Sei sulla pagina 1di 18

PROPERTIES OF THE ANISOTROPIC ELASTICITY TENSOR

By STEPHEN C. COWINf {Department of Biomedical Engineering, Tulane University, New Orleans, Louisiana 70118, USA)
[Received 18 April 1988. Revise 1 August 1988]
Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

SUMMARY A unified presentation of some properties of the fourth-rank tensor of anisotropic elasticity is given. The unified presentation involves both established concepts such as the Cauchy relations, the Voigt and Reuss bounds, planes of symmetry and specific directions of longitudinal wave propagation, and a new concept, the specific axis for pure shear wave amplitudes. The unified treatment employs the decomposition of the fourth-rank tensor of anisotropic elasticity into two symmetric secondrank tensors and an irreducible, completely symmetric and traceless, fourth-rank tensor. It is shown that a necessary and sufficient condition for a direction to be a normal to a plane of symmetry is that it be both a specific axis and a specific direction.

1. Introduction
ANISOTROPIC elastic materials are, both qualitatively and quantitatively, different from isotropic elastic materials. This paper contains a unified presentation of some of the anisotropic elastic properties represented by the fourth-rank tensor of elastic coefficients called the elasticity tensor. The constitutive relation for linear anisotropic elasticity is the generalized Hooke's law,

7]y = C / y t m w

(1.1)

which is the most general linear relation between the stress tensor whose components are Ttj and the strain tensor whose components are Etj, where the strain has been assumed to be measured from an unstressed reference state. The coefficients of linearity, namely Cijkm, are the components of the fourth-rank elasticity tensor. There are three important symmetry restrictions on the elasticity tensor Cijkm, restrictions that are independent of those imposed by material symmetry. These are the symmetries
Cijkm
=

Cjlkm>

Cijkm = Cijmk,

Cijkm Ckmij>

(1-2)

which follow from the symmetry of the stress tensor, the symmetry of the strain tensor, and the thermodynamic requirement that no work be
t Present address: Department of Mechanical Engineering, City College of the City University of New York, New York 10031, USA. [Q. Jl MM*. ppl. Math., Vol. 42, PI. 2,1989) Oxford IMnnitj Pros 1989

250

STEPHEN C. COWIN

produced by the elastic material in a closed loading cycle, respectively. The number of independent components of a fourth-rank tensor in three dimensions is 81, but the restrictions (1.2) reduce the number of independent components of Cijkm to 21. Since it has 21 independent components there is considerable information on the material properties represented by Ctjkrn. In order to make these properties apparent a decomposition of Ci/km into two symmetric second-rank tensors and an irreducible, completely symmetric and completely traceless fourth-rank tensor denoted by Zljkm is introduced. The two symmetric second-rank tensors employed in the decomposition were suggested by an earlier study by Cowin and Mehrabadi (1). These tensors are defined by
Q / Cijkkt *IJ C-ikjky (1-3)

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

and are called the dilatational modulus and the Voigt tensor, respectively. The rationale for this nomenclature is discussed below. The symmetry of C and V follow from (1.2). The traces of C and V are the only linear invariants of Cijkm (see, for example, Ting (2)). The inverse of the stress-strain relations (1.1) are the strain-stress relations E^K^T^, (1.4) where Kijkm is the compliance tensor which is related to the elasticity tensor CIJkm by K^C^ = HSirdj, + 6u6jr). (1.5) A decomposition of Kljkm similar to that employed for Cijkm is accomplished. The irreducible, completely symmetric, traceless fourth-rank tensor is denoted by Yijkm, and the two symmetric second-rank tensors are denoted by K and R, and their components are defined by These tensors are called the hydrostatic pressure modulus and the Reuss tensor, respectively. The rationale for this nomenclature is discussed in the next paragraph. Nomenclature was an important consideration in the preparation of this paper. Five new terms are introduced and specific nomenclatures are selected from a literature which often contains several choices. Sometimes authors introduce terminology to reflect their impression of the literature on the subject. For example, the phrase 'Cauchy relations' was introduced by Love (3, p. 14), but the equations described by this phrase first appeared explicitly in the monumental 1855 work of St Venant on torsion. The term does, however, reflect the history of the subject in that the molecular model that leads to the relations was one model considered by Cauchy in his developments of elasticity theory. The five terms introduced in this paper

