Jump to content

Continuum mechanics/Spectral decomposition

From Wikiversity

Spectral decompositions

[edit | edit source]

Many numerical algorithms use spectral decompositions to compute material behavior.

Spectral decompositions of stretch tensors

[edit | edit source]

Infinitesimal line segments in the material and spatial configurations are related by

d๐ฑ=๐‘ญโ‹…d๐‘ฟ=๐‘นโ‹…(๐‘ผโ‹…d๐‘ฟ)=๐‘ฝโ‹…(๐‘นโ‹…d๐‘ฟ) .

So the sequence of operations may be either considered as a stretch of in the material configuration followed by a rotation or a rotation followed by a stretch.

Also note that

๐‘ฝ=๐‘นโ‹…๐‘ผโ‹…๐‘นT .

Let the spectral decomposition of ๐‘ผ be

๐‘ผ=โˆ‘i=13ฮปi ๐‘ตiโŠ—๐‘ตi

and the spectral decomposition of ๐‘ฝ be

๐‘ฝ=โˆ‘i=13ฮปฬ‚i ๐งiโŠ—๐งi .

Then

๐‘ฝ=๐‘นโ‹…๐‘ผโ‹…๐‘นT=โˆ‘i=13ฮปi ๐‘นโ‹…(๐‘ตiโŠ—๐‘ตi)โ‹…๐‘นT=โˆ‘i=13ฮปi (๐‘นโ‹…๐‘ตi)โŠ—(๐‘นโ‹…๐‘ตi)

Therefore the uniqueness of the spectral decomposition implies that

ฮปi=ฮปฬ‚iand๐งi=๐‘นโ‹…๐‘ตi

The left stretch (๐‘ฝ) is also called the spatial stretch tensor while the right stretch (๐‘ผ) is called the material stretch tensor.

Spectral decompositions of deformation gradient

[edit | edit source]

The deformation gradient is given by

๐‘ญ=๐‘นโ‹…๐‘ผ

In terms of the spectral decomposition of ๐‘ผ we have

๐‘ญ=โˆ‘i=13ฮปi ๐‘นโ‹…(๐‘ตiโŠ—๐‘ตi)=โˆ‘i=13ฮปi (๐‘นโ‹…๐‘ตi)โŠ—๐‘ตi=โˆ‘i=13ฮปi ๐งiโŠ—๐‘ตi

Therefore the spectral decomposition of ๐‘ญ can be written as

๐‘ญ=โˆ‘i=13ฮปi ๐งiโŠ—๐‘ตi

Let us now see what effect the deformation gradient has when it is applied to the eigenvector ๐‘ตi.

We have

๐‘ญโ‹…๐‘ตi=๐‘นโ‹…๐‘ผโ‹…๐‘ตi=๐‘นโ‹…(โˆ‘j=13ฮปj ๐‘ตjโŠ—๐‘ตj)โ‹…๐‘ตi

From the definition of the dyadic product

(๐ฎโŠ—๐ฏ)โ‹…๐ฐ=(๐ฐโ‹…๐ฏ) ๐ฎ

Since the eigenvectors are orthonormal, we have

(๐‘ตjโŠ—๐‘ตj)โ‹…๐‘ตi={0if iโ‰ j๐‘ตiif i=j

Therefore,

(โˆ‘j=13ฮปj ๐‘ตjโŠ—๐‘ตj)โ‹…๐‘ตi=ฮปi ๐‘ตino sum on i

That leads to

๐‘ญโ‹…๐‘ตi=ฮปi (๐‘นโ‹…๐‘ตi)=ฮปi ๐งi

So the effect of ๐‘ญ on ๐‘ตi is to stretch the vector by ฮปi and to rotate it to the new orientation ๐งi.

We can also show that

๐‘ญโˆ’Tโ‹…๐‘ตi=1ฮปi ๐งi ;  ๐‘ญTโ‹…๐งi=ฮปi ๐‘ตi ;  ๐‘ญโˆ’1โ‹…๐งi=1ฮปi ๐‘ตi

Spectral decompositions of strains

[edit | edit source]

Recall that the Lagrangian Green strain and its Eulerian counterpart are defined as

๐‘ฌ=12 (๐‘ญTโ‹…๐‘ญโˆ’1) ;  ๐’†=12 (1โˆ’(๐‘ญโ‹…๐‘ญT)โˆ’1)

Now,

๐‘ญTโ‹…๐‘ญ=๐‘ผโ‹…๐‘นTโ‹…๐‘นโ‹…๐‘ผ=๐‘ผ2 ;  ๐‘ญโ‹…๐‘ญT=๐‘ฝโ‹…๐‘นโ‹…๐‘นTโ‹…๐‘ฝ=๐‘ฝ2

Therefore we can write

๐‘ฌ=12 (๐‘ผ2โˆ’1) ;  ๐’†=12 (1โˆ’๐‘ฝโˆ’2)

Hence the spectral decompositions of these strain tensors are

๐‘ฌ=โˆ‘i=1312(ฮปi2โˆ’1) ๐‘ตiโŠ—๐‘ตi ;  ๐ž=โˆ‘i=1312(1โˆ’1ฮปi2) ๐งiโŠ—๐งi

Generalized strain measures

[edit | edit source]

We can generalize these strain measures by defining strains as

๐‘ฌ(n)=1n (๐‘ผnโˆ’1) ;  ๐’†(n)=1n (1โˆ’๐‘ฝโˆ’n)

The spectral decomposition is

๐‘ฌ(n)=โˆ‘i=131n(ฮปinโˆ’1) ๐‘ตiโŠ—๐‘ตi ;  ๐ž(n)=โˆ‘i=131n(1โˆ’1ฮปin) ๐งiโŠ—๐งi

Clearly, the usual Green strains are obtained when n=2.

Logarithmic strain measure

[edit | edit source]

A strain measure that is commonly used is the logarithmic strain measure. This strain measure is obtained when we have nโ†’0. Thus

๐‘ฌ(0)=ln⁡(๐‘ผ) ;  ๐’†(0)=ln⁡(๐‘ฝ)

The spectral decomposition is

๐‘ฌ(0)=โˆ‘i=13ln⁡ฮปi ๐‘ตiโŠ—๐‘ตi ;  ๐ž(0)=โˆ‘i=13ln⁡ฮปi ๐งiโŠ—๐งi