k.l = 1, 2, 3), generally asymmetric with respect to
the (ij ) ↔ (j i) interchange. Construction of this matrix is crucial for stability assess-
ment. As was shown by Yip et al. [19], the system becomes unstable once its sym-
metrized counterpart
 
A = 21 B T + B (5)
attains a zero determinant, i.e.
det |A| = 0 (6)
during the loading. It should be emphasized that the elastic moduli in Eq. (4) are
the local ones, i.e. corresponding to different points on the deformation path. Thus,
in order to assess the stability, their values should be determined by introducing a
sufficient number of independent small deviations (strain increments) away from the
original deformation path at each point, in accordance with the symmetry of the par-
ticular crystal lattice. The solution of Eq. (6) yields a different number of possible
stability conditions for different crystal lattice symmetries as well as different load-
ing modes. The smallest possible number of necessary stability conditions (only two)
corresponds to the isotropic solid.
The stability conditions for cubic crystals loaded in uniaxial tension or compres-
sion along the [0 0 1] direction – the so-called Bain’s path – can serve as a suitable
example (see Figure 1). The tetragonal symmetry induced by the uniaxial loading
means C11 = C22 = 0, C33 = 0, C12 = 0, C13 = C23 = 0, C44 = C55 = 0, C66 = 0, and Cij = 0,
others and the simple relation σij = σ δi3 δj 3 holds for the stress tensor. By introducing
these relations to Eq. (6), one obtains the following stability conditions:
(C33 + σ )(C12 + C11 ) − 2(C13 − σ/2)2 > 0, (7a)
C11 − C12 > 0, (7b)
2C44 + σ > 0, (7c)
C66 > 0. (7d)
The left-hand side of the first condition differs from the tetragonal E[001] modulus for
the stress-free state σ = 0 only by a multiplication constant. Therefore, the violation
4 J. Pokluda et al.

[001]

[100]
a

Figure
√ 1. The fcc structure which is obtained at the Bain’s path (starting with the bcc cell) for c/a =
2.

of that condition must be closely related to the first inflexion point on the energy
vs. strain curve along the [0 0 1] deformation path. In any case, the maximum value
of the stress determining the TS along the [0 0 1] path is that at the point of inflex-
ion. This is the main reason why the testing of the criterion (7a) can actually be
omitted. All the other conditions prevent the crystal from shear instabilities. Breaking
the second condition (7b) causes a shear bifurcation from the tetragonal deformation
path to the orthorhombic one [22, 23]. In case of the fcc crystal, the second instabil-
ity induces branching to the tetragonal face centered orthorhombic path – it is the
so-called Born’s instability. The instabilities (7c) and (7d) are related to C44 and C66
shear moduli, respectively.
In order to test the shear-related criteria, special local Lagrangian deformations
(determined by corresponding Jacobian matrixes) are to be applied to the crystal at
each point of the deformation path. For the tetragonal path [24], the Jacobian
√ 
 1 + 2e 0 0 
 √ 

JC  =  0 1 − 2e 0 
 0 0 1

and the corresponding deformation matrix

 
e 0 0
 
ηC  =  0 −e 0 
0 0 0

lead to the following change of the system energy (per volume unit) according to Eq. (1):

Eu = (C11 − C12 )e2 + · · ·

Then, the tetragonal shear modulus

1 1 ∂ 2 Eu
C = (C11 − C12 ) = (8)
2 4 ∂e2
 Calculations of theoretical strength 5

enables us to test the second stability condition. Using

 
1 0√ 0 
 √ 
 
 0 1 + 2e + 1 − 2e √ 2e
√ 

JC44 =  2 + + − ,
√ 1 2e √1 2e 
 2e 1 + 2e + 1 − 2e 
0 √ √
 
1 + 2e + 1 − 2e 2
 
0 0 0
 
ηC44 =  0 0 e 
0 e 0

we obtain

Eu = 2C44 e2 + · · · and
1 ∂ 2 Eu
C44 = (9)
4 ∂e2
for testing the third stability condition. Finally,
√ √ 
 1 + 2e + 1 − 2e 2e 
 √ √ 0 
 2 + + − 
 √ 1 2e √1 2e 

JC66 =  2e 1 + 2e + 1 − 2e  ,
√ √ 0
 1 + 2e + 1 − 2e 2 
 0 0 1

 
0 e 0
 
ηC66 =  e 0 0 
0 0 0

lead to Eu = 2C66 e2 + · · · and C66 = (1/4)(∂ 2 Eu /∂e2 ) enabling to test the validity of
the last criterion.
A more extended analysis of elastic stability conditions is presented in Appendix
B for isotropic solid as well as for cubic and tetragonal crystals under various load-
ing conditions. It should be noted, however, that the stability conditions may attain
somewhat different form when using other definitions of elastic constants [19, 20], as
shown in Appendix C.
In general, two basic cases of instability behavior related to the TS can be distin-
guished when analyzing the crystal deformation
(i) Instability occurs along the original deformation path
(ii) Instability changes the loading mode or the type of the deformation path.
The first kind of instability means that the process of unstable crystal collapse
starts at the point of inflexion on the same (original) deformation path. Assuming
the constant stress ensembles (i.e. the stress-controlled loading), the crystal starts
to spontaneously disintegrate after reaching this point. During this process, how-
ever, strain induced phase transformations (so-called displacive transformations) may
appear along the deformation path [25–27]. These transformations proceed by means
6 J. Pokluda et al.

of cooperative displacements of atoms away from their lattice sides and alter the crys-
tal symmetry without changing the atomic order or composition. They are of the
first order and, therefore, accompanied by a symmetry-dictated extrema on the stress-
strain curve. The tetragonal Bain’s path also induces a typical displacive transforma-
tion. Starting with the bcc cubic structure (considered as a tetragonal one with a
ratio c/a = 1), we arrive at body-centered tetragonal structures
√ (c/a > 1) with a lower
symmetry. There is, however, one exception – for c/a = 2 the structure becomes
fcc (see Figure 1) and its higher symmetry dictates a maximum on the energy-strain
curve. Moreover, additional extrema may occur that are not dictated by the symme-
try and reflect properties of the specific material [28]. Consequently, more “TS val-
ues” can be found related to different points of inflexion on the energy–strain curve.
However, the TS is determined by the stress associated with the first point of inflex-
ion on the original strain–energy curve, which corresponds to the maximal energy gra-
dient. All other inflexion points are preceded by a break down of at least one stability
condition. Note that atomic configurations related to energy minima behind the first
point of inflexion may mimic stable or metastable atomic arrangements that could
be encountered when investigating thin films or extended defects such as interfaces
or dislocations [29]. Such configurations can also be reached during the strain-con-
trolled deformation path of a crystal (the constant strain ensembles), provided that
they are not preceded by any instabilities of the second kind.
The second-kind instability occurs before reaching the first point on the original
energy vs. strain curve. It changes the type of the deformation path (e.g. the bifurca-
tion from the tetragonal to the trigonal tensile path) or it may simply cause a change
in the loading mode (e.g. from the tensile to the shear or vice versa). In that case, the
TS corresponds to the stress related to the point, where the first shear-related stability
condition breaks down. Note that this value of the stress is smaller than that at the
point of inflexion, though it still corresponds to the maximum reachable energy gra-
dient.
A schematic energy vs. strain curve, including all the instabilities described above,
is shown in Figure 2(a) and (b) together with the related stress vs. strain curve. It
corresponds to the uniaxial tension and compression paths of a hypothetical crystal.
The left-hand side values of associated four stability conditions SC1, SC2, SC3 and
SC4 are plotted in Figure 2(c).
No instabilities occur before the first point of inflexion I1 on the tensile part of
the loading path as can be simply deduced from the positive values of all the “stabil-
ity curves” within this range (see Figure 2(c)). The SC1 curve intersects the abscissa
at the coordinate closely behind the I1 point. Hence, the stress related to I1 is the
maximal one and determines the σiut value. Extrema either dictated by occurrence of
higher-symmetry structures (denoted as PT1 and PT2) or without symmetry reasons
(WS1 and WS2) appear further on the original tensile deformation path. However,
the PT1 crystal configuration is closely preceded by the negative values of both the
SC2 and SC3 stability curves in Figure 2(c). Therefore, it cannot be reached during
the deformation of the crystal. The WS1 and WS3 extrema are enforced either by
PT2, WS2 and PT3 extrema, respectively, or by the fact that the energy curve must
increase when the distance between atoms approaches infinity or zero.
The SC4 shear instability in the compression region precedes the first point of
inflexion I6. Therefore, the value σiuc is determined by the stress related to the point
 Calculations of theoretical strength 7
(a) energy