ANISOTROPIC ELASTICITY

251

are put forward with less enthusiasm than the interpretation of their physical significance. They are introduced more from the need to make discussion smooth than from a conviction that they represent the best definitions. The term specific axis introduced in section 6 is reasonable given the Borgnis (4) definition of specific direction. The naming of the four tensors C, K, V and R as the dilatational modulus, hydrostatic pressure modulus, Voigt and Reuss tensors, respectively, has a lesser recommendation. The traces of C and V determine the estimate of Voigt for the effective isotropic moduli of a polycrystalline material containing randomly oriented single crystals with elastic constants described by Cijkm. The traces of K and V determine a similar estimate of Reuss where the elastic properties of the single crystals are represented by the compliance constants Kijkm. The dilatation modulus tensor C is the second-rank tensor of elastic constants representing the stress response of the elastic material to a pure dilatational deformation. The hydrostatic pressure modulus tensor K is the second-rank tensor of elastic constants representing the strain response of the elastic material to a pure hydrostatic stress. It can be shown that, if the deviatoric parts of C and K vanish, as they do in the cases of isotropic and cubic symmetry, then the associated form of Hooke's law permits the uncoupling of the hydrostatic stress from the deviatoric strain and the distortional stress from the dilatational strain. Information on the traditional 6 x 6 matrix notation for Cijkm and Kljkm is given in section 2 together with the introduction of some special notation employed in this paper. In section 3 it is shown that overly restrictive conditions on Cijkm known as the Cauchy relations correspond to the condition that the dilatational modulus tensor C and the Voigt tensor V be equal. The decompositions of the elasticity tensor Ctjkm and the compliance tensor Ky^ are described in section 4. In section 5, it is noted that the estimate of the effective isotropic moduli of a polycrystalline material given by Voigt (5) and based on Cijkm, and the one given by Reuss (6) and based on Kijkm, are determined by the traces of C and V and the traces of K and R, respectively. The acoustic tensor Q based on Cijkm is introduced in section 6 and the concepts of a specific direction and a specific axis are described. A specific direction was defined by Borgnis (4) to be a direction in which a pure longitudinal wave may propagate in an anisotropic elastic material. A specific axis is a pure shear amplitude axis, an axis about which it is possible to propagate a pure shear wave in any perpendicular direction with an amplitude lying along the axis. It is shown that a set of necessary and sufficient conditions that a direction a be a specific direction and a specific axis is that a be an eigenvector of C, V, the acoustic tensor based on Zijkm in the a-direction, Z(a), and the acoustic tensors Z(b) based on Zijkm in all directions b such that a . b is zero. Section 7 concerns planes of symmetry. Cowin and Mehrabadi (1) showed

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

252

STEPHEN C. COWIN

that the ten distinct material symmetries of linear elasticity could be classified by the number and orientation of their planes of symmetry. It is shown in section 7 that a set of necessary and sufficient conditions that a be a normal to a plane of symmetry is that it be a specific direction and a specific axis. Also, in section 7, formulae are presented for the calculation of the normals to the planes of symmetry given the numerical values of the components of Cijkm relative to an arbitrary coordinate system. The Appendix contains the representations for the different material symmetries of the components of Cijkm and the elements of its decomposition, Ctj, Vi, and Zljkm. 2. Notation As is customary in the discussion of linear anisotropic elasticity, we introduce the single index notation for stress (a, = Tn, o2= T-&, a 3 = T33,
CT4= 7^3 = 732, a 5 = T,3 = T31, a 6 = TX2 = T2l) and strain (, = , , , 2 = 22,

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

3 = ^33, 4 = 223 = 2 3 2 ) J = 2 1 3 = 2 3 1 , 6 = 2 1 2 = 2 2 1 ); thus (1.1) is replaced by aa = ca0Ep, (2.1)

where the 6 x 6 matrix with components cap represents the components of CQkm. The shift from an index system with a range of three (i, j = 1, 2, 3) to one with a range of 6 (or, /3 = 1,..., 6) is accomplished by the following rules for replacing a pair of indices ij by a single index: change 11 to 1, 22 to 2, 33 to 3, 23 to 4, 13 to 5 and 12 to 6. This change of notation incorporates the first two symmetries of (1.2) and the third is reflected in the symmetry of the 6 x 6 matrix c, where
iC\\
C

Cx2
C

C 13
C

C 14
C

Cls
C

Ci6\
C

12

22

23

24

25

26

Cl3

C-22 C33
C C

C34
C C

C35
C

C36
C

c=

(2.2)

14

24 25

C C

34 35

44 45

45

46

C]5
\C16

C55
C

C^
C

26

36

4<S

56

66/

The notational conventions adopted for the elasticity tensor, namely capital C for fourth-rank tensor components and lower case c for two suffix components and the matrix of such components, applies to the other fourth-rank tensors that appear, for example KIJkm, Zljkm and Y,jkm. For specific material symmetries the matrix (2.2) will have fewer than 21 non-zero and distinct components in special Cartesian-coordinate systems associated with the specific material symmetry involved. These special Cartesian-coordinate systems are referred to here as symmetry Cartesiancoordinate systems. The specific forms of the matrix (2.2) for particular

ANISOTROPIC ELASTICITY

253

material symmetries are given in the Appendix. The double-index notation for Kijkm is different from the double-index notation for C * ^ because of the factors of two occurring in the single-index notation for strain. The difference is that while Ck^ corresponds to caP with no numerical factors, the correspondence between Kijkm and kaP involves factors of 2 and 4.
Specifically, X m i , ^2222. ^3333. ^1122. ^1133. ^2233 correspond to kn, k^, ^-33> ^i2 *i3> ^23 respectively; 2X2311, 2Ki3n, 2Ki2u, 2X23221 2X 13 22, 2X1222)

2X2333, 2X 13 33, 2Kl233 correspond t o k4l, k5i, k6i, k42, k52, *62, * 4 3 , k53, * respectively; and 4X2323, 4X 1 3 1 3 , 4X 1 2 1 2 , 4X1323, 4X 1 3 1 2 , 4X 122 3 correspond to *44, ^5s ^66. ^54, ^56, ^64 respectively. T h e 6 x 6 matrix of components kaP is the matrix reciprocal of t h e 6 x 6 matrix of c o m p o n e n t s caP,

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

= 6a6.

