7

Review of
Continuum Mechanics:
Kinematics

7–1
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

TABLE OF CONTENTS
Page
§7.1 Introduction . . . . . . . . . . . . . . . . . . . . . 7–3
§7.2 Notational Systems . . . . . . . . . . . . . . . . . . 7–3
§7.3 The Continuum Model . . . . . . . . . . . . . . . . . 7–4
§7.3.1 Particle Motion . . . . . . . . . . . . . . . . 7–5
§7.3.2 Configurations . . . . . . . . . . . . . . . . . 7–5
§7.3.3 Distinguished Configurations . . . . . . . . . . . 7–5
§7.3.4 Kinematic Descriptions . . . . . . . . . . . . . . 7–7
§7.3.5 Coordinate Systems . . . . . . . . . . . . . . . 7–8
§7.3.6 Finite Element Frames . . . . . . . . . . . . . . 7–9
§7.3.7 Configurations and Staged Analysis . . . . . . . . . 7–11
§7.4 Deformation Measures . . . . . . . . . . . . . . . . . 7–11
§7.4.1 Deformation Gradients . . . . . . . . . . . . . . 7–12
§7.4.2 Displacement Gradients . . . . . . . . . . . . . . 7–12
§7.4.3 Stretch Tensors . . . . . . . . . . . . . . . . 7–13
§7.5 *Useful Vector and Matrix Formulas . . . . . . . . . . . . 7–17
§7.5.1 *Vector Identities . . . . . . . . . . . . . . . 7–17
§7.5.2 *Integral Identities . . . . . . . . . . . . . . . 7–17
§7.5.3 *Quick Refresher on Matrix Calculus . . . . . . . . . 7–18
§7. Notes and Bibliography . . . . . . . . . . . . . . . . . 7–19
§7. Exercises . . . . . . . . . . . . . . . . . . . . . . 7–20

7–2
 §7.2 NOTATIONAL SYSTEMS

§7.1. Introduction

Chapters 3 through 6 have covered general properties of the governing force residual equations of
geometrically nonlinear structural systems with finite number of degrees of freedom (DOF). The
DOFs are collected in the state vector, and driven by control parameters. The residual equations,
being algebraic, are well suited for numerical computations.
Continuum models of actual structures, however, are expressed as ordinary or partial differential
equations in space or space-time. As such, those models possess an infinite number of DOFs. Except
for simple (typically linear) models, they cannot be directly solved analytically. The reduction to
a finite number is accomplished by discretization methods. It was observed in Chapter 1 that for
nonlinear problems in solid and structural mechanics the finite element method (FEM) is the most
widely used discretization method.
This Chapter and the next one provide theoretical background for deriving geometrically nonlinear
finite elements from continuum models. Collectively they give an overview of kinematic, kinetic
and constitutive relations for a three-dimensional elastic deformable body, as needed in subsequent
Chapters. Readers familiar with continuum mechanics may skim this and the next Chapter to get
familiar with the notation.

§7.2. Notational Systems

Continuum mechanics deals with vector and tensor fields such as displacements, strains and stresses.
Four notational systems are in common use.
Indicial Notation. Also called component notation. The key concept is that of an index. Indices
identify components of vectors and tensors. It has convenient abbreviation rules, such as commas
for partial derivatives and Einstein’s summation convention. The notation is general and powerful,
and as such is preferred in analytic developments as well as publication in theory journals and
monographs. It readily handles arbitrary tensors of any order, curvilinear coordinate systems and
nonlinear expressions. When used in non-Cartesian coordinates, it sharply distinguishes between
covariant and contravariant quantities. The main disadvantages are: (i) physics is concealed behind
the index jungle, and (ii) highly inefficient for expressing numeric computations. Because of (i), it
is not suitable for first-level (undergraduate) instruction.
Direct Notation. Sometimes called algebraic notation. Vectors and tensors are represented by
single symbols, usually bold letters. These are linked by the well known operators of mathematical
physics, such as . for dot product, × for cross products, and ∇ for gradient (or divergence). Has
the advantage of compactness and quick visualization of intrinsic properties. Certain operations,
however, become undefined beyond a certain range. Some of them overlap with matrix algebra
while others do not. This fuzzyness can lead to confusion in computational work.
Matrix Notation. This is similar to the previous one, but entities are appropriately recast so that
only matrix operations are used. This form can be directly mapped to discrete equations as well
as high-level, matrix-oriented programming languages such as Matlab. It has the disadvantage of
losing contact with the original physical entities along the way. For example, stress is a symmetric
second-order tensor, but is recast as a 6-vector for FEM developments. This “covnience” change
may forgo essential properties. For instance, it makes sense to say that principal stresses are

7–3
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

eigenvalues of the stress tensor. But the statement becomes nonsense when stresses are recast as
vectors.1
Full Notation. In full form notation every term is spelled out. No ambiguities of interpretation can
arise. This works well as a notation of last resort, and also as a “comparison template” against which
one can check out the meaning of more compact expressions. It is also useful for programming in
low-order languages.
In this and following Chapters the direct, matrix and full notations are preferred. The indicial
notation is used if either complicated tensor forms are needed, or nonlinear expressions not amenable
to other notations appear. Often the expression is first given in direct form and confirmed by full form
if feasible. It is then mapped to matrix notation for use in FEM developments. The decision chain
leads to possible ambiguities against reuse of vector symbols in two contexts: continuum mechanics
and FEM discretizations. Such ambiguities are resolved in favor of keeping FEM notation simple.
Example 7.1. Consider the well known dot product between two physical vectors in 3D space, a = (a1 , a2 , a3 )
and b = (b1 , b2 , b3 ) written in the four different notations:
a b = 
a.b = 
aT b = a1 b1 + a2 b2 + a3 b3 . (7.1)

i i
  
indicial direct matrix full

Example 7.2. Take the internal static equilibrium equations of a continuum body, expressed in terms of Cauchy
stresses and body forces per unit volume:
∂σ11 ∂σ12 ∂σ13
σi j, j + bi = 0,
   ∇σ +b = 0, DT σv + b = 0,
   ∂ x1
+
∂ x2
+
∂ x3
+ b1 = 0, plus 2 more. (7.2)
indicial direct matrix   
full
In the matrix version, σv denotes the stress tensor cast as a 6-vector (in the sequel this is simply called σ).

Example 7.3. The strain-to-stress constitutive equations of a linearly elastic solid, writen in a RCC frame and
excluding initial strain effects, can be written as the generalized Hooke’s law
σi j = E i jk ek varies σ = E e σ11 = E 1111 e11 + E 1112 e12 + . . . 9 equations, 81 terms. (7.3)
           
indicial direct matrix full
Here the indicial and matrix forms are compact. The latter, however, requires that stresses σi j and strains ek
be cast as 6-vectors while the elastic moduli tensor E i jk is cast to a 6 × 6 matrix. The direct form requires
the introduction of an ad-hoc operator that effects the product of a 4-dimensional tensor times a second order
one; this varies from author to author. The full form will take up at least half a page.

Example 7.4. If a discrete mechanical system is conservative, it was shown in Chapter 6 that the total force
residual is the gradient of a total potential energy function with respect to the state:
∂
def ∂
∂
∂

ri = =
,i , r =∇
, r= , r1 = , r2 = ,... (7.4)
∂u i  ∂u ∂u 1 ∂u 2
   direct   
indicial matrix full
The indicial form requires defining
,i as abbreviation of partial derivative with respect to u i rather than xi .
The direct form depends on the gradient operator symbol chosen. For example, r = grad
is also used.
1 Stilted authors make a big deal about distinguishing between tensors, which are entities independent of any coordinate
system, and their matrix and indicial representations, which are not. This is largely an exercise in peripatetic angelology,
á la “How many angels can dance on the head of a pin?” Such scholastic preoccupations are eschewed here.

7–4
 §7.3 THE CONTINUUM MODEL

§7.3. The Continuum Model

In this and the next Chapter the structure is mathematically treated as a continuum body B. In this
model, the body is considered as being formed by a set of points P called particles, which are
endowed with certain mechanical properties. For FEM analysis the body is divided into elements,
which inherit the properties of the continuum model.

§7.3.1. Particle Motion

Particles displace or move in response to external actions, such as forces, characterized by the
control parameters i introduced in Chapter 3. Following the reduction process discussed there,
in each stage the body responds to the single stage parameter λ.
A one-parameter series of positions occupied by the particles as they move in space is called a
motion. The motion may be described by the displacement u(P) ≡ u(x) of the particles2 with
respect to a base or reference state in which particle P is labeled P0 . The displacements of all
particles u(x) such that x ≡ {x, y, z} ∈ B, constitutes the displacement field.
The motion is said to be kinematically admissible if:
1. Continuity of particle positions is preserved so that no gaps or interpenetration occurs.3
2. Kinematic constraints on the motion (for example, support conditions) are preserved.
A kinematically admissible motion along a stage will be called a stage motion. For one such motion
the displacements u(x) characterize the state and the stage control parameter λ characterizes the
control. Both will be generally parametrized by the pseudo-time t introduced in Chapter 3. Thus a
stage motion can be generally represented by

λ = λ(t), u = u(x, t), x ∈ B. (7.5)

§7.3.2. Configurations

If in (7.5) we freeze t, we have a configuration of the structure. Thus a configuration is formally the
union of state and control. It may be informally viewed as a “snapshot” taken of the structure and its
environment when the pseudotime is frozen. If the configuration satisfies the equilibrium equations,
it is called an equilibrium configuration. In general, however, a randomly given configuration is
not in equilibrium unless artificial body and surface forces are applied to it.
A staged response, or simply response, can be now mathematically defined as a series of equilibrium
configurations obtained as λ is continuously varied, starting from zero.

2 The underlining in u is used to distinguish the physical (particle level) displacement vector from the finite element node
displacement array, which is a computational vector.
3 The mathematical statement of these compatibility conditions is quite complicated for finite displacements and will not be
given here. See for example, [828, §34]. The displacement-based finite element formulations worked out in subsequent
Chapters will automatically satisfy the requirement.

7–5
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

Name Alias Definition Equilibrium Identification

Required?

Admissible A kinematically admissible configuration No

Perturbed Kinematically admissible variation No δ

of an admissible configuration

Current Deformed Any admissible configuration taken during the No or

Spatial analysis process. Contains all others as special cases

Base Initial The configuration defined as the origin of Yes 0, B or

Undeformed displacements. Strain free but not necessarily stress free
Material

Reference Configuration to which stepping computations TL,UL: yes TL: 0 UL: n−1
in an incremental solution process are referred CR: R no, 0 yes CR: R and 0

Iterated Configuration taken at the kth iteration No n,k

of the nth increment step

Target Equilibrium configuration accepted Yes n

after completing the nth increment step

Corotated Shadow Body- or element-attached configuration obtained No R

Ghost from through a RBM (CR description only)

Aligned Preferred A fictitious body ot element configuration aligned No A

Directed with a particular set of axes (usually global axes)

Definitions with blue background are used only in theoretical and applied mechanics for analytical formulations.
Definitions with yellow background are only used in computational mechanics.
Definitions with green background are used in both, hence the color choice. The meaning of reference configuration,
however, may differ. The one stated above is for computational mechanics.

The base configuration 0 is often the same as the natural state in which body (or element) is undeformed and stress free.

In dynamic analysis using the CR kinematic description, 0 and R are called the inertial and dynamic reference
configurations, respectively, when applied to an entire structure such as an airplane (e.g., autopilot simulations)

Figure 7.1. Distinguished configurations in geometrically nonlinear analysis.

§7.3.3. Distinguished Configurations

A distinctive feature of geometrically nonlinear analysis is the need to carefully distinguish among
different configurations of the structure. As defined above, a set of kinematically admissible
displacements u(x) plus a staged control parameter λ at a frozen t defines a configuration.
This is not necessarily an equilibrium configuration. In fact it will not usually be one. It is important
to realize that an equilibrium configuration is not necessarily a physical configuration assumed by
the actual structure.4 Configurations that are important in geometrically nonlinear analysis receive
special qualifiers:
admissible, perturbed, deformed, base, reference, iterated, target, corotated, aligned

This terminology is collected in Figure 7.1 in tabular format. Of the nine instances listed there,

4 Recall the suspension bridge under zero gravity of Chapter 3.

7–6
 §7.3 THE CONTINUUM MODEL

Current Configuration
or

Base Configuration
B or

Reference Configuration
(identifier depends on
kinematic description chosen)

Figure 7.2. The three most important configurations for geometrically nonlinear analysis.
Specialization to the kinematic descriptions tabulated in Figure 7.3 are pictured in Figures 7.4, 7.5,
and 7.6.

the first two are used extensively in theoretical and applied mechanics, the last four exclusively in
computational mechanics, and three (current, base, and reference) used in both, although sometimes
with different meanings.
The three most important configurations insofar as a FEM implementation is concerned, are: base,
reference and current. These are pictured in Figure 7.2. An examination of this figure indicates that
the choice of the reference configuration depends primarily on the kinematic description chosen.
Such a choice is examined in the next subsection.
Remark 7.1. Many names can be found for the configurations pictured in Figure 7.1 in the literature dealing
with finite elements and continuum mechanics. Here are some of those alternative names.
Perturbed configuration: adjacent, deviated, disturbed, incremented, neighboring, varied, virtual.
Current configuration: deformed, distorted, pull-forward, moving, present, space, spatial, varying.
Base configuration: baseline, body, initial, global, laboratory, material, natural, original, overall,
˙ undeformed, undistorted.
Reference configuration: fixed, frozen, known, pull-back.
Iterated configuration: corrected, intermediate, stepped, transient, transitory.
Target configuration: converged, equilibrated spatial, unknown.
Corotated configuration: attached, convected, ghost, phantom, shadow.
Aligned configuration: directed, body-matched, body-fitted, preferred.

§7.3.4. Kinematic Descriptions

Three kinematic descriptions of geometrically nonlinear finite element analysis are in current use in
programs that solve nonlinear structural problems: Total Lagrangian or TL, Updated Lagrangian
or UL, and Corotational or CR. They are described in Figure 7.3 They can be distinguished by the
choice of reference configuration. The important configurations for these three descriptions are
pictured in Figures 7.4, 7.5, and 7.6, respectively.
The TL formulation remains the most widely used in continuum-based FEM codes.5 The CR

5 A key reason is historic: three of the original nonlinear FEM codes: MARC, ABAQUS and ANSYS, originally imple-
mented elements based on that description.

7–7
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

Name Acronym Definition Primary applications

Total Lagrangian TL Base and reference configurations Solid and structural mechanics with finite
coalesce and remain fixed throughout but moderate displacements and strains.
the solution process Primarily used for elastic materials. Unreliable
for flow-like behavior or topology changes

Updated Lagrangian UL Base configuration remains fixed but Solid and structural mechanics with finite
reference configuration is periodically. displacements and possibly large strains.
updated. Common update strategies: Handles material flow-like behavior well,
either set reference configuration to current (e.g., forming processes) as well as topology
solution, or to last converged solution changes (fracture)

Corotational CR Reference configuration is split into base Solid and structural mechanics with arbitrarily
and corotated. Strains and stresses are large finite motions, but small strains and
measured from corotated to current, while elastic material behavior. Extendible to nonlinear
base configuration is maintained as materials if inelasticity is localized so most of
reference to measure rigid body motions structure stays elastic.

TL and CR are Lagrangian: computations are always referred to a previous configuration (base and/or reference). UL is
either Eulerian or a Lagrangian-Eulerian hybrid. Eulerian descriptions are popular in fluid mechanics, but less so in solid
and structural mechanics unless the behavior is fluid-like (e.g., metal forming, nonlinear viscoelasticity)

Figure 7.3. Kinematic descriptions used in FEM programs that handle geometrically
nonlinear problems in solid and structural mechanics.

formulation is gaining in popularity for structural elements such as beams, plates and shells, espe-
cially in Aerospace.6 The UL formulation is useful in treatments of vary large strains and flow-like
behavior, as well as in processes involving topology and/or phase changes; e.g., metal forming.
§7.3.5. Coordinate Systems
Configurations taken by a body or element during the response analysis are linked by a Cartesian
global frame, to which all computations are ultimately referred.7 There are actually two such
frames:
(i) The material global frame with axes {X i } or {X, Y, Z }.
(ii) The spatial global frame with axes {xi } or {x, y, z}.8
The material frame tracks the base configuration whereas the spatial frame tracks all others. This
distinction agrees with the usual conventions of classical continuum mechanics. In this book both
frames are taken to be identical, as nothing is gained by separating them. Thus only one set of
global axes, with dual labels, is drawn in Figure 7.7

6 Geometrically nonlinear problems in Aerospace Engineering tend to involve large motions, in particular large rotations,
but small strains. Reason: structures are comparatively thin to save weight.
7 In dynamic analysis the global frame may be moving in time as a Galilean or inertial frame. This is convenient to track
the trajectory motion of objects such as aircraft or satellites.
8 The choice between {X 1 , X 2 , X 3 } versus {X, Y, Z } and likewise {x1 , x2 , x3 } versus {x, y, z} is a matter of notational
convenience. For example, when developing specific finite elements it is preferable to use {X, Y, Z } or {x, y, z} so as
to reserve coordinate subscripts for node numbers. On the other hand, in derivations that make heavy use of indicial
notation, the X i and xi notation is more appropriate.