WS1
PT1
I5
I2
I3
I1
PT2 I4
PT3 WS2
I7
I6

WS3 0 strain

(b) stress
iut

0 strain
iuc

(c) SC1
stability
conditions
SC2
SC3

strain
0
SC4

Figure 2. Schematic curves for a virtual crystal under uniaxial tension-compression. (a) Energy–strain,
(b) stress–strain, (c) stability conditions.

where the SC4 curve intersects the abscissa. Beyond this point, the crystal becomes
unstable and the deformation path is changing to a shear one. No metastable TSs
occur on the whole tension-compression deformation path since the inflexion points
I3, I5 and I6 lie already within the unstable region. The I2, I4 and I7 points are
connected with opposite slopes of the strain vs. energy curve with respect to the
8 J. Pokluda et al.

tension and compression loadings and, therefore, they could not be related to the TS
values.
In addition to the violation of the conditions of the mechanical stability, some
phonon instabilities may occur along the deformation path. A straightforward
coupling of elastic instabilities and long wave-lengthy phonons can be established
owing to the connection between elastic constants and the slope of acoustic pho-
non branches at small wave vectors. Indeed, a typical soft mode phonon was found
to be directly associated with the vanishing of the C44 modulus under uniaxial ten-
sion of the α-Fe lattice cell [30]. This implies the occurrence of a stationary wave
over the entire cell. Another soft mode phonon was detected in the short wave length
region implying atom shuffling from bcc to hcp structure according to the Burgers
transition mechanism. Very recently, pressure-induced phonon instability at low tem-
perature was detected in the copper chloride, leading to a shear-controlled phase
transformation [31], and the phonon instabilities were also predicted in Al under
uniaxial tensile and shear loading [32]. Pressure-induced phonon instabilities at low
temperature in copper chloride were studied using density functional linear-response
theory [31]. At 2.6 GPa, a soft transverse-acoustic-phonon mode was found to ini-
tiate the phase transition from the zinc-blende structure CuCl-II to a cubic struc-
ture CuCl-IV. This phase transformation is assumed to be driven by screening of
non-central long-range forces needed to stabilize the crystal against short-wavelength
shear distortions. Therefore, a possible coupling with a violation of some shear sta-
bility condition can be deduced in spite of a lack of the elastic stability analysis in
this paper.
However, other phonon instabilities cannot be linked with the above mentioned
mechanical instability. Clatterbuck et al. [32] performed ab initio calculations of the
phonon spectra as a function of strain for uniaxial tension along the [0 0 1], [1 1 0]
and [1 1 1] directions as well as for relaxed 1 1 2{1 1 1} shear. They found that in all
four cases, phonon instabilities determine the theoretical strength of Al. As already
mentioned, the first instability (violation of some elastic stability condition, soft pho-
non modes, magnetic spin arrangement, etc. [20, 22, 24, 29, 33–36]) occurring prior to
reaching the inflexion point, determines the theoretical strength. In principle, only the
analysis of the phonon spectrum of a strained crystal at each point of the defor-
mation path would be necessary and sufficient condition for ascertaining the sta-
bility of the investigated material. Such analysis is, however, extremely demanding
and, to our knowledge, it has been done only in the study of Clatterbuck et al.
[32].
Thus, the currently used methodology for calculating the TS of the particular crys-
tal can be summarized into the following points:
1. Construction of a suitable empirical interatomic potential or calculation of the
electronic structure.
2. Calculation of the strain–energy curve and the related stress vs. strain dependence
for a specific deformation path.
3. Establishment of elastic and phonon instability ranges on the strain vs. energy
and/or stress vs. strain curves.
4. Determination of TS value as a stress related to the first point of inflexion or to
the first instability point on the energy vs. strain curve.
 Calculations of theoretical strength 9

Until about 1990, however, the 3rd and, partially, the 4th point were usually omit-
ted in the TS analysis. Therefore, almost all history of TS calculation belongs to the
evolution within the frame of the first two points.

3. History of computational methods

The first well-known estimations of TS values τis and σiut were presented by Frenkel
(F) [37] and Orowan-Polanyi (O–P) [38]
Gb
τis = , (10)
2πh
 
Eγ 1/2
σiut = , (11)
a0
where E and G are elastic modules in tension and shear, b the size of the Burgers
vector, h the distance between shear planes, γ the surface energy and a0 is the lattice
constant in the tensile direction. Both expressions can be easily derived assuming the
sinusoidal stress–strain curve. In spite of a simple pair atomic interaction approach,
both relations give a plausible orientation results τis ≈ G/10 and σiut ≈ E/10 for TS
values.
The Mackenzie theory [39] can be considered to be a further milestone in the
development of TS calculations. Here the first three terms of the Fourier expan-
sion are used as the energy–strain curve and, therefore, it appears to be a consid-
erably more general approach. As found by Šandera and Pokluda [40], however, it
was not based on physically legitimate assumptions and led to very small values of
τis = G/30 for fcc crystals. Since that value seemed to be quite close to some exper-
imental results, an optimistic conclusion was accepted simultaneously with a term
change from the theoretical to the ideal strength [3, 39, 41]. More recent semiempir-
ical and ab initio approaches [40, 42] confirmed a very good validity of the classical
estimation (10). Therefore, we keep use the term “theoretical strength” as a more rel-
evant one.
During the last 30 years, many types of empirical interatomic potentials were pro-
posed, such as Johnson, Morse, Born-Mayer, Lennard-Jones, Stillinger-Weber, ionic,
polynomial, etc. Empirical parameters in these potentials have been adjusted by fit-
ting to experimental data at equilibrium, e.g., to the equilibrium lattice parameter a0 ,
elastic moduli Cij and the cohesive energy U0 . For a more extensive historical over-
view of empirical approaches to TS see, e.g. Refs. 3, 41 and 43.
In the last 25 years, semiempirical potentials constructed according to either Finnis
and Sinclair scheme (FS) [44] or embedded-atom method (EAM) [45] became very
popular. They were used for modeling behavior of materials under large deformations
and extended lattice defects in alloys. In the FS approach the total energy of the sys-
tem of N atoms is written as
 

N


 1 √
E= VSi Sj (Rij ) − ρSi  , ρSi = Si Sj (Rij ).
2 j =i j =i
i=1

Suffices i and j refer to individual atoms and suffices Si and Sj refer to species of
atoms involved. The first term is a sum of pair-potential interactions and the second
10 J. Pokluda et al.

term the many-body attractive part of the cohesive energy. The distance Rij between
atoms is limited by the cut-off radii of these potentials. This potential was con-
structed for many types of binary alloys reproducing exactly the values a0 , cij , U0 ,
the vacancy formation energies Uv as well as the stacking fault energies Usf [46]. The
EAM potentials allow exact fits to elastic moduli of the third order Cij k and yield
also reasonable phonon frequency spectra [47]. The values of a0 , Cij and Cij k charac-
terize the closest vicinity (essentially the infinitesimal neighborhood) of the unstressed
equilibrium state and the quantities U0 , Uv and Usf are of integral character. The TS
value, however, corresponds typically to the 10–20% atomic stretch from their equi-
librium (unstressed) lattice positions and its reproduction could substantially improve
the precision of semiempirical potentials.
Studies based on electronic structure calculations – so called ab initio (or first prin-
ciples) methods – started to appear since 1980. These calculations have all been per-
formed within the density functional theory [48, 49] in which the problem of many
interacting electrons is transformed into the study of the motion of a single electron
in an effective potential. This is described by the Kohn–Sham equation, which is for-
mally similar to the Schrödinger equation. In case of periodic crystalline materials,
the one-electron wave functions are expanded into variously chosen basis sets and
satisfy the Bloch theorem.
Various methods used in the ES calculations may be distinguished according to
the choice of the basis functions. The better we choose them (according to the char-
acter of the problem), the smaller number of them is needed for the description
of one-electron wave functions. Commonly used bases are augmented (APW) and
orthogonalized (OPW) plane waves, linear muffin-tin orbitals (LMTO), linear com-
bination of atomic orbitals (LCAO), Gaussian (LCGO) and augmented Slater-type
(LASTO) orbitals, augmented spherical waves (ASW), etc. The Korringa–Kohn–Ro-
stoker (KKR) method proceeds by the use of the Green function of the Kohn-Sham
equation and is also called Green’s function (GF) method. The pseudopotential
approach applied mostly to solids containing no d- or f-electrons is also widely used.
A detailed description of these methods may be found in many books and articles
(e.g., in Refs. 50–53).
After choosing an appropriate basis, the Kohn–Sham equation is solved iteratively
in order to attain self-consistency, i.e. the electron density, determined from the effec-
tive one-electron potential, must generate the same effective potential (which is again
a functional of the electron density). The quality and speed of the convergence of
such calculations is related not only to the choice of a suitable basis, but also to
the sophistication of the iterative process. As a plausible input, atomic-like poten-
tials are usually employed and input and output potentials are appropriately mixed
before starting a new iteration. The atomic configurations corresponding to deformed
structures usually have lower symmetries and, at the strength limit, they are very
far from the lowest-energy equilibrium state. Therefore, to obtain reliable structural
energy differences, the full-potential methods (i.e. without any shape approximation
of the crystal potential and electronic charge density) have to be utilized in such stud-
ies. At present, several codes are available, e.g. WIEN, VASP, FHI, FLEUR, FPLO,
FPLMTO, ABINIT, SIESTA, etc.
The first study dealing with the ideal tensile strength from the first principles was
probably that of Esposito et al. [54] who investigated tensile deformation of Cu.
 Calculations of theoretical strength 11