(2.3)

Using the caP notation rather than the Cijkm notation, the dilatational modulus, Voigt, hydrostatic pressure modulus and Reuss tensors are given by the following matrices:

( (

C11 + C12+C13 C16 + C26 + C36

CHJ + C ^ + C-H; CJ2 + C22 + C23 Cu + C24 + CM I, J (2.4)

C1S + C2J + C35


CH+CSS + CK

C M + C24 + C34
C16 + C26 + C45

C16+C26+C4S C15 + C46 + C3J

C22 + C44 + C66 C24 + C34+C56

C24 + C34 + C56 I, C^ + C^ + CsJ

(2.5)

2(ku + ka + kl3) k^ k3S


5

kl6 + k26 + k3 2(kl2 + k22 + k^) kis + kn + kn


2kl6 + 2*26 + 4A:45 4*22 + *44 + *66 2*24 + 2*34 + 4*56

k15 kl4 + k2A + kM

\ I, )

(2.6)

2*24 + 2*34 + 4* |.

(2.7)

2*35 + 4*46

The traces of these matrices are


tr C = c,, + C22 + C33 + 2(c, 2 + c, 3 + C23), tr V = cn + C22 + c 33 + 2(044 + cS5 + C&), tr K = * n + *22 + *33 + 2(*12 + *13 + *23). tr R = *,, + *22 + *33 + K*44 + *55 + *)

(2.8)

The deviatoric part f of a second-rank tensor T is related to T by f = T-(trT/3)l. (2.9) Two special index-notation conventions are introduced for open products of two symmetric second-rank tensors whose components are AVl and Bkm.

254

STEPHEN C. COWIN means the sum of A,jBkm and (2.10)

The curly brackets around the product AljBkm A^B^ thus

{AyB,^} = AtjB^ + A^B,,.

The pointed brackets around the product of Ai,Bkm means the sum of AyB^ and a term obtained by interchanging one index of Ay with one index of Bfcn, A^Bkj, thus (AtjB^ > = AltBkm + A^B.j. (2.11)
Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

A combination of these two definitions has the following representation which is deducible from the definitions (2.10) and (2.11)

{(A^)}

= ({AyBton}) = AllBkm+AkmBli

+ AlmBk/ + AkiBim.

(2.12)

3. The Cauchy relations The symmetry relations (1.2) require that the components of Cljkm satisfy the following three relationships:
CmJcji = Cijlon = Cl/mk = Cjikm = Cjtmk = Ckmij = Ciunjj = Cmkjj, I Cmikj = Cjjnj/c = Ctnjy = Cji^^ = CjUm = Wy m / = Cyfa = Cmijk, r Cmjik ~ Cuunj = Cucjm = Cjmik = Cimki = CUjm = Ci^nj = Cmjkh J (3.1)

It is easy to see that if in addition to the indicial symmetries represented by (1.2) the components of Cljkm enjoy the symmetry Cijkm
=

Cikjm,

(3-2)

then the three relationships of (3.1) collapse to one. Thus, if (3.2) as well as (1.2) is satisfied, the tensor Ctjkm would be completely symmetric. It follows that Cijkm may be decomposed into a sum of two fourth-rank tensors, one of which is completely symmetric: 2C/ytm = (Cijkm + Q*/m) + {Ciikm Cikjm). (3-3)

The second summand on the right-hand side of (2.8) has the representation

- Clqmq)},

(3.4)

which is an identity. It is easily verified by direct integer substitution and it always yields the null identity except for the ijkm sets 1122, 1133, 2233, 1123, 2213 and 3312. Substitution of the definitions (1.3) for the dilatational modulus and Voigt tensors into (3.4) gives Qfa, - C a ^ = (tr C - tr V)(5flt<5ym - <5,y<5tm)/2 + + { W * - V^,)} - { ^ - Vjm)). (3.5) Equation (3.5) shows that (3.2) is equivalent to the condition of equality

ANISOTROPIC ELASTICITY between the dilatational modulus and Voigt tensors, C = V.

255

(3.6)

The use of (2.4) and (2.5) shows that the conditions (3.2) and (3.6) are also equivalent to the following six scalar relations between the elasticity tensor components
C

44 = C23,

5 5 = Ci3,

C6e> = Cl2,

C45 = C36>

C46 = C25,

56 = C U .

(3.7)

The relations (3.7) are called the Cauchy relations (3, p. 100). The validity of the Cauchy relations was a subject of major controversy in the last century, Love (3). The resolution of the controversy was that the relations do not hold for most elastic materials, but only for materials which can be described as having central-force laws operating between points of a simple lattice, Musgrave (7). 4. Decomposition of the elasticity and compliance tensors Backus (8) showed that the elasticity tensor Cijkm can be decomposed into two scalars, denoted by a and b, two traceless symmetric second-rank tensors A and 6, and one completely traceless and completely symmetric fourth-rank tensor, denoted here by Zijkm, thus Citkm = adySto + b < <5*<5,m> + { V W + {< My >} + Zijkm, where Ziikk = Zikjk = 0 (4.2) and Zyian enjoys all the symmetries (1.2) and (3.2). In order for the relationship (4.1) to be consistent with the definitions (1.3) of the dilatational modulus and Voigt tensors it is necessary that (4.1)

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

7A = 5t - 4V,
and a = (2trC-trV)/15, thus (4.1) takes the form Cijkm = ((2 tr C - tr V)/15) 5^^

7ft = 3V - it,
b = (3 tr V - tr C)/30; + ((3 tr V - tr C)/30)<<5<5,m> +

(4.3)
(4.4)

The formula (4.5) gives a representation of Cijkm in terms of Cijt Vtl and Zijkm. There are six independent components of both C and V and nine independent components of Zl/km due to the restrictions of complete symmetry (1.2) and (3.2), and complete tracelessness (4.2). These sum to 21 independent components corresponding to the 21 independent components

256

STEPHEN C. COWIN

of Cijkm. Using the double-index notation for the components of Zijkm, they are expressed as
Zn 2 2 Z3)
255
2*6 = 2i2 = 2 3 ,
2

233 Z\ Z 2 ,
=

2i3 =

(4.6)
224 2 7 Z 4 , Zis = Zg ~ Z5, Z26
=

36 = 245 = 2Z6

Z34 = - (2 7 + Z 4 ) , 2i6
=

2 3 5 = - ( 2 5 + 28), Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

29 Z5,

( 2 6 + 29),

where 2l = Men ~ 4(C22 + C33) + 9(C23 + 2cJ) - (c 1 2 + 2 c + C13 + 2 c 5 5 ) V ! - 4(C U + C33) + 9(C 13 + 2C55) - (C12 + 2C66
23 = Z4 = 25 = Z6 = Z7 =

9(c 1 2 + 2 c 6 6 ) - ( c 1 3 + 2c 5 5
-14 + 4C56 -C24C34),

(4.7)

; + 4C46 - C 15 - CJJ), , + 4 c 4 3 - Cj 6 - Cje), ' C34) 2g = j ( c 1 5 C35), Zg = 2( C 16 C26

Because ZIJkm is completely symmetric, the components of Zijkm must satisfy the Cauchy relations (3.7) and, from (4.6), this is indeed the case. The other relationships between the components of Z,jkm expressed by (4.6) follow from the total tracelessness condition (4.2). The component form of the decomposition (4.5) breaks into four distinct sets of equations: 70c n = 18C,, - 2C22 - 2C33 + 36V n - 4V22 - 4V33 - 70(z2 7OC22 = 1 8 ^ - 2CU - 2C33 + 36V22 - 4V33 - 4K,, - 70(z, + z 3 ), 70c33 = I8C33 - 2CU ~ 2CU + 36V33 - 4 V n - 4 ^ - 70(z, + z 2 ), 70c12 = 26(C n + C^) - 24C33 - 18(V n + V^) + 22F 33 + 70z3, 70c13 = 26(C n + C-n) - 24C22 - 18(V n + V33) + 2 2 ^ + 70z2, 70C23 = 26(C22 + C33) - 24C n - 18(^22 + V33) + 22V,, + 70z,, 70C+4 = l i d - 9(C22 + C33) + 17(V22 + V33) - 13Vn + 70z,, 70c3S = IIC22 - 9 ( C n + C M ) + 17(V,, + V33) - 13V22 + 70z2, 70C66 = IIC33 - 9 ( C n + C22) + 17(VU + ^ 7c, 4 = 5C23 - 4V23 + 14z4,
lcM = C23 + 2V23 - 7z4 - 7z 7 ,

(4.8a)

- 13V33 + 70z3, 1 + 14z4,

7c 24 = C23 + 2V23 - 7z4 + 7z 7 ,


7 Cs6 = -2C23 + 7 C 2 5 = 5 C 1 3 - 4V 13 + 14z 5 , 7 C 4 6 = - 2 C 1 3 + 3V I3 + 14z 5

7c 1 5 = C 1 3 + 2K 13 - 7z 5 + 7z 8 , Vc33 = C 1 3 + 2V 13 - 7z 5 - 7z8,

] (4.8c)

ANISOTROPIC ELASTICITY 7c16 = C12 + 2Vl2 - 72c + 7z<,, 7c M = C12 + 2V12 - lz6 - Iz?,' '} 7 Cj6 = 5C12 - 4V12 + 14z6) 7c4J = - 2 C , 2 + 3V,2 + 14%,

257

(4.8d)

