This action might not be possible to undo. Are you sure you want to continue?

BooksAudiobooksComicsSheet Music### Categories

### Categories

### Categories

Editors' Picks Books

Hand-picked favorites from

our editors

our editors

Editors' Picks Audiobooks

Hand-picked favorites from

our editors

our editors

Editors' Picks Comics

Hand-picked favorites from

our editors

our editors

Editors' Picks Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

A study of the possibilities opened up by Kelvin 150 years ago

**Klaus Helbig, Hannover
**

The Structure of Tensors 1

**The customary ³official´ representation
**

The tensor cijkl connects the symmetric stress tensor Wij symmetrically with the symmetric strain tensor Ikl The tensor cijkl has 34=81components The two ³symmetries´ of stress and strain mean Wij = Wji , Ikl = Ilk , thus cijkl = cjikl = cijlk The ³symmetrical´ connection means cijkl = cklij Thus only 21 of the 81 components are significant!

The Structure of Tensors 2

Hooke¶s Law

**There are still 36 terms, but, e.g., c2313=c1323
**

The Structure of Tensors

3

**The Voigt mapping
**

With only 21 significant components, the elastic tensor can obviously be mapped on a symmetric 6v6 matrix Such ³mapping´ should preserve the elastic energy density

**2E = Wij Iij = WpIp
**

The Voigt mapping achieves this by the mapping rules

**p = i Hij+ (1±Hij)(9±i±j ) , q = k Hkl+ (1±Hkl)(9±k±l ) Wp = Wij , cpq = cijkl , Iq = (2 ± Hkl )Ikl
**

The Structure of Tensors 4

The Voigt mapping. visual W11 W12 W13 W12 W22 W23 W13 W23 W33 W11 W22 W33 W23 W13 W12 I11 I12 I13 I12 I22 I23 I13 I23 I33 I W1 1 11 I W 22 I W 33 Mapping for the stress tensor Mapping for the strain tensor I23 W 23 13 I13 W But this does.I! The Structure of Tensors I12 W But this would not keep 5 . 12 the scalar product W.

Properties of the Voigt mapping Advantages The Voigt mapping preserves the elastic energy density The Voigt mapping preserves the elastic stiffnesses Disadvantages Stress and strain are treated differently The norms of the three tensors are not preserved The entries in all three Voigt arrays are not tensoror vector components. thus we loose all advantages of tensor algebra The Structure of Tensors 6 .

Mohr¶s Circle.Lost Advantages of Tensor Algebra There is no ³invariant representation´ Representation in another coordinate system by the simple rule I¶kl = rki rlj Iij is not possible For rotation of the coordinate system. and other constructs. one has to use the Bond relations. The Structure of Tensors 7 .

The Kelvin mapping 2E = Wij Iij = WpIp The Kelvin mapping preserves E by the mapping rules p = i Hij+(1-Hij)(9±i±j ) . q = k Hkl+ (1±Hkl)(9±k±l ) Wp = (Hij+¥2(1±Hij )) Wij . The Structure of Tensors 8 . Iq = (Hkl+¥2(1±Hkl)) Ikl . Cpq = (Hij+¥2(1± Hij )) (Hkl+¥2(1± Hkl)) cijkl .

thus we keep all advantages of tensor algebra Only disadvantage: the values of the stiffness components are changed The Structure of Tensors 9 .Properties of the Kelvin mapping Advantages The Kelvin mapping preserves the elastic energy density The norms of the three tensors are preserved Stress and strain are treated identically The ³maps´ of stress. and stiffness have all properties of tensors of 1st respectively 2nd rank in 6D-space. strain.

The Structure of Tensors 10 . For some problems the 4th-rank tensor is the most convenient notation. it does not mean that the norm is lost. The description we choose is not a question of ideology. only that we cannot access it easily as long as we use this description. but there are problems that are much more easily solved in the Kelvin form.A tensor is a tensor by any name! It is important to realize that a tensor is a physical entity that does not depend on the way we describe it. For instance: if in a mapping (a change of description) the norm of the tensor is not defined (as in the Voigt mapping). but of scientific economy: for many problems the Voigt mapping is the natural choice.