However, these authors did not perform relaxations of the dimensions of the loaded
crystal in the directions perpendicular to the loading axis (Poisson type expansion or
contraction). Probably the first ab initio simulation of a tensile test, including the
relaxation in perpendicular directions to the loading axis, was performed by Price
et al. [55] for uniaxial loading of TiC along the [0 0 1] axis. Later, the theoretical ten-
sile strength σiut was calculated for [0 0 1] and [1 1 1] loading axes in a variety of cubic
crystals by Šob et al. [56, 57]. Kitagawa and Ogata [58, 59] studied the tensile strength
of Al and AlN, but also did not include Poisson contraction. Further calculations of
σiut were performed for β-SiC [23], diamond [60], Si and Ge [61], Mo and Nb [35],
and for Si3 N4 [62–64].
The group of Šob et al. determined the value of σiut for iron loaded in tension
along [0 0 1]. As discussed in [36, 65, 66], no magnetic instabilities occur prior to
reaching the inflexion point in the energy vs. elongation curve for uniaxial loading
along the [0 0 1] and [1 1 1] directions as well as for the hydrostatic loading and,
therefore, the calculated values of theoretical strength of iron are not influenced by
magnetic effects.
It should be noted that most ab initio calculations of the theoretical strength ana-
lyzed only the position of the inflexion point in the dependence of the total energy on
the elongation. In some materials, another elastic stability criterion may be violated
prior to reaching the inflexion point at the energy vs. elongation curve. It was shown
that this is the case of the [0 0 1] uniaxial loading in Al [34], Nb [35] and Cu [24].
Ideal shear strength τis was first calculated by Paxton et al. [42] for V, Cr, Nb, Mo,
W, Al, Cu, Ir and later for Mo [67] and Ta [68]). These calculations did not include
any relaxation. Recently, the relaxed values of τis were calculated for TiC, TiN and
HfC [69], Mo and Nb [35], Si [70] and for Al, Cu and W [71]. Some calculations have
been done for nanowires (amorphous Si [72], MoSe nanowires [73]), grain boundaries
[74, 75], and interfaces [76, 77]. The group of M.L. Cohen and J.W. Morris, Jr. et al.
calculated τis values for Al and Cu [78, 79], for W [80, 81] and performed a thorough
theoretical analysis of the problem of strength and elastic stability [22].
Since 1997, ab initio calculations of the TS under hydrostatic tension (i.e., negative
hydrostatic pressure) have been performed by the group of Pokluda et al. [82–84] and
others (e.g. Ref. 85). As the symmetry of the structure does not change during this
deformation, simpler ab initio approaches, e.g. LMTO-ASA (Atomic Sphere Approx-
imation) may be applied.
At present, ab initio approaches are capable to yield a sufficiently precise predic-
tion of the mechanical behavior only in very simple cases but far enough from the
unstressed equilibrium states. Therefore, they may be used for calibration of semi-
empirical potentials. Starting from the early 1990s, the ab initio methods are also
combined with sophisticated semiempirical potentials to study extended defects in
materials [86].

4. Influence of lattice defects and temperature

In real perfect crystals (whiskers), the presence of some imperfections is to be

expected, at least, the equilibrium concentration of vacancies and a certain level
of surface roughness. Thus, it seems to be correct to mention the effects of lattice
defects, surface roughness and temperature on TS. From the principal reasons, the
12 J. Pokluda et al.

strength of imperfect crystals cannot be called as TS any more. In this paragraph,

therefore, the term “strength” will be used instead of TS.
In general, the presence of lattice defects decreases the strength value. The only
exception from this rule might be some special impurity dopants and alloying atoms
(see further). Since no grain (or subgrain) boundaries and secondary phase particles
are expected in perfect single crystals, only point defects, dislocations, stacking faults,
free surface, cracks and phonons (temperature) are mentioned in the following brief
summary.
(a) Point defects. Vacancies are inevitable in real crystals just near the zero Kelvin
temperature. However, the effect of monovacancies on strength is negligible. It can be
clearly seen from the fact that the simple formula (11) for σiut is equal to the Griffith
criterion for a nanocrack of the atomic size – a vacancy. Large clusters of vacancies,
which can be considered to be microcracks, may probably exist only by temperatures
near the melting point, i.e., far beyond the temperature range relevant for strength
studies.
The effect of impurities and alloying atoms on the strength was studied in several
works based on ab initio methods. Godwin et al. [87] found that Ge and As impuri-
ties in aluminum increase the cohesive energy by up to 8% but no clear conclusions
concerning the change in the strength were made. Huang et al. [88] reported that no
effect on the σiut value of silicon was found in case of p-type doping unlike the 6%
decrease in case of very high levels of n-dopants. Both works used the pseudopoten-
tial approach combined with the molecular dynamics. Song et al. [89] used the DVM
method to realize that the 7% alloying of V, Cr, Fe and Mo slightly improves the σiht
value unlike the 4% alloying causing a slight decrease of σiht . Many authors (e.g. Ref.
90) theoretically expect a slight decrease of τis owing to the impurity content. In sum-
mary, the influence of a low concentration of point defects on the strength seems to
be very small.
(b) Dislocations. Dislocation slip is possible under shear stresses above the
Peierls–Nabbarro stress. When that stress is very low as in metals, the strength might
be dramatically reduced (by 4 orders of magnitude or even more). On the other hand,
the P–N stress is extremely high in ceramic covalent crystals (C, Si, SiC, ZnS, Si3 N4 )
and complex ionic crystals (MgAl2 O4 , Al2 O3 , Al2 O3 .MgO) so that the dislocations
are practically sessile at near zero temperatures. Therefore, the tensile strength can be
reduced by, say, tens of percents owing to microcracks initiating by the stress relax-
ation around dislocations with long Burgers vectors.
(c) Stacking faults. As far as we know, no special studies regarding the effect of
stacking faults (SF) on strength were performed. Atoms on the stable SF plane lie in
the local energy minimum and no stress is induced in the surrounding volume. The
energy of the SF per atom is about a two order of magnitude lower than that of the
free surface. We may, therefore, deduce that the influence of SFs on the strength value
can be neglected.
(d) Free surface. The effect of a perfectly flat surface on the strength can be also
considered to be negligible. However, small imperfections like scratches or dimples act
as stress concentrators (notches). Their maximum effect on the strength value can be
roughly estimated by a factor 1 + 2(l/ρ)1/2 , where l is the notch depth and ρ is the
curvature of the notch root (ρ > 0) [91]. Therefore, the sharp notches can significantly
reduce the nominal strength value.
 Calculations of theoretical strength 13

(e) Cracks. According to the Griffith law for perfectly brittle materials, an atomi-
cally sharp crack of the length a causes a drop in the tensile or shear strength value
by a factor of (a/a0 )1/2 (a0 is the lattice parameter). However, only ceramics, semi-
conductors and, most probably, molybdenum and tungsten can be considered to be
intrinsically brittle crystals at the zero absolute temperature [92]. In all other metals,
the dislocation emission precedes the unstable crack growth. This process increases
the effective surface energy and blunts the crack tip. Consequently, the drop in the
strength must be much less than that predicted by the Griffith law.
(f) Phonons (temperature). The strength variation with the temperature corre-
sponds, basically, to the problem of the role of phonons in deformation and frac-
ture processes. Although the Frenkel (10) suggests simply that the variation should
be the same as that of the shear modulus, the O–P relation (11) brings the com-
plication with the temperature dependence of surface energy. In principle, two suffi-
ciently relevant methods were historically applied to predict the temperature influence
– the Einstein model of harmonic oscillators combined with the elastic instability cri-
terion and the model of dislocation nucleation supported by phonon fluctuations [3].
The first approaches predicted the drop in the uniaxial strength within the range of
several to tens percents when changing the temperature from 0 to 1000 K, depend-
ing on the type of the empirical interatomic potential (short- or long range). How-
ever, this drop was very close to the predicted change in the Young’s modulus. The
second approach deals with the maximum energy of up to 50 kT the thermal fluc-
tuations can supply at any temperature. Typically, the results show 10% decrease in
the strength within that temperature range. An analysis of possible thermal initia-
tion of a crack has shown that the relatively small amounts of available thermal
energy are unlikely to affect the process very much up to 1200 K (in comparison with
the total amount of the strain energy conversion into the surface energy). It should
be emphasized, however, that the most physically justified approach to the tempera-
ture problem would be a complex analysis of phonon spectra. Nevertheless, one can
assume that the temperature change in the relevant elastic modulus might be consid-
ered to be an acceptable lower-band first approximation to the strength temperature
dependence.