In the first set the nine elasticities cn, c^,, c33, c12) c13, C23, C44, c 5J and c^ are related to Cn, C22, C-u, Vu, V^, V33, zu z2, z3. In the second, third and fourth sets, cu, c24, C34 and c M are related to C^, V^, z4 and z7; c15, C25, c35, and c ^ are related to C13, Vi3, zs and z8; and c16, c M , c-*, c4J are related to Cl2, Vl2, z* and Zg, respectively. These results are interesting because they illustrate the interconnected patterns in which different coefficients of c vanish as different material symmetries are considered. For example, for orthotropic symmetry, transversely isotropic symmetry (that is, the five-constant hexagonal symmetry), cubic symmetry and isotropic symmetry the second, third, and fourth sets of equations all vanish. For monoclinic symmetry the second and third sets all vanish. For the seven-constant tetragonal symmetry the second and third sets and the last two equations of the fourth set all vanish. For the seven-constant hexagonal symmetry the third and fourth equations of the second and third sets vanish as well as the fourth set of equations. Similar results apply to the six-constant hexagonal and tetragonal symmetries. The components of C, V and z for each of the ten distinct elastic symmetries are given in the Appendix. It is interesting to note that while the components of C, V and z are different for each crystal system (triclinic, monoclinic, or orthorhombic, tetragonal, hexagonal, cubic) and isotropy, the components of C and V are the same for each class contained in a system and the classes within a specific system differ only by differences in the components of z. In the Appendix one can see that this is true for the three classes of the hexagonal system and the two classes of the tetragonal system with distinct elastic symmetries. For these classes, distinction occurs only on the basis of z within the crystal system. The decomposition of Kijkm into Ktj, RtJ and a completely symmetric and traceless tensor Yijkm is given by Kqhn = ((2 tr K - tr R ) / 1 5 ) A + ((3 tr R - tr K)/30)(6 lk 6 /m ) + (4.9) The formula (4.9) gives a representation of Kijkm in terms of Ktj, R,j and Y,lkm. The formulae (4.6) and (4.7) for the components of zaP can be used to write formulae similar to (4.6), (4.7) and (4.8) for yaP if the factors of 2 and 4 discussed in section 2 are properly accounted for. 5. Effective isotropic moduli for anisotropic materials Voigt (5, p. 956) addressed the question of determining the effective isotropic elastic moduli of a polycrystalline material composed of a large number of small crystal grains whose orientation within the material is
/ y^.

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

258

STEPHEN C. COWIN

completely random. The approach of Voigt employs the elasticity tensor Cijkm and the fourth-rank-tensor transformation law for Ctjkm, CIJkm = QlpQtoQkrQnvC'^, (5.1) where the orthogonal tensor Q represents an orthogonal transformation from the xt to the x\ basis. Let C'Mn denote the actual constants of the single crystals that are randomly distributed in their spatial orientation and let 6, <p, ip be Eulerian angles. The average of C^ over all angular orientations is denoted by ClJkm and is denned by
(5.2)

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

where the tensor components C^ are not functions of the Eulerian angles, but where Q is a function of the Eulerian angles. It is straightforward, but tedious, to show that the average over all orientations of any completely symmetric and completely traceless second- or fourth-rank tensor is zero. For second-rank tensors the formula (5.2) is modified in the obvious way. This result means that the averages of C,jt Vu and Ztlkm over all orientations are all zero. It follows then that when (4.5) is substituted into (5.2) the only terms that survive the averaging process are those involving Kronecker deltas, thus CQkm = (Kv - IGvWdifitn, + Gv (6lk6jm >, (5.3) where Gv = i ( 3 t r V - t r C ) . (5.4) These are Voigt's estimates for the isotropic bulk and shear moduli, respectively. The method of Voigt described above is equivalent to assuming that the strain is homogeneous; Reuss (6) suggested a second method of orientationaveraging that is based on the compliance tensor. The Reuss method is equivalent to assuming that the stress is homogeneous. The space average is then defined by the formula (5.2) with Cljkm replaced by Kljkm. Employing the algebraic operations similar to those used in obtaining (5.3) it is easy to show that Kiikm = ((1/9KR) - (1/6GK))6<A + (U4GR)(6ik6im), where KR = (l/tTK), GR = 15/(2(3 t r R - t r K ) ) . (5.6) Hill (9) showed that the actual effective bulk and shear moduli, denoted by K and G respectively, are bounded by the Voigt and Reuss estimates KR^K^KV, GR^G^ Gv, (5.7) with a similar inequality holding for the Young's modulus. (5.5)

ANISOTROPIC ELASTICITY

259

The relationship of trC and trV to the estimates of Voigt (5) and the relationship of tr K and tr R to the estimates of Reuss (6) are the reason that V and R are called the Voigt and Reuss tensors, respectively. The relationships (5.4) and (5.6) appear in the articles of Walpole (10) and Katz and Meunier (11) in different notations. 6. Specific directions and specific axes In this section the relationship of the results presented in previous sections to the conditions for the propagation of plane waves in anisotropic elastic solids is considered. Let n denote the direction in which the wave is propagating and let q denote a unit vector in the direction of the amplitude vector. The value of q for a wave in the direction n is the solution to the eigenvalue problem, Green (12), Kelvin (13), Qik(n)qk = pv2qh (6.1) where p is the density of the anisotropic material, v is the phase velocity of the wave and Q(n) is the acoustic tensor in the direction n defined in terms
of Cijkm by G ( ) C (6.2)

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

If a, b and c are an orthonormal triad of vectors, it follows from the definition (1.3)2 of the Voigt tensor V and (6.2) that G*(a) + G*(b) + Qu.it) = Vlk. (6.3) This identity will be useful in the proof of a theorem later in this section. In a different notation it appears in Hayes (14). A wave is said to be longitudinal if q = n and to be transverse if q. n is zero. In isotropic materials waves are either longitudinal or transverse, but this is not the case for anisotropic elastic media. In anisotropic media, the amplitude vector does not have to be parallel or perpendicular to the direction of wave motion. Borgnis (4) introduced the concept of specific directions for longitudinal waves in anisotropic elastic materials. He noted that there are only specific directions in which a pure longitudinal wave can propagate in an anisotropic elastic material. For a longitudinal wave, the n and q of (6.1) must coincide, so the specific directions a are the solutions to the nonlinear eigenvalue problem Qik(*)ak = pvjah (6.4) where u, is the phase velocity of the longitudinal wave. The concept of a specific axis related to pure transverse wave propagation will now be introduced. Let a denote a unit vector and let b denote a direction perpendicular to a. If, for all directions of propagation b such that a. b = 0, it is possible to propagate a pure transverse wave with amplitude in the a-direction, the axis a is said to be a pure transverse amplitude specific

260

STEPHEN C. COWIN

axis or, for brevity, a specific axis. The condition for a specific axis a to exist is that a be an eigenvector of all Q(b) for which a . b is zero, thus Qik(b)ak = pvfa for all b such that a . b = 0, (6.5) where v, is the shear-wave velocity in the direction b. From the definition (6.5) of a specific axis a, it is easy to see that, if (6.5) holds for b, it also holds for c where c . a = 0 and c . b = 0. Thus, if a is both a specific direction and a specific axis, then it follows from (6.3) that it is also an eigenvector ofV. A representation for the propagation condition (6.1) in terms of the dilatational modulus and Voigt tensors is obtained by substituting the decomposition of C,jkm given by (4.5) into (6.2), using the definitions (5.4), and then placing the result into (6.1), thus Qut(n)qk = (n . q)((Kv + Gv/3)ni + ((3<?* - Vlk)H)nk) + + qt{Gv + ((3?,m - 2Cjm)n)njnm) + n,((3Cb - ^km)/7)nmqk + + ((3<k - 2 t t )/7) 9 t + Zik{n)qk, (6.6) where Zlk(n) is the reduced acoustic tensor defined by Zu&n) = Zljkmnjnm - Zlkjnjitnm. (6.7) From (4.2) it follows that tr Z(n) vanishes. The identity analogous to (6.3) for Z rt (n) is Z,t(a) + Ztk(b) + Z*(c) = Z, w = 0 (6.8) for an orthonormal triad of vectors a, b and c. If both n and q in (6.6) are set equal to a one obtains the relation G*(a)* = M + ((<?* + 29lk)/7)ak + Z ^ a K , where A = Kv + 4G v /3 + ((<?*,, + 2fkm)/7)akam. e*0)fl* = Aa, + <t>b, + ((3<k - 2 ^ ) / 7 ) a t + Z(b)a t for all b such that a . b = 0, where A = Gv + ((3<L - 2<?fcn)/7)^ftm, = ( ( 3 ^ - <L)/7)6Ma*. D (6.12) The following theorem may be proved easily using the equations (6.9) and (6.11). A set of necessary and sufficient conditions that a direction a be both a specific direction and a specific axis is that a be an eigenvector ofC,\, Z(a) and all Z(b) for which b . a is zero. (6.10) (6.11) If n in (6.6) is set equal to b and q set equal to a, where a. b is zero, then (6.9)

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

ANISOTROPIC ELASTICITY

261

Sufficiency follows directly from (6.9) and (6.11) since if a is an eigenvector of C, V, Z(a) and all Z(b) for which b . a is zero, then it is also an eigenvector of Q(a) and all Q(b) for which b . a is zero. The necessity proof begins with the observation in the paragraph following (6.5) that if a is both a specific direction and a specific axis, then it is also an eigenvector of V. Substituting (6.4) into (6.9) and (6.5) into (6.11) and using the fact that a is an eigenvector of V, (6.9) and (6.11) take the following forms: ipa, = $Cikak + Z*(aK, Za, = (^to,ftmfl*)6.- - iCufit + Z*0K for all b such that a . b is zero, where n)bkbm. When (6.14) is contracted with bt it follows that =0 ^Ctmfl* + Zfcn(bK = Tam. Elimination of Z ^ b ) ^ between (6.17) and (6.14) yields b, - Cu.0, (6.18) for all b such that a . b is zero. Contraction of (6.18) with at shows that l(T - 2) is equal to C^a^t, thus (6.18) can be rewritten as taflk ~ (*,*)/ = (.hbmak)b, (6.19) for all b such that a . b is zero. Since (6.19) is true for all b such that a . b is zero, it is also true if b is replaced by c where c . b and c . a are both zero, thus (6.20) From (6.19) and (6.20) it follows that (C^b^^b, = (C^c^c, and since b . c is zero one concludes that ^kn,bmak = Cbn^at = 0, (6.22) where a, b and c form an orthonormal triad. From (6.19) and (6.22) it follows that a is an eigenvector of C, and from (6.13), (6.14) and (6.22) it follows that a is also an eigenvector of Z(a) and all Z(b) for which a . b is zero. This completes the proof of the therorem. (6.21) (6.16) (6.17) for all b such that a . b = 0. It also follows then that there exists T such that
{

(6.13)
Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

(6.14)