7–8
 §7.3 THE CONTINUUM MODEL

Current Configuration

Base and Reference

Configuration
0=

TOTAL LAGRANGIAN (TL)

Kinematic Description

Figure 7.4. Important configurations in Total Lagrangian (TL) kinematic description.

Current Configuration

Base Configuration
B

UPDATED LAGRANGIAN (UL)

Kinematic Description Reference Configuration n
either reset to current after each incremental step,
or continuously reset to current after each iteration

Figure 7.5. Important configurations in Updated Lagrangian (UL) kinematic description.

Current Configuration

Base Configuration 0

COROTATIONAL (CR) Corotated Configuration R

Kinematic Description a rigid motion of the
base configuration

Figure 7.6. Important configurations in Corotational (CR) kinematic description. The Note:
the corotated and current configuration are shown highly offset for visualization convenience.
In practical use they highly overlap; for example, the centroids coincide.

7–9
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

Current configuration
P(x)

x = x(t)
Z, z

u=x−X
V
X, x Y, y X ≡ x0 P0 (X)
Base configuration (for
drawing simplicity, assumed to
coalesce with reference, as in TL)
Figure 7.7. The geometrically nonlinear problem in a Lagrangian kinematics: coordinate systems, reference
and current configurations, and displacements. To keep the figure simple it has been assumed that base and
reference configurations coalesce, as happens in the Total Lagrangian (TL) description.

§7.3.6. Finite Element Frames

In stark contrast to the uniqueness of global frames, the presence of finite elements means there
are many local frames to keep track of. More precisely, each element is endowed with two local
Cartesian frames:
(iii) The element base frame with axes { X̃ i } or { X̃ , Ỹ , Z̃ }.
(iv) The element reference frame with axes {x̄i } or {x̄, ȳ, z̄}.
The base frame is attached to the base configuration. It remains fixed if the base is fixed. It is
chosen according to usual FEM practices. For example, in a 2-node spatial beam element, X̃ 1 is
defined by the two end nodes whereas X̃ 2 and X̃ 3 lie along principal inertia directions. The origin
is typically placed at the element centroid.
The meaning of the reference frame depends on the description chosen:
Total Lagrangian (TL). The reference and base frames coalesce, whence the X i and xi coalesce.
Updated Lagrangian (UL). This is described differently by different authors. Some specify that
the reference configuration is set to be the current configuration, so it is effectively updated in each
increment or iteration. Others authors relax this policy and specify that the reference configuration
is updated once the process converges.
Corotational description (CR). The reference frame is renamed corotated frame or CR frame. It
remains attached to the element and continuously moves with it.
The transformation
x = X + u, (7.6)
maps the position of base particle P(X, Y, Z ) to P(x, y, z). See Figure 7.7. Consequently the
particle displacement vector is defined as
   
uX x−X
u= uY = y−Y = x − X. (7.7)
uZ z−Z

7–10
 §7.4 DEFORMATION MEASURES

in which (X, Y, Z ) and (x, y, z) pertain to the same particle.

Remark 7.2. Variations of this notation scheme are employed as appropriate to the subject under consideration.
For example, the coordinates of P in a target configuration Cn may be called (xn , yn , z n ).

Remark 7.3. In continuum mechanics, (X, Y, Z ) and (x, y, z) are called material and spatial coordinates,
respectively. In general treatments both systems are curvilinear and need not coalesce. The foregoing relations
are restrictive in two ways: the base coordinate systems for the reference and current configurations coincide,
and that system is Cartesian. These assumptions are sufficient, however, for the problems addressed here.

Remark 7.4. The dual notation (X, Y, Z ) ≡ (x 0 , y0 , z 0 ) is introduced on two accounts: (1) the use of
(x0 , y0 , z 0 ) sometimes introduces a profusion of additional subscripts, and (2) the notation agrees with that
traditionally adopted in continuum mechanics for the material coordinates, as noted in the previous remark.
The identification X ≡ x0 , Y ≡ y0 , Z ≡ z 0 will be employed when it is convenient to consider the reference
configuration as the initial target configuration; cf. Remark 7.1.

Remark 7.5. The TL and CR descriptions may be unified for a restricted class of elements: those containing
only translational freedoms. The unification freamework is presented in Chapters 14ee.