5. Theoretical and experimental results

Owing to the long history of TS calculations, a number of theoretical results con-

cerning crystals of pure elements and compounds is available in the literature. On the
other hand, the experimental data are rather limited, which is related to the problems
associated with both the specimen preparation and the experimental arrangement. It
is particularly difficult to measure the TS for pure shear τis and hydrostatic tension
σiht – in the latter case, in fact, no measurements have been performed.
Theoretical and experimental TS values for various crystals are presented in
Appendix D. The selection was made in order to show all kinds of crystals –
metals, semiconductors, intermetalics, inert gases, ceramics (ionic and covalent) and
polymers. There is a three-order difference between the strongest and the weakest
crystal. As expected, the highest TS values belong to the diamond crystal with pure
covalent bonds and the lowest ones to van der Waals crystals of inert gases stable
only in the low temperature range. The lowest values of τis for a particular crystal
14 J. Pokluda et al.

have always been found in calculations with full relaxations that take into account
the elastic stability conditions. In case of fcc crystals, the shear plane is mostly
{1 1 1} unlike the {1 1 2} for bcc crystals. Similarly, the lowest σiut values correspond
to ab initio calculations taking into account the elastic stability conditions. It can
be seen that theoretical and experimental values can differ substantially according to
the applied computational or experimental procedure. Therefore, it is difficult to per-
form a general comparison of theoretical approaches. It should be emphasized that
a great majority of σiut values corresponds to the inflexion point on the energy vs.
strain curve when omitting the stability analysis. Hence, most of the calculated val-
ues are probably overestimated. On the other hand, the stability analysis applied to
cubic crystals under hydrostatic tension has revealed that, in most cases, no shear
instabilities appeared before reaching the inflexion point [93]. Therefore, values σiht
obtained by ab initio methods can be considered to be very reasonable estimations
of TS. A comparison between ab initio σiht values and those obtained by empirical
approaches is displayed in Figure 3 for selected crystals. It is clear that, in most cases,
the sinusoidal potential overestimates the TS whereas the Morse approximation yields
underestimations (Figure 3). Results well comparable with the ab initio calculations
using the LDA can be obtained by employing the polynomial potential, as can be
seen from Figure 3(a). Application of the GGA, however, indicates an overestimation
of the polynomial approach (Figure 3(b)).
Most experimental data obtained on perfect large single crystals or even whis-
kers loaded in uniaxial tension are an order of magnitude lower than calculated
TS values. This might be due to the dislocation-assisted shear instability control-
ling the final failure process. The very high Peierls–Nabarro stress in covalent and
complex ionic ceramic crystals resists to nucleation and motion of dislocations. For
such crystals, indeed, the difference between the theory and the experiment is rela-
tively small (see Appendix D). The probably highest ever reported experimental value
σiut = 40 GPa ≈ E/20 corresponds to the ZnO whisker [41]. However, experiments
performed on tungsten whiskers by Mikhailovskii et al. [94] approach closely the cal-
culated results [56].
The discrepancy between the theory and the experiment for metallic crystals
becomes much smaller when the elastic- and phonon-based stability conditions are
taken into account (e.g. for Cu Ref. 24 and for Al Ref. 32 crystals), though it still
remains significant. Values of the maximum resolved shear stress τmax ≈ G/50 at frac-
ture [41] are substantially lower than the theoretical τis ≈ G/10, but already of the
same order. Let us recall that the lowest ever estimated theoretical value τis ≈ G/30
for fcc lattice and {1 1 1}1 1 2 shear is not based on physically legitimate assump-
tions. Therefore, the difference between the theory and the experiment can be under-
stood particularly in terms of various imperfections of the experimental procedure,
including the above mentioned crystal defects. It should be noted, for example, that
the stability conditions imply the loading mechanism to be able to maintain the
Cauchy stress and the crystal symmetry along the whole deformation path. It means
particularly that the shear instability {1 1 1}1 1 2  in case of [0 0 1] tension of fcc crys-
tal is not controlled by the stability conditions (7) although it is a usually observed
failure mode in experiments. Namely this instability requires a finite shear in the
{1 1 1} plane changing the symmetry to triclinic or monoclinic. Additionally, there is
the resonance of short-wavelength phonons. All such effects are clearly beyond the
 Calculations of theoretical strength 15
2.4
(a)
2.2
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
sinus
0.4
polynom
0.2 Morse
0.0
Li Na Al K V Cu Nb Mo Ag Ba Ta W Pt Au Pb

2.4
2.2 (b)
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6 sinus
0.4 polynom
0.2 Morse
0.0
Li Na Al K V Cu Nb Mo Ag Ba Ta W Pt Au Pb

Figure 3. Comparison between theoretical strength values for isotropic tension computed by means of
empirical potentials and ab initio methods using: (a) LDA and (b) GGA.

description supplied by the mechanical stability conditions based on the continuum

mechanics.
Anyway, a lot of work is needed in both the theoretical and, predominantly, in the
experimental investigation of TS. Let us note, finally, that the currently used ultra-
high strength steels exhibit only σiut /10 value of ultimate strength. From the theo-
retical point of view, however, materials of an extreme dislocation density could, in
principal, achieve the strength level of σiut /2 [5].

6. Conclusions

A significant advance in the solution of the theoretical strength problem was achieved
by application of ab initio electronic structure calculations. Theoretical predictions
of mechanical characteristics of perfect crystals from the first principles enable to
construct very reliable and powerful interatomic potentials which can be used for
analyzing complex problems in physics of deformation and fracture on an atomic
level. The best documentation of that statement is the success of ab initio supported
semi-empirical interatomic potentials used in recent studies on extended crystal
16 J. Pokluda et al.

defects as grain boundary structures and metastable phases. The ab initio calculated
TS values are plausible stress parameters characterizing the mechanical state of a
crystal close to its mechanical instability.
From the theoretical point of view, the following items remain the main challenges
in the near future: (i) ab initio supported semiempirical potentials; (ii) extended appli-
cation and improvement of stability conditions including phonon resonance; (iii) the
first principle TS studies. On the other hand, the theory was insufficiently supported
by experimental results hitherto. In this sense, the nanoindentation experiments are
very promising at least for testing the shear TS. Additionally, there is an increas-
ing demand to use whiskers and single crystals in advanced composite materials and
components. Let us hope that the technological progress would enable to perform
not only sophisticated nanoindentation tests but also additional tensile experiments
on whiskers and, particularly, on large perfect crystals by means of (nearly) perfect
loading devices. It seems to be one of the crucial points not only in the evolution of
TS investigations but, generally, in the physics of deformation and fracture.

Acknowledgements

This research was supported by the Ministry of Education, Youth and Sport of the
Czech Republic within the Research Project MSM 262100002, by the Grant Agency
of the Academy of Sciences of the Czech Republic (Project No. IAA1041302), by the
Grant Agency of the Czech Republic (Project No. 202/03/1351), and by the Research
Project Z2041904 of the Academy of Sciences of the Czech Republic. A part of this
study has been performed in the framework of the COST Project No. OC 523.90.
The use of the computer facilities at the MetaCenter of the Masaryk University,
Brno is acknowledged.