The strength of the Kelvin notation is the possibility to reduce an elastic tensor to its invariant (coordinate free) representation. and conversely to ³construct´ tensors with given invariants. elastic parameters are listed in this notation. For this reason. algorithms to deal with tensors in Voigt notation are useful. The 4-subscript tensor notation is convenient for operations as the change of coordinate systems: the very definition of a tensor is based on this operation. and ³users´ expect that results are listed in this form.Which notation? Each has is place! The Voigt notation is the de facto standard in the ³outside world´: in the entire literature. The Structure of Tensors 11 . It should be used in the analysis of tensors.

and the relevant routines can be freely exchanged. In ³Reader´ format it can be used on any computer without the program. and the 6561 multiplications needed for a transformation of the coordinate system might have taken up to a minute.. Today a scientist is hardly ever without access to a computer.e.Even ten years ago. i. Most of the important operations give results ³instantly´. a conversion from one notation to another was not a trivial matter. with response times below two seconds. The Structure of Tensors 12 Practical aspects of multiple notations . The current project will be completed with a Tensor Toolbox written as a Mathematica notebook.

A simple tool for conversion For conversion between the Voigt. only one array is needed: G= 1 1 1 ¥2 ¥2 ¥2 WK = WV G WV = WK/G IK = IV / G IV = IK G For the stiffnesses.and Kelvin notation. one uses the ³outer product´ of G 1 1 1 1 ¥2 ¥2 ¥2 1 1 1 ¥2 ¥2 ¥2 ¥2 ¥2 ¥2 2 2 2 ¥2 ¥2 ¥2 2 2 2 ¥2 ¥2 ¥2 2 2 2 13 CK = 1 1 ¥2 ¥2 ¥2 CV The Structure of Tensors .

A strain can be described by 6 linearly independent ³base strains´. The line of thought can be described like this: 1. It is convenient to use a set ³orthogonal types´ for the common basis. A stress can be described by 6 linearly independent ³base stresses´.How did Kelvin come to his description? Not as a mapping of a 4th-rank 3D tensor on a 2nd-rank 6D tensor. 2. It is convenient to use bases of the same ³type´ for stress and strain. The ³weight´ of the different components should be such that the product of a³parallel´ pair of stress and strain is preserved under coordinate transformations from one to another orthonormal base (hence the ¥2). In this way Kelvin had defined a 6D Cartesian vector space The Structure of Tensors 14 . 5. 3. 4. because neither did exist then.

In a base consisting of (orthogonal) eigenstrains. with the ³eigenstiffnesses´ the only non-zero components. The Structure of Tensors 15 . Kelvin thus had produced the ³eigenvalue decomposition´ ± or the ³canonical representation of the stiffness tensor by its invariants´. The ideal base strain generates a parallel base stress (of the same type). 7. the 6v6 representation of the stiffness tensor is diagonal. As examples he gave for isotropic media ³hydrostatic pressure´ -> ³uniform volume compression´ and ³shear strain´ -> ³shear stress´. Kelvin called such strains ³principle strains´. we call them ³eigenstrains´.Kelvin went on to the invariant description Hooke¶s law can be thought of as a linear mapping of the strain space on the stress space The map is described by the 6v6 stiffness matrix 6.

Eigensystem of isotropic media The Structure of Tensors 16 .

The Structure of Tensors 17 . The elements of these representations are called eigenvalues and eigenvectors. The nature of the 6 eigenstrains is of great importance for many tasks. The resulting items are called eigencompliances and eigenstresses. If these details are important. but not easily seen in an arbitrarily oriented coordinate system.Nomenclature I have called the eigenvalues eigenstiffnesses. Why these special terms? A similar eigenvalue problem exists for the inverse of Hooke¶s Law. and the eigenvectors eigenstrains. the eigenstrain vectors are treated as 3v3 tensors and brought into invariant (canonical) form.