§7.3.7. Configurations and Staged Analysis

The meaning of some special configurations can be made more precise if the nonlinear analysis
process is viewed as a sequence of analysis stages, as discussed in Chapter 3. We restrict attention
to the Total Lagrangian (TL) and Corotational (CR) kinematic descriptions, which are the only
ones covered in this book. In a staged TL nonlinear analysis, two common choices for the reference
configuration are:
(1) Reference ≡ base. The base configuration is kept as reference configuration for all stages.
(2) Reference ≡ stage start. The configuration at the start of an analysis stage, i.e. at λ = 0, is
chosen as reference configuration.
A combination of these two strategies can be of course adopted. In a staged CR analysis the
reference is split between base and corotated. The same update choices are available for the base.
This may be necessary when rotations exceed 2π; for example in aircraft maneuvers.
The admissible configuration is a “catch all” concept that embodies all others as particular cases. The
perturbed configuration is an admissible variation from a admissible configuration. An ensemble
of perturbed configurations is used to establish incremental or rate equations.
The iterated and target configurations are introduced in the context of incremental-iterative solution
procedures for numerically tracing equilibrium paths. The target configuration is the “next solution”.
More precisely, an equilibrium solution (assumed to exist) which satisfies the total residual equations
for a given value of the stage control parameter λ. While working to reach the target, a typical
solution process goes through a sequence of iterated configurations that are not in equilibrium.
The corotated configuration is a rigid-body rotation of the reference configuration that “follows”
the current configuration like a “shadow”. It is used in the corotational (CR) kinematic description
of nonlinear finite elements. Strains measured with respect to the corotated configuration may
be considered “small” in many applications, a circumstance that allows linearization of several
relations and efficient treatment of stability conditions.

7–11
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

0
dX F
F−1
X2, x2 X u dx

X1, x1 X1,X2,X3 : material frame

X3, x3 x1,x2,x3 : spatial frame
Figure 7.8. Interpretation of F and its inverse as dX 
 dx action mappings as per (7.10).

§7.4. Deformation Measures

Characterizing the deformation of the material with respect to the reference configuration requires
the introduction of gradients of the motion.
§7.4.1. Deformation Gradients
The derivatives of (x, y, z) with respect to (X, Y, Z ), arranged in Jacobian format, constitute the
so-called deformation gradient matrix:
 ∂x ∂x ∂x 
∂(x, y, z) ∂ X ∂Y ∂Z
 ∂y ∂ y ∂y 
F= = . (7.8)
∂(X, Y, Z ) ∂ X ∂Y ∂ Z
∂z ∂z ∂z
∂ X ∂Y ∂ Z
which represents a second order tensor. The inverse relation gives the derivatives of (X, Y, Z ) with
respect to (x, y, z) as  ∂X ∂X ∂X 
∂x ∂y ∂z
−1 ∂(X, Y, Z )   ∂Y ∂Y ∂Y 

F = =  ∂x ∂y ∂z  . (7.9)
∂(x, y, z)
∂Z ∂Z ∂Z
∂x ∂y ∂z
The indicial versions of (7.8) and (7.9) are more compact: Fi, j = ∂ xi /∂ X j , Fi,−1j = ∂ X i /∂ x j . This
is convenient when F and F−1 are interpreted as second order tensors. As illustrated in Figure 7.8,
F and F−1 can be used to relate the coordinate differentials
   
dx dX
dx = dy = F dY = F dX, dX = F−1 dx. (7.10)
dz dZ
§7.4.2. Displacement Gradients
Since u = x − X, the displacement gradients with respect to C0 can be presented as the 3 × 3 matrix
 ∂x ∂x ∂ x   ∂u X ∂u X ∂u X 
∂X − 1 ∂Y ∂Z
 ∂y   ∂u ∂X ∂Y ∂Z 
G=F−I= ∂ y ∂ y = Y ∂u ∂u Y  = ∇u. (7.11)
∂Y − 1
Y
 ∂X ∂Z   ∂X ∂Y ∂Z 
∂z ∂z ∂z ∂u Z ∂u Z ∂u Z
∂X ∂Y ∂Z − 1 ∂X ∂Y ∂Z
7–12
 §7.4 DEFORMATION MEASURES

Likewise, displacement gradients with respect to C are given by

 ∂ Z   ∂u X ∂u X ∂u X 
1 − ∂∂ X
x
∂X
∂y ∂x   ∂x ∂y ∂z
 
 =I−F = −1  ∂Y 1 − ∂y∂Y ∂Y   ∂u Y ∂u Y ∂u Y 
G ∂ ∂z  =  ∂x ∂y ∂z . (7.12)
 x   
∂Z ∂Z ∂
1 − ∂z Z ∂u Z ∂u Z ∂u