References

1. Brenner, S.S., J. Appl. Phys., 27 (1956) 1484.

2. Nadgorny, E.M., Sov. Phys. Uspekhi, 5 (1962) 462.
3. Kelly, A. and Macmillan N.H., Strong Solids, Oxford: Clarendon Press, 1986.
4. Jokl, M.L., Vitek V. and McMahon, Jr. C.J., Acta Metall., 28 (1980) 147.
5. Kroupa, F., In Pokluda, J. and Staněk, P. (Eds.) Prediction of Mechanical Properties of Metal-
lic Materials by Means of Structural Characteristics. VÚ070 Brno, 1981, p. 32.
6. Thomson, R., Solid State Phys., 39 (1986) 1.
7. Huang, H. and Gerberich, W.W., Acta Metall. Mat., 40 (1992) 2873.
8. Kelly, A., Tyson, W.R. and Cottrell, A.H., Philos. Mag., 15 (1967) 567.
9. Pokluda, J. and Šandera, P., Phys. Stat. Solidi., 167b (1991) 543.
10. Pokluda, J. and Šandera, P., Key Eng. Mater., 97–98 (1994) 467.
11. Bahr, D.F., Kramer, D.E. and Gerberich W.W., Acta Mater., 46 (1998) 3605.
12. Woodcock, C.L. and Bahr, D.F., Scripta Mater., 43 (2000) 783.
13. Gouldstone, A., Koh, H.J., Zeng, K.Y., Giannakopoulos, A.E. and Suresh, S., Acta Mater., 48
(2000) 2277.
14. delaFuente, O.R., Zimmerman, J.A., Gonzales, M.A., delaFiguera, J., Hamilton, J.C., Pai, W.W.
and Rojo, J.M., Phys. Rev. Lett., 88 (2002) 036101.
15. Goldschmidt, D., In Coutsouradis, D. et al. (Eds.) Materials for Advanced Power Engineering
1994, Part I. Dordrecht, Kluwer Academic Publishers, 1994, p. 661.
16. Wallace, D.C. (Ed.) Thermodynamics of Crystals. New York: John Wiley & Sons, 1972.
 Calculations of theoretical strength 17

17. Wang, J., Li, J., Yip, S., Phillpot, S.R. and Wolf, D., Phys. Rev., B52 (1995) 627.
18. Zhou, Z. and Joós, B., Phys. Rev., B54 (1996) 3841.
19. Yip, S., Li, J., Tang, M. and Wang, J., Mater. Sci. Eng., A317 (2001) 236.
20. Hill, R. and Milstein, F., Phys. Rev., B15 (1977) 3087.
21. Milstein F. and Rasky, D.J., Phys. Rev., B54 (1996) 7016.
22. Morris, J.W. and Krenn, C.R., Philos. Mag., A80 (2000) 2827.
23. Li, W. and Wang, T., Phys. Rev., B59 (1999) 3993.
24. Černý, M., Šob, M., Pokluda, J. and Šandera, P., J. Phys. Cond. Mater., 16 (2004) 1045.
25. Craievich, P.J., Weinert, M., Sanchez, J.M. and Watson, R.E., Phys. Rev. Lett., 72 (1994) 3076.
26. Šob, M., Wang, L.G. and Vitek, V., Comp. Mater. Sci., 8 (1997) 100.
27. Paidar, V., Wang, L.G., Šob, M. and Vitek, V., Model. Simul. Mat. Sci., 7 (1999) 369.
28. Šob, M., Friák, M., Wang, L.G. and Vitek V., In Bulatov, V.V., Diaz de la Rubia, T.,
Phillips, R., Kaxiras, E. and Ghoniem, N. (Eds.) Multiscale Modelling of Materials, Mat. Res.
Soc. Symp. Proc, Vol. 538, Warrendale, PA: Materials Reseach Society, 1999, pp. 523–527.
29. Šob, M., Friák, M., Legut, D. and Vitek V., In: 3rd Int. Alloy Conf. An Interdisciplinary
Approach to the Science of Alloys in Metals, Minerals, eds. P.E.A. Turchi and A. Gonis,
Estoril/Cascais, Portugal, June 30–July 5, 2002 (accepted in Mater. Sci. Eng. A).
30. Shibutani, Y., Krasko, G.L., Šob, M. and Yip, S., Mater. Sci. Res. Int., 5 (1999) 225.
31. Ma, Y., Tse, J. and Sand Klug, D.D., Phys. Rev., B67 (2003) 140301.
32. Clatterbuck, D.M., Krenn, C.R., Cohen, M.L. and Morris, Jr. J.W., Phys. Rev. Lett., 91 (2003)
135501.
33. van Vliet, K.J., Li, J., Zhu, T., Yip, S. and Suresh, S., Phys. Rev., B67 (2003) 104105.
34. Li, W. and Wang, T., J. Phys. Cond. Mat., 10 (1998) 9889.
35. Luo, W., Roundy, D., Cohen, M.L. and Morris, Jr., J.W., Phys. Rev., B66 (2002) 094110.
36. Clatterbuck, D.M., Chrzan, D.C. and Morris, J.W. Jr., Acta Mater., 51 (2003) 2271.
37. Frenkel J., Z. Phys., 37 (1926) 572.
38. Orowan E., Rep. Progr. Phys., 12 (1949) 185.
39. Kelly A., Strong Solids, Oxford: Clarendon Press, 1973.
40. Šandera, P. and Pokluda, J., Scr. Metall. Mat., 29 (1993) 1445.
41. Macmillan, N.H., In Latanision, R.M. and Pickens, J.R. (Eds.) “Atomistics of Fracture”,
Plenum Press, New York, 1983, p. 95.
42. Paxton, A.T., Gumbsch, P. and Methfessel, M., Philos. Mag. Lett., 63 (1991) 267.
43. Šandera, P. and Pokluda, J., Metall. Mat., 32 (1994) 180.
44. Finnis, M.W. and Sinclair, J.E., Philos. Mag., A50 (1984) 45.
45. Daw, M.S. and Baskes, M.I., Phys. Rev., B29 (1984) 6443.
46. Luzzi, D.E., Yan, M., Šob, M. and Vitek, V., Phys. Rev. Lett., 67 (1991) 1894.
47. Chantasiriwan, S. and Milstein, F., Phys. Rev., B67 (1996) 14081.
48. Hohenberg, P. and Kohn, W., Phys. Rev., 136 (1964) B864.
49. Kohn, W., Sham, L.J. and Phys. Rev., 140 (1965) A1133.
50. Singh, D.J., Planewaves, Pseudopotentials and the LAPW Method, Boston, Dordrecht, London:
Kluwer Academic Publishers, 1994.
51. Blaha, P., Schwarz, K. and Luitz, J., A Full Potential Linearized Augmented Plane Wave Pack-
age for Calculating Crystal Properties, Vienna: Vienna University of Technology, 1997.
52. Turek, I., Drchal, V., Kudrnovský, J., Šob, M. and Weinberger P., Electronic Structure of Dis-
ordered Alloys, Surfaces and Interfaces, Boston: Kluwer Academic Publishers, 1997.
53. Springborg, M., Methods of Electronic-Structure Calculations, Chichester: Wiley, 2000.
54. Esposito, E., Carlson, A.E., Ling, B.D., Ehrenreich, H. and Gelatt, Jr., C.D., Philos. Mag., A41
(1980) 251.
55. Price, D.L., Cooper, B.R. and Wills, J.M., Phys. Rev., B46 (1992) 11368.
56. Šob, M., Wang, L.G. and Vitek, V., Mater. Sci. Eng., A234–236 (1997) 1075.
57. Šob, M., Wang, L.G. and Vitek, V, Metall. Mat., 36 (1998) 145.
58. Kitagawa, H. and Ogata, S., Key Eng. Mat., 161–163 (1999) 443.
59. Ogata, S. and Kitagawa, H., Comput. Mat. Sci., 15 (1999) 435.
60. Telling, R.H., Pickard, C.J., Payne, M.C. and Field, J.E., Phys. Rev. Lett., 84 (2000) 5160.
61. Roundy, D. and Cohen, M.L., Phys. Rev., B64 (2001) 212103.
18 J. Pokluda et al.

62. Ogata, S., Hirosaki, N., Kocer, C. and Kitagawa, H., Phys. Rev., B64 (2001) 172102.
63. Ogata, S., Hirosaki, N., Kocer, C. and Shibutani, Y., J. Mat. Res., 18 (2003) 1168.
64. Kocer, C. Hirosaki, N. and Ogata, S., Phys. Rev., B67 (2003) 035210.
65. Friák, M., Šob, M. and Vitek, V., In: Juniormat’01, Institute of Materials Engineering, Brno:
Brno University of Technology, 2001, pp. 117–120.
66. Friák, M., Šob, M. and Vitek, V., Philos. Mag., A83 (2003) 3529.
67. Xu, W. and Moriarty, J.A., Phys. Rev., B54 (1996) 6941.
68. Söderlind, P. and Moriarty, J.A., Phys. Rev., B57 (1998) 10340.
69. Jhi, S.H., Louie, S.G., Cohen, M.L. and Morris, J.W. Jr., Phys. Rev. Lett., 87 (2001) 075503.
70. Umeno, Y. and Kitamura, T., Mater. Sci. Eng., B88 (2002) 79.
71. Ogata, S., Li, J. and Yip, S., Science, 298 (2002) 807.
72. Galli, G., Gygi, F. and Catellani, A., Phys. Rev. Lett., 82 (1999) 3476.
73. Ribeiro, F.J., Roundy, D.J. and Cohen, M.L., Phys. Rev., B65 (2002) 153401.
74. Kohyama, M., Mater. Sci. Forum 294–296 (1999) 657.
75. Kitamura, T. and Umeno, Y., Model. Simul. Mat. Sci. Eng., 11 (2003) 127.
76. Batirev, I.G., Alavi, A., Finnis, M.W. and Deutsch, T., Phys. Rev. Lett., 82 (1999) 1510.
77. Tanaka, S., Yang, R. and Kohyama, M., to be published in Proc. Int. Symp. on Micro-Mech.
Eng., -Heat Transfer, Fluid Dynamics, Reliability and Mechanotronics (ISMME2003, Tsuchiura
and Tsukuba, Japan, Dec. 1–3, 2003), The Japan Soc. Mech., Tokyo, 2003, B26–053.
78. Roundy, D., Krenn, C.R., Cohen, M.L. and Morris, J.W. Jr., Phys. Rev. Lett., 82 (1999) 2713.
79. Roundy, D., Krenn, C.R., Morris, J.W. Jr. and Cohen, M.L., Mat. Sci. Eng., A317 (2001) 44.
80. Roundy, D., Krenn, C.R., Morris, J.W. Jr. and Cohen, M.L., Mat. Sci. Eng., A319–321 (2001)
111.
81. Roundy, D., Krenn, C.R., Cohen, M.L. and Morris, J.W. Jr., Philos. Mag., A81 (2001) 1725.
82. Šandera, P., Pokluda, J., Wang, L.G. and Šob, M., Mater. Sci. Eng., A234–236 (1997) 370.
83. Černý, M., Šandera, P. and Pokluda, J., Czech. J. Phys., B49 (1999) 1495.
84. Černý, M., Pokluda, J., Šandera, P. and Šob, M., Phys. Rev., B67 (2003) 035116.
85. Son, Y. and Yang, R., Phys. Rev., B59 (1999) 14220.
86. Yan, M., Šob, M., Luzzi, D.E., Vitek, V., Ackland, G.J., Methfessel, M. and Rodriguez, C.O.,
Phys. Rev., B47 (1993) 5571.
87. Goodwin, L., Needs, R. and Heine, V., Phys. Rev. Lett., 60 (1988) 2050.
88. Huang, Y.M., Spence, J.C.H. and Sankey, O.F., Philos. Mag., 70 (1994) 53.
89. Song, Y., Xu, D.S., Yang, R., Li, D., Wu, W.T. and Guo, Z.X., Mater. Sci. Eng., A260 (1999)
269.
90. Haydock, R., J. Phys., 14C (1981) 3807.
91. Neuber, H., Kerbspannungslehre, Berlin: Springer Verlag, 1958.
92. Pokluda, J., Kroupa, F. and Obdržálek, L., Mechanical Properties and Structure of Solids,
Brno: PC Dir, 1994.
93. Černý, M. and Pokluda, J., J. Alloys Comp., 378 (2004) 159.
94. Mikhailovskii, I.M., Poltinin, I.Ya. and Fedorova, L.I., Sov. Phys. Solid State, 23 (1981) 757.
95. Murnaghan, F.D., Finite Deformation of an Elastic Solid, New York: Wiley, 1951.
96. Mehl, J.M., Klein, B.M. and Papaconstantopoulos, D.A., In Westbrook, J.H. and Fleisher, R.L.
(Eds.) Intermetallic Compounds 1, New York: Wiley, 1994.
97. Fast, L., Elastic Properties and Structural Stabilities of Elements and Compounds, Uppsala:
Acta Universitatis Upsaliensis, 1998.
98. Söderlind, P., Theoretical Studies of Elastic, Thermal and Structural Properties of Solids,
Uppsala: Acta Universitatis Upsaliensis, 1994.
99. Clatterbuck, D.M., Chrzan, D.C. and Morris, Jr., J.W., Philos. Mag. Lett., 82 (2002) 141.
100. Roundy, D., Krenn, D., Cohen, M.L. and Morris, Jr., J.W., Philos. Mag., A81 (2001) 1725.
101. Verma, M.V., Verma, A. and Rathere, R.P.S., Acta Phys. Pol., A93 (1998) 479.
102. Friák, M., Šob, M. and Vitek, V., In: Schneibel, J.H., Hemker, K.J., Noebe, R.D., Hanada, S.
and Sauthoff, G. High-Temperature Ordered Intermetallic Alloys IX, Mater. Res. Soc. Symp.
Proc. Vol. 646, Warrendale, PA: Materials Reseach Society, 2001, paper N4.8.
 Calculations of theoretical strength 19

103. Černý, M., Pokluda, J. and Šandera, P., Mater. Sci. Eng. A, 387–389 (2004) 923.
104. Friák, M., Šob, M. and Vitek, V., Phys. Rev., B68 (2003) 184101.
105. Šob, M., Friák, M., Legut, D., Fiala, J. and Vitek, V., Mater. Sci. Eng. A, 387–389 (2004) 148.
106. Morris, Jr., J.W., Krenn, D., Roundy, D. and Cohen, M.L., In: Turchi, P.E. and Gonis, A.
(Eds.) 2000 Hume-Rothery Symp. in honor of A.G. Khatchaturyan, Warrendale, PA: TMS,
2000, pp. 187–207.
107. Šob, M., Wang, L.G. and Vitek, V., Philos. Mag., B78 (1998) 653.
108. Černý, M., Šandera, P. and Pokluda J., In: Šandera, P. (Ed.) MSMF3 VUTIUM, Brno Univ.
Technol., 2001, pp. 140–145.

Appendix A: Basic quantities

Let us consider a system in some reference state (e.g. in one of infinitely many config-
urations along a certain deformation path). Let us focus on one point in the system
with a position described by vector a. When the system is subjected to a deforma-
tion, this point moves by a displacement u from its original position to a new one
assigned x. So, it holds x = a + u. Such a change can be depicted by a Jacobian
matrix in the following manner: x = J· a. This Jacobian matrix has the form
 
 ∂x1 ∂x1 ∂x1 
 
 ∂a1 ∂a2 ∂a3 
 ∂x ∂x ∂x 
 2 2 
Jˆ = 
2
.
 ∂a1 ∂a2 ∂a3 
 ∂x3 ∂x3 ∂x3 
 
 
∂a1 ∂a2 ∂a3
The finite strain (sometimes called Lagrangian strain [18, 19]) matrix can then be
derived according to the relation [16, 95]
1  ˆT ˆ ˆ
η̂ = J J −I ,
2

where the upper index T means transposition and I is the identity matrix of the same
dimension as the Jacobian matrix. For the strain matrix components it follows:
   
1 ∂xk ∂xk 1 ∂ui ∂uj ∂uk ∂uk
ηij = − δij = + + .
2 ∂ai ∂aj 2 ∂aj ∂ai ∂ai ∂aj

The energy Eu can be expanded in the Fourier series as

1


Eu = σij ηij + Cij kl ηij ηkl + · · ·
ij
2 ij kl

Using the Voigt notation to reduce number of indexes, we can substitute two indexes
by the only one according to the rule 1 → 11, 2 → 22, 3 → 33, 4 → 23, 5 → 13 and
6 → 12. However, sometimes it is more convenient [96] to transform the strain matrix
in a different manner: η1 = η11 , η2 = η22 , η3 = η33 , η4 = 2η23 , η5 = 2η13 and η6 = 2η12 .
Then, the energy expansion can be written as shown in Eq. (1).
20 J. Pokluda et al.

Appendix B: Stability conditions for some particular structures and loading types

B.1. Tetragonal system under uniaxial loading

When the uniaxial stress is applied along the [0 0 1] direction, the elastic stiffness
matrix can be expressed as follows:
 
 C11 C C 0 0 0 
 12 13 
 C12 C11 C13 0 0 0 

 C13 − σ C13 − σ C33 + σ 0 0 0 
 σ

B̂ =  0 0 0 C44 + 0 0  .
 2 σ 
 
 0 0 0 0 C 44 + 0 
 2 
 0 0 0 0 0 C66 

The determinant of the related matrix A

 
det(Â) = 1/4C66 (2C44 + σ )2 (C11 − C12 ) C11 C33 + C12 C33 − 2C13
2
+ 2C13 σ + C11 σ + C12 σ − σ 2 /2

can be simply decomposed into a set of four stability conditions presented in Eqs.
(7a)–(7d).

B.2. Tetragonal system under biaxial (epitaxial) loading

There are only two non-zero components of the stress tensor. When the system is
loaded in the xy-plane, the components of the stress tensor are σ1 = σ2 = σ . The elas-
tic stiffness matrix then obtains the form
 
 C11 + σ C12 − σ C13 − σ 0 0 0 

 C12 − σ C11 + σ C13 − σ 0 0 0 

 C13 C C 0 0 0 
 13 33
σ 

B̂ =  0 0 0 C44 + 0 0  .
 2 σ 
 
 0 0 0 0 C44 + 0 
 2 
 0 0 0 0 0 C66 + σ 

The corresponding determinant is

 
det(Â) = 1/4 (C66 + σ ) (2C44 + σ )2 (C11 − C12 + 2σ ) C11 C33 + C12 C33 − 2C13
2
+ 2C13 σ − σ 2 /2

and the resulting stability conditions

C33 (C11 + C12 ) − 2 (C13 − σ/2)2 > 0,

C11 − C12 + 2σ > 0,
C44 + σ/2 > 0,
C66 + σ > 0

are somewhat different from Eq. (7) for uniaxial loading.

 Calculations of theoretical strength 21

B.3. Cubic system under triaxial isotropic (hydrostatic) loading

This is the only type of loading preserving the cubic symmetry. The stress tensor is
given by σ1 = σ2 = σ3 = σ , and the elastic stiffness matrix is as follows:
 
 C11 + σ C12 − σ C12 − σ 0 0 0 

 C12 − σ C11 + σ C12 − σ 0 0 0 

 C12 − σ C12 − σ C11 + σ 0 0 0 
B̂ =  .
 0 0 0 C44 + σ 0 0 
 0 0 0 0 C44 + σ 0 

 0 0 0 0 0 C44 + σ 

The determinant of the matrix B (that is already symmetric in this case)

det(B̂) = (C44 + σ )3 (C11 − C12 + 2σ )2 (C11 + 2C12 − σ )

can be decomposed into the following three stability conditions

C11 + 2C12 − σ > 0,

C11 − C12 + 2σ > 0,
C44 + σ > 0. (B.1)

Isotropic loading can be described using the following Jacobian and strain matrices:
√   
 1 + 2e  e 0 0
 √ 0 0   
Jizo =  0 1 + 2e √ 0 ,
 ηizo =  0 e 0  . (B.2)
 0 0 1 + 2e   0 0 e 

Then, the energy expansion (1) reduces to

1 
Eu = 3σ e + 3C11 e2 + 6C12 e2 + · · · .
2
and results in the relation

1 ∂ 2 Eu
C11 + 2C12 = . (B.3)
3 ∂e2
The left-hand side of Eq. (B.3) is proportional to the bulk modulus
B = (1/3) (C11 + 2C12 ) that is related to the first of stability conditions (B.1).

B.4. Isotropic solid

Applying the hydrostatic loading to an isotropic crystal, the number of stability con-
ditions reduces to two. It is caused by an additional relation C11 − C12 = 2C44 :

C11 + 2C12 − σ > 0,

C44 + σ > 0.
22 J. Pokluda et al.

Appendix C: Different definitions of elastic constants and related stability conditions

C.1. A case study – isotropic loading of cubic crystals

The Jacobian and strain matrices can be defined in a slightly different way than those
shown in Appendix B (Eq. (B.2)), e.g. as follows:
 
 x2 
   x + 0 0 
1+x 0   
 0   2 
 x 2 
Jiso =  0 1 + x 0  , ηiso =  0 x + 0  (C.1)
 0   2
0 1+x 
 x 2 
 0 0 x+ 
2
Here the relative extension x = (a/a0 ) − 1 is used as a strain variable. Then, the energy
expansion has the form
   2
x2 1 x2
Eu = 3σ x + + (3C11 + 6C12 ) x + +···
2 2 2
and results in the relation
 
1 ∂ 2 Eu 3
= (C11 + 2C12 ) 1 + 3x + x + σ. (C.2)
3 ∂x 2 2
One can clearly see that such a definition of elastic constants (used by some authors
[97, 98]) is different from the previous case – see Eq. (B.3).
Similarly, the Jacobian and deformation matrixes for testing the stability condition
related to the tetragonal shear can have the form
 
   x2 
1+x 0 0  x + 0 0 
   2 
   
JC =  0 1 − x 0  ,
 ηC = 
 x .
2
 0  0 −x + 0
0 1  2 
 0 0 0
The corresponding energy expansion
1
Eu = σ x 2 + (C11 − C12 ) x 2 + (C11 + C12 ) x 4 + · · ·
2
results in the relation
1 ∂ 2 Eu 3
= (C11 − C12 ) + (C11 + C12 ) x 2 + σ. (C.3)
2 ∂x 2 2
To obtain the C44 constant, the following matrices can be used:
 
  0 0 0 
1 0 0  
   x2 
 
JC44 =  0 1 x  , ηC44 =  0 2 x 
0 x 1  
 x2 
0 x 
2
 Calculations of theoretical strength 23

and the energy is expanded as

1
Eu = σ x 2 + 2C44 x 2 + (C11 + C12 ) x 4 + · · · .
4
Then, one obtains
1 ∂ 2 Eu 3
= 2C44 + (C11 + C12 ) x 2 + σ. (C.4)
2 ∂x 2 2
Now, considering x as well as e to be small deformations (from some reference
state at the volume V), we can simply compare obtained values of C11 + 2C12 using
Eqs. (C.1) and (B.2)
1 ∂ 2E 1 ∂ 2E
− σ =
3 ∂x 2 3 ∂e2
to see that such different definitions (based on different strain variables) of elas-
tic constants differ by the value of applied stress. Comparison between C11 − C12
obtained from Eq. (C.3) and Eq. (8) gives the same difference. Finally, equalizing the
C44 values using Eqs. (C4) and (9) yields:
1 ∂ 2E σ 1 ∂ 2E
− = .
4 ∂x 2 2 4 ∂e2
Therefore, if one calculates the elastic moduli using the second derivative of the
energy with respect to parameter x, a different set of stability conditions must be
employed
σ
C11 + 2C12 − 2σ > 0, C11 − C12 + σ > 0, C44 + > 0. (C.5)
2
There are, however, some other definitions of elastic constants, (e.g., Ref. [21])
∂ 2 Eu
Cij = αi αj ,
∂ai ∂aj
where a1 , a2 and a3 are three principal edges of the conventional crystallographic cell
and a4 , a5 and a6 are their included angles; αi = ai if i = 1, 2, 3 and αi = 1 if i = 4, 5, 6.
One can easily derive that the definition of, e.g. C11 and C12 is equivalent to the pre-
vious definition using the relative extension x as a strain variable. In the equilibrium
(without external load), all definitions result in identical elastic constants and stability
conditions. Under an applied load, however, the different definitions of elastic con-
stants lead to different stability conditions. In the literature, therefore, we can find
also the following stability conditions for hydrostatic loading [21] (see relations (B.1)
and (C.5)):
C11 + 2C12 − 2σ > 0,
C11 − C12 + σ > 0,
C44 + σ > 0.
Appendix D: Comparison between theoretical and experimental strength values 24
Crystal Structure Theoretical strength [GPa]
Threeaxial Uniaxial Shear
Theory Calculation method Theory Calculation method Experiment Theory Calculation method Experiment
Li bcc 1.92 AI [85] 0.319 111{110} SE [43]
3.89 AI [83]
3.13 AI [108]
2.49 SE [43]
J. Pokluda et al.

Na bcc 1.77 AI [85] 0.036 100 SE [41] 0.209 111{110} SE [40]

2.18 AI [83] 0.191 110 SE [41] 0.20 111{112} AI [106]
1.55 AI [108] 0.848 111 SE [41]
1.20 SE [43]
K bcc 1.0 AI [85] 0.112 111{110} SE [40]
1.06 AI [83]
0.701 AI [108]
0.659 SE [43]
Ge dia 18.6 SE [43] 15.2 100 SE [41] 3.82 bend 78K [41] 12.0 110{111} SE [40]
37.1 111 SE [41] 2.94 bend 300K [41]
15.2 100 SE [43] 4.3 bend [41]
14.0 111 AI [61]
Si dia 25.0 SE [43] 22.4 100 SE [41] 4.14 bend [41] 14.7 110{111} SE [40]
26.4 AI [82] 47.3 111 SE [41] 7.60 bend [41]
32 111 SE [41]
24.8 111 AI [88]
32.0 111 SE [43]
22.4 100 SE [43]
22.0 111 AI [61]
C dia 138 SE [43] 168 100 SE [41] 20.7 (graphite) [41] 131 110{111} SE [40]
205 111 SE [41] 19.6 (graphite) [3]
205 111 SE [43]
168 110 SE [43]
90 111 AI [60]
95 111 AI [61]
225 100 AI [60]
130 100 AI [60]
Al fcc 12.0 AI [108] 12.1 100 AI [81] 2.27 bend [41] 0.75 112{111} SE [41]
11.9 SE [43] 11.1 111 AI [34] 0.9 112{111} SE [3]
11.0 AI [93] 3.49 100 SE [43] 2.62 110{111} SE [41]
11.0 111 AI [58,59] 2.56 112{111} SE [40]
13.1 100 AI [75] 3.50 112{111} AI [106]
9.2 100 AI [32] 3.40 112{111} AI [106]
8.95 111 AI [32] 1.85 112{111} AI [106]
4.89 110 AI [32]
Cu fcc 29.1 AI [83] 32.0 100 AI [54] 1.25 111 [3] 1.1 112{111} SE [41] 0.80 011 {011} [3]
26.0 AI [82] 36.0 100 AI [54] 2.94 111 [41] 1.29 112{111} SE [3] 1K 0.61 110 {111} [3]
19.9 SE [43] 41.0 100 SE [54] 1.50 100 [41] 1.20 112{111} SE [3] 10K 0.46 110 {111} [3]
20.9 AI [108] 39 111 SE [41] 1.74 100 [41] 2.93 112{111} SE [40]
20.2 AI [93] 38.7 100 SE [41] 1.59 110 [41] 5.30 112{111} AI [106]
25.0 100 SE [43] 1.71 111 [41] 4.00 112{111} AI [106]
38.0 100 SE [43] 2.65 112{111} AI [106]
33.0 001 AI [57]
31.0 110 AI [57]
29.0 111 AI [57]
9.3 001 AI [23]
9.4 001 AI [24]
Ag fcc 20.4 AI [83] 24 111 SE [41] 3.80 001 [3] 0.66 112{111} SE [3] 0.71 011{111} [3]
13.7 SE [43] 1.73 100 [41] 0.77 110{111} SE [3]
11.4 AI [93] 1.90 112{111} SE [40]
Au fcc 26.2 AI [83] 27.0 110 SE [41] 0.65 112{111} SE [41]
20.9 SE [43] 27.0 111 SE [43] 0.74 112{111} SE [3]
17.6 AI [108] 1.78 112{111} SE [40]
Pt fcc 35.1 SE [43] 5.26 112{111} SE [43]
33.6 AI [108]
Nb bcc 31.6 AI [108] 13.1 001 AI [35] 7.5 111{112} AI [106]
37.6 AI [83] 8.2 111{110} SE [43]
25.5 SE [43]
Calculations of theoretical strength 25
Appendix D: Continued 26
Crystal Structure Theoretical strength [GPa]
Threeaxial Uniaxial Shear
Theory Calculation method Theory Calculation method Experiment Theory Calculation method Experiment
Ta bcc 28.5 SE [43] 11.7 112{110} SE [43]
42.3 AI [83] 7.4 111{112} AI [106]
36.4 AI [108]
Fe bcc 24.1 SE [43] 47.9 111 SE [43] 13.1 111 [3] 7.3 111{110} SE [41] 3.56 111{011} [3]
J. Pokluda et al.

21.5 SE [43] 30.0 100 SE [43] 6.6 111{110} SE [3]

26.7 AI [84] 12.7 001 AI [65, 66] 11.5 111{110} SE [40]
14.2 001 AI [99]
27.3 111 AI [29, 66]
12.6 001 AI [36]
Zr bcc 20.7 100 SE [101]
Hf bcc 17.6 100 SE [101]
Mo bcc 50.4 AI [83] 28.8 001 AI [35] 17.8 111{112} AI [106]
42.7 AI [108] 20.9 111{112} AI [106]
34.9 SE [43] 16.4 111{112} AI [106]
15.6 111{112} AI [106]
V bcc 23.5 SE [43] 7.3 111{112} AI [106]
40.2 AI [83] 9.72 111{110} SE [43]
33.2 AI [108]
Cr bcc 25.9 SE [43] 20.5 111{112} AI [106]
21.0 AI [84]
Ni fcc 26.9 SE [43] 5.79 112{111} SE [43]
27.4 AI [84] 9.1 010{100} SE [41]
28.9 AI [103] 2.4 112{111} SE [3]
7.1 110{111} SE [3]
W bcc 53.1 AI [85] 28.9 001 AI [56] 24.7 [8] 18.2 111{110} SE [41]
57.4 AI [83, 82] 40.1 111 AI [56] 16.5 111{110} SE [3]
42.2 SE [43] 90.8 100 SE [41] 30.6 111{110} SE [40]
50.6 AI [108] 61.0 100 SE [43] 19.2 111{112} AI [106]
29.5 001 AI [81] 18.1 111{112} AI [106]
54.3 110 AI [56] 20.8 111{110} AI [106]
17.7 111{110} AI [106]
19.6 111{123} AI [106]
Ir fcc 45.7 SE [43] 18.6 110{111} SE [43]
57.7 AI [82] 24.1 112{111} AI [106]
Pb fcc 8.47 AI [83] 0.472 112{111} SE [43]
5.47 SE [43]
Co hcp 30.7 AI [84]
Ar fcc 0.25 SE [41] 0.32 100 SE [43] 0.14 110{110} SE [40]
0.146 SE [43] 0.08 112{110} SE [40]
TiC B1 44.0 001 AI [55]
NaCl B1 4.3 SE [41] 2.4 100 SE [43] 1.08 100 [41] 2.3 112{111} SE [41] 0.8 110{110} [3]
3.40 SE [43] 4.6 110 SE [41] 2.9 112{111} SE [3]
12.4 111 SE [43] 5.75 110{110} SE [43]
4.3 100 SE [41]
KCl B1 2.40 SE [43] 2.66 100 SE [43] 5.40 110{110} SE [40]
NaJ B1 2.22 SE [43] 1.89 100 SE [43]
NaBr B1 2.84 SE [43] 2.48 100 SE [43] 4.68 110{110} SE [43]
NaF B1 6.69 SE [43] 5.0 100 SE [43] 11.6 110{110} SE [43]
LiF B1 10.9 SE [43] 4.73 100 SE [43]
MgO B1 38.5 SE [43] 37.0 100 SE [43] 23.7 bend [41] 32.0 110{110} SE [40]
CsCl B2 2.10 SE [43] 1.82 100{110} SE [43]
TlBr B2 2.47 SE [43] 1.70 100{110} SE [43]
AgCl B1 4.09 SE [43] 3.70 111{110} SE [43]
ZnO B1 50.3 SE [43] 40.0 [3]
Al2 O3 46.0 0001 SE [41] 6.85 bend [41] 16.9 1120{0001} SE [41]
15.2 [41] 19.9 1120{0001} SE [3]
22.3 1120 [41]
10.9 0001 [41]
15.3 1010 [41]
SiO2 16.0 111 SE [41] 13.8 [41]
13.1 bend [41]
5.68 bend [41]
Ni3 Al L12 28.1 AI [103] 17.5 100 AI [105]
28.2 111 AI [105]
Al3 Ni 17.2 AI [103]
NiAl B2 25.6 AI [103] 46.0 001 AI [107]
25.0 111 AI [57]
β−SiC B3 101.0 001 AI [23]
50.8 111 AI [23]
Calculations of theoretical strength 27
View publication stats
Appendix D: Continued 28
Crystal Structure Theoretical strength [GPa]
Threeaxial Uniaxial Shear
Theory Calculation method Theory Calculation method Experiment Theory Calculation method Experiment
AlN B4 50.0 0001 AI [58,59]
MoSi2 WSi2 C11b 37.0 001 AI [102,104]
38.0 001 AI [102,104]
β−Si3 N4 P63 m 72.2 100 AI [43] 13.5 [41]
J. Pokluda et al.

57.0 100 AI [44] 14.0 [3]

75.0 001 AI [43]
55.0 001 AI [44]
c−Si3 N4 Fd3̄ m 45.0 001 AI [64]
Polyethylen 26.0 SE [3] 3.04 [41]
4.6 [3]
Kevlar 49 2.8–3.6 [3]
Nylon-66 0.8 [3]
Cellulose 16.0 SE [3]
AI, ab initio; SE, semiempirical.