The Structure of Tensors 18 . solution has to be programmed Kelvin notation: Wp = Cpq Iq = 0Ip => (cpq .Hpq0) Iq= 0 Standard eigenvalue/eigenvector problem.How does eigenvalue decomposition work in the different notations? The ³eigenvalue decomposition´ can be formulated as: find a stress that is generated by a parallel strain Four-subscripts: Wij = cijkl Ikl=0Iij =>(cijkl ± HikHjl 0) Ikl = 0 Meaningful problem. since stresses and strains are expressed in terms of different bases. routines for solution in every math-package Voigt notation: the problem cannot be formulated easily.

Only the eigenstiffnesses can assume complex values.What does the invariant description mean? Of the 21 parameters of a general elastic tensor. The Structure of Tensors 19 . only the six eigenstiffnesses are genuinely elastic The remaining 15 parameters are geometric: the 15 free parameters that describe a set of six mutually perpendicular 6D unit vectors Of these 15. three are extraneous to the problem: they are the ³Euler angles´ that describe the orientation of the material with respect to the global coordinate system Conjecture: All geometric parameters must be real.

In modeling a dissipative medium one cannot freely assign imaginary parts to the stiffnesses. If we would do that. i. a ³eigenvalue composition´. Obviously the medium is stable if all six eigenstiffnesses are positive. How do we assign imaginary parts to the 21 stiffnesses? We need a method to construct tensors from their eigensystem.e. The Structure of Tensors 20 . we might end up with complex eigenstrains.How does all this affect our work? The ³invariant description´ is possible for every medium.. in violation of the concept of a ³strain type´.

then C = E .ET is a symmetric. The Structure of Tensors 21 .Eigenvalue composition of a tensor If E is a 6v6 orthonormal matrix and L is a diagonal matrix with 6 positive elements.L. we can generate a stiffness tensor C with an arbitrarily chosen eigensystem. positive definite. 6v6 matrix with the elements of L as eigenvalues and the column vectors of E as eigenvectors If we choose the intended eigenstrains as column vectors of E and place the intended eigenstiffnesses on the diagonal of L.

add imaginary parts (of the correct sign) to the eigenstiffnesses. The Structure of Tensors 22 . An ³educated´ choice requires that we know more about the nature of the eigenstrains and the way they influence (together with the eigenstiffnesses) the stiffness tensor. and re-compose. the symmetry of the stiffness tensor ± and some other properties like the ³order´ of elastic waves ± is controlled by the chosen eigensystem.How should the eigensystem be chosen? Obviously. Such a study is not necessary if we just want to make an existing tensor ³dissipative´: we just have to determine the eigensystem.

Are these data consistent with our assumptions? The Structure of Tensors 23 .01i 4.00-.03i 3.36-.03i 2.84-.16i 2. the imaginary part came from a different source.00-.16i 5.01i Kelvin form The real part was ³invented´ in 1991.Example: a dissipative orthotropic medium The following complex-valued stiffness tensor C was used: 9.01i 3.94-1.3i 4.25-.20-.60-.02i 9.40-.

36-.99 i -.93 0 0 0 4.024i .737 -.015 i .37 i .359-.68-. but so have two components of each of three eigenstrains.073i -. What does that mean? Moreover.29-.26+.20-0. this matrix is not orthonormal ! The Structure of Tensors 24 .622-.67+.00-.80-.24+0.066i 0 0 0 5.011 i 0 0 0 1 0 0 3.033-.002i .693 .011i 0 0 0 0 0 1 4.037i 0 0 0 4.02i .11 i 0 0 0 0 1 0 All eigenstiffnesses have a small imaginary part.The eigensystem of the tensor C 14.