262

STEPHEN C. COWIN

7. Planes of symmetry The problem of determining the type of elastic symmetry possessed by a material given the numerical values of the fourth-rank elasticity tensor relative to an arbitrary coordinate system was considered by Cowin and Mehrabadi (1). They solved the problem by constructing the number and orientation of the planes of symmetry characterizing each of the ten traditional and distinct elastic symmetries and by proving a theorem which provides the machinery for determining the number and orientation of the planes of symmetry of a given Cljkm relative to an arbitrary coordinate system. A plane of symmetry at a point in an elastic material is a plane with respect to which the material has reflective symmetry. The main result of Cowin and Mehrabadi (1) is that a set of necessary and sufficient conditions for the vector a to be a normal to a plane of symmetry of a material of given elasticity is that a be an eigenvector of C and V, a specific direction (6.4) and a specific axis (6.5). In section 6 it was shown that if a is a specific direction and a specific axis, then it is also an eigenvector of C and V. Thus, the combination of the results of section 6 with the theorem of Cowin and Mehrabadi (1) proves the following theorem. A necessary and sufficient condition that a be a normal to a plane of symmetry in a linear elastic material is that it be a specific direction and a specific axis. Another combination of the theorem obtained in section 6 and the theorem of Cowin and Mehrabadi (1) produces a simplified calculational procedure for determining the normals to planes of symmetry of a material characterized by Cljkm. A set of necessary and sufficient conditions for the vector a to be a normal to a plane of symmetry of a material with given elasticities ClJkm is that a be an eigenvector of C, V, Z(a) and all Z(b) for which a . b is zero. The matrices C and V are given by (2.4) and (2.5) respectively. The matrices Z(a) and Z(b) can easily be obtained from formulae for the components of Z(n): Z n ( n ) = z3{n\ - ni) + z2{n\ - n\) + 2(z<> - z^)nxn2 + 2(z 8 - z5)ntn3 + 4zAn2n3,y ? ~ n\) + z^n\ - n\) + 4z6n1n2 - 2(z 5 + z*)nxn3 - 2(z4 + z7)n2n3, Z 12 (n) = z9(n\ - n2) + z6(3n\ - 1) + 2z3nxn2 + 4zAnxn3 + 4zsn2n3, Z 13 (n) = zs(n\ - n\) + zs{3n\ - 1) + 4zAnxn2 + 2z2/z,/z3 + 4zf>n2n3, Z ^ n ) = z-fai\ - n\) + zA(3n\ - 1) + 4z 6 n,n 3 + 4z5nln2 + 2z1n2"3> (7.1) where the constants zt to z^ are related to the components of c ^ by (4.8).

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

ANISOTROPIC ELASTICITY
Acknowledgments

263

A portion of this work was done when I was supported on a fellowship from the Japan Society for the Promotion of Science (JSPS) and a guest in the laboratory of Professor M. Satake at T6hoku University. I thank JSPS for their support and Professor Satake and his laboratory for their great hospitality. The remainder of this investigation was supported by USPHS, grant DEO6859 from the National Institute of Dental Research, National Institutes of Health, Bethesda, Maryland 20205. I am indebted to Dr L. J. Walpole for pointing out the Backus (8) reference and to Professor M. M. Mehrabadi for comments on the present paper at its formative stages.
REFERENCES 1. S. C. COWIN and M. M. MEHRABADI, Q. Jl Mech. appl. Math. 40 (1987) 451. 2. T. C. T. TINO, ibid. 40 (1987) 432. 3. A. E. H. LOVE, A Treatise on the Mathematical Theory of Elasticity, 4th edition (Dover, New York 1927). 4. F. E. BORGNIS, Phys. Rev. 98 (1955) 1000. 5. W. VOIGT, Lehrbuch der Kristallphysik (Teubner, Leipzig 1910). 6. A. REUSS, Z. angew. Math. Mech. 9 (1929) 55. 7. M. J. P. MUSGRAVE, Crystal Acoustics (Holden Day, San Francisco 1970). 8. G. BACKUS, Rev. Geophys. Space Phys. 8 (1970) 633. 9. R. HILL, Proc. Phys. Soc. London A65 (1952) 349. 10. L. J. WALPOLE, Adv. appl. Mech. 21 (1981) 169.
11. J. L. KATZ and A. MEUNIER, / . Biomech. 20 (1987) 1063.

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

12. G. GREEN, Trans. Camb. phil. Soc. 7 (1839) 121; Mathematical (Macmillan, London 1871) 307.
13. Lord KELVIN (WILLIAM THOMPSON), Elasticity, Encyclopedia

Papers

Britannica

Chapter XVII (Adam and Charles Black, Edinburgh 1878). 14. M. HAYES, Arch, ration. Mech. Analysis 46 (1972) 105.

APPENDIX The purpose of this appendix is to record the component forms for c, z, C and V for each of the ten distinct elastic symmetries. Similar results for k, y, K and R can be obtained using the notational conventions described in section 2. A number of devices are used to present this information in a manner that reduces its space on the printed page. First, four of the ten distinct symmetries are presented as special cases of the remaining six. Both of the distinct symmetries of the tetragonal system, the seven- and six-constant symmetries, are presented together. All three of the distinct symmetries of the hexagonal system, the seven-, six- and five-constant symmetries, are presented together. Lastly, isotropic symmetry is presented as a special case of the distinct symmetry of the cubic system. Secondly, in a seven by seven format the components of c are given under the diagonal and the components of z are given over the diagonal, c\z. The components of c are given exactly in the notation of Voigt (6). Thirdly, in a four by four format, the components of V are given under the diagonal and the components of C over the diagonal, V \ C . Using these

264

STEPHEN C. COWIN

conventions, the components of c\z and V\C for each of the distinct elastic symmetries are given below. I. Tnclinic system 2z4 Z2 Z7 Z h z, - Z i - z2 z 7 z .
4

_
Cn
C\i

Z2 Z3
C22

z3

Zg-Z,

Zg
-

2z5
Z , Zg

"Z9

Cn Cu
Cl5

c-a
C24 C25 C26

C33

CM

Zi -

2z6
Zz C

cM
C33 C36

C45
C

C33 C56

Cl6

46

66

2^ 2z5 2z4 z3

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

C n + C55 + C M

C,, + C u + C13 -

C16 + C-K, + C36 C12 + C22 + C23 C33 + C M + C35

C15 + C25 + C35 C14 + C-n + C35

The relationships between z, to z, and the components of c ^ are given by (4.8). The number of distinct constants associated with c, z, V and C is 21, 9, 6 and 6 respectively. II. Monoclinic system
cn C12 C,3 0 0
C16

-Zi-Zi C22 Ca 0 0
C26

Z3 Zi z 3 C33 0 0
C36

Z2 Zi z, z2 CM c
0

0 0 0 Z, c55
0

0 0 0 2z,s Z2 CM

Zg-Zf, z , Z(, 2z6 0 0 z3

Cll + C55 + CM C16 + C26 + C45

Cu + C,2 + C13 ~ C ^ + C ^ + CM

CI6 + CM + Cj* C,2 + C22 + C23 -

0 0 C,3 + C23 + C33

The relationships between zt, z2, z3, z,, and z? and the components of c are those given by (4.8). The number of distinct constants associated with c, r, V and C is 13, 5, 4 and 4 respectively. HI. Rhombic system z2 0
Zi -Zi - z 2
-

Cn c, 2
C13

-z2-z3
CZJ
C23

z3
C33

0
0 0
0

0
0 0
0

-Zi - z 3

0 0
Zi

0 0 0

0 0 0

0 0 0

c 0 0

c55 0 0

z2 CM 0

0 z3 -

c n + c 1 2 + c13

C33 + C^ + C33

ANISOTROPIC ELASTICITY

265

The relationships between zt, z2 and z3 and the components of c are those given by (4.8). The number of distinct constants associated with c, z, V and C is 9, 3, 3 and 3 respectively. TV. Tetragonal system z, 0 0 Z-22, 0
2,
CM

^1

^3

Z\

0 0 0 0
2,

29 -2,

0 0
Cl6

0 0

0 0 0
23

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

CM

0 + C,, + C,

0 0 0 0
C33 + 2c 1 3

c n + c 0 0

CU + CM

The components zu 23 and Z9 are related to the c by 2, = i ( 8 ( c I 3 + 2cM) - (cl2 + 2c) - 3 c n - 4c 33 ), 23 = is (-2(c I 3 + 2c) + 9(c 12 + 2c) - 8 c n + C3,), The representations given above are for the tetragonal seven-constant symmetry. The tetragonal six-constant is the same except that c16 and hence 2, is zero. The number of distinct constants associated with c, z, V and C is 7, 3, 2 and 2 respectively for the tetragonal seven system, and 6, 2, 2 and 2 respectively for the tetragonal six system. V. Hexagonal system
Cu
C12 Cl3
CM

32 3 -

23

cn
C13
- C M

32 3 C33

-423 -423 823 CM

2z -22, 0 -42 3
CM

-225 22 5

0 0
423
2AC11 C12)

0 0 0 2z5
22,
23 -

-C25

C25

0 _
-Cl2)

0
C

0 0 0

0
C25
C

CM

: + cl2-\ "
0

13

0 C.l+C,2
-

+ C.3

0 0
C33 + 2c 1 3 _

0 0

+ i(3c,, c,i)

Cn + 2

The components z3, Z4 and Zj are related to the c^ by

The representations given above are for the hexagonal seven-constant material symmetry. The hexagonal six-constant symmetry is the same except that z, and ca are zero, and the hexagonal five-constant symmetry (that is, transverse isotropy) is also the same except that 2,, 2, and c u , c^ are zero. The number of distinct constants associated with c, z, V and C is 7, 3, 2, 2 respectively for the seven-constant case, 6, 2, 2, 2 respectively for the six-constant case and 5, 1, 2, 2 respectively for the five-constant case.

266

STEPHEN C. COWIN VI. Cubic system and isotropic -2zi Zi z, 0 -2z, Z! 0 cu -2z, 0
0 0 0 /
~\2

cu c 12
/
~12

symmetry 0 0 0
n
\J

0 0 0
n
\J

/ 0 0 0

''ll

T
Z\

0 0 0

C44 0 0

044 0

z, C44

0 z, -

c n + 2c44 0 0 For the cubic system

c n +2c 1 2 cn+2cM 0

0 c n + 2c12 cu + 2cM

0 0 cn+2c12

Downloaded from qjmam.oxfordjournals.org at Sana'a University on January 31, 2011

\ 3^ C i2 -T ZC44 ~

X\)

and the number of distinct components associated with c, z, C and V is 3, 1, 1 and 1 respectively. For isotropic symmetry 2cM = cu - c12, Zt=0 and the value c n + 2c*4 appearing along the diagonal of V is replaced by 2c u c ]2 . For isotropic symmetry the number of distinct components associated with c, z, V and C is 2, 0, 1 and 1 respectively.

Potrebbero piacerti anche