The ³shear´eigenstiffnesses are equal to the corresponding shear stiffnesses (in Kelvin form) The Structure of Tensors 25 .16i 2. but now everything is consistent.00-.90-.22i 3.01i 3.13i 5.78-.01i 4.01i Kelvin form The changes are moderate.20-.22i 9.00-.58-.37-.17-.The improved dissipative tensor C¶ The real part of the eigenstrains together with the complex eigenstiffnesses result in: 9.36-. Note: the shear stiffnesses have not changed at all.90i 4.24i 2.

Anatomy of the strain tensor Since the appearance of the eigenstrains depends on their orientation with respect to the global coordinate system (three Euler angles!). Any (unit) strain tensor can be part of the eigensystem of a stable stiffness tensor Two invariants of a strain tensor are the trace and the determinant Strains with vanishing trace are called ³isochoric´ Strains with vanishing determinant are called ³wave-compatible´ The Structure of Tensors 26 . we have to look at the invariant representation of strains.

Some homogeneous strains can exist along a plane. The Structure of Tensors 27 . Homogeneous strain in unbounded media cannot exist. Consider a homogeneous strain in an unbounded medium.Wave compatibility? The term ³wave compatible´ sounds oddly out of place in a discussion that so far concerned only static aspects. Even an infinitesimal strain can lead to very large displacements at large distance from the reference point.

³Wave compatibility´ explained Ultimately we are interested in plane waves. The Structure of Tensors 28 . Any displacement can be attached to a plane wave. Such strains are ³compatible´ with wave propagation perpendicular to the plane. Hence their name. but only those strains that can exist along a plane without creating infinite displacements.

and 23-planes would be enforced The Structure of Tensors 29 . and a shear strain in the 13.Strains that are not ³wave compatible´ These two strains could not travel as a plane wave in 3-direction: with distance from the center the displacement would grow without limit.

Three strains that are ³wave compatible´ These three strains could travel as a plane wave in 3-direction without generating locally large displacements The plane shear strain (lowest example) could travel not only in 3-direction. but also in 1-direction The Structure of Tensors 30 .

All strains that are ³wave compatible´in a given coordinate direction Six mutually orthogonal unit strains Note that each shear strain could travel in two directions! The Structure of Tensors 31 .

All strains (as 6D-vectors) that are ³wave compatible´in a given coordinate direction Six mutually orthogonal unit strains Note that each shear strain could travel in two directions! Of the two shear strains that can travel in 3-direction. the other also in 2-direction. etc. one can also travel in 1-direction. The Structure of Tensors 32 .

but for an eigenstrain this is irrelevant.Symmetry planes and shear eigenstrains The symmetry of a medium is controlled by its symmetry planes. for elements with an even number of subscripts ³3´. two of them plane shear strains in the 23. for all but the two shear strains. i. It has the form The 12-plane is a symmetry plane if the set is invariant under a change of sign of the 3axis.e.planes.e.. The Structure of Tensors 33 . Consider a set of six orthonormal eigenstrains. i. are arbitrary entries The two shear eigenstrains change sign.and 13. Proposition: Two shear eigenstrains that share an axis define a symmetry plane perpendicular to this axis.

Symmetry planes and shear waves The symmetry of a medium is controlled by its symmetry planes A symmetry plane supports shear wave with cross-plane polar. Conversely. a plane that does support these waves is a symmetry plane Two shear eigenstrains that share an axis thus define a symmetry plane perpendicular to this axis This holds for any two shear eigenstrains The Structure of Tensors 34 .ization in all direction.

. we can choose the shear eigenstrains to let the symmetry planes coincide with the coordinate planes The Structure of Tensors 35 . but its orientation with respect to the ³default´ system (e. These are the elements of the ³equivalence set´. If we construct a tensor with symmetry planes.The equivalence set We had found that three of the 21 parameters describing an elastic tensor are ³extraneous´: they describe not the tensor. N±E±down) For a tensors without any symmetry plane that means that there is a three-parametric manifold of ³equivalent´ tensors that differ only in orientation.g.

The 1. Let the normal to this plane be the 3-axis. which have only one symmetry plane (but also with tetragonal tensors. In a different (crystallographic) coordinate system.Equivalence for a single symmetry plane Two shear eigentensors define a single plane of symmetry. trigo & tetra-> 6->7).and 2-axis can have still any orientation in the symmetry plane. The Structure of Tensors 36 . This problem occurs with monoclinic and trigonal tensors. We choose the orientation which gives the lowest number of stiffnesses. thus there is a 1-parametric manifold of equivalent tensors. As far as elasticity is concerned. where the plane orthogonal to the fourfold axis is regarded as the symmetry plane). this is always possible. the number is higher (mono:12->13.

B. This orthonormal eigensystem leads to this Voigt stiffness matrix with 16 independent stiffnesses The A.Effect of one single shear eigenstrain A single shear eigenstrain has no effect on symmetry. but a strong effect on the stiffness matrix.«F are arbitrary positive numbers The black disks represent real numbers The tensor is stable ³by design´ The Structure of Tensors 37 .

The Structure of Tensors 38 . thus the tensor has ³monoclinic´ symmetry.Two shear eigenstrains -> monoclinic Two shear eigenstrains generate a plane of symmetry. This orthonormal eigensystem leads to this Voigt stiffness matrix with twelve independent stiffnesses The matrix on the right is automatically in the coordinate system by the two shear planes and the symmetry planes and therefore has twelve ± not thirteen ± stiffnesses.

The Structure of Tensors 39 . From now on. the ³shape´ of the matrix has reached its final form.Three shear eigenstrains ±> orthotropic Three shear eigenstrains generate three planes of symmetry. Also. This orthonormal eigensystem leads to this Voigt stiffness matrix with nine independent stiffnesses Orthotropy marks an important point: up to now we had just to add new planes to increase the symmetry. thus the tensor has ³orthotropic´ symmetry. co-planar shear tensors and identical eigenstiffnesses will be needed to increase symmetry.

The Structure of Tensors 40 . Thus the eigenstiffnesses must be identical. On the other hand.Example: Transverse Isotropy TI is the anisotropy best known. since the two other shear planes are symmetry planes too. the symmetry is high enough to give a good example of how additional symmetry planes depend on the eigensystem. This makes the 3-axis at least a 4-fold axis. The eigenstiffnesses control the velocities of the waves with the appropriate in-plane shears along the three red lines. In addition we certainly have a shear eigenstrain in the 12-plane. The equatorial plane must be a symmetry plane.

I determined with Mathematica the eigensystem of a general TI medium.The eigensystem of Transverse Isotropy To simplify the discussion. I obtained for 1±4 Estff 2c55 11 0 22 0 33 0 23 0 13 1 12 0 2c55 0 0 0 1 0 0 2c66 0 0 0 0 0 1 2c66 ±1 1 0 0 0 0 41 The Structure of Tensors .

we need C.c66)/2.The eigensystem of TI: #5 and #6 p = c11 + (c33 .D = ±2 A.B The Structure of Tensors 42 . To be orthogonal to each other. q = ¥[(p-c33)2 + 2c132]/2 Estff p ± q A 11 A 22 C 33 0 23 0 13 0 12 p+q B B D 0 0 0 Each of the two tensors is orthogonal to the other four.

orthotropic. asterisk: d* isochoric strain The Structure of Tensors 43 . and trigonal symmetry Capital letters: E plane shear.Overview of Eigensystems I Eigensystems of monoclinic.

cubic. TI.Overview of Eigensystems II Eigensystems of tetragonal. and isotropic symmetry Capital letters: E plane shear. asterisk: d* isochoric strain Underlined: coplanar shear tensors The Structure of Tensors 44 .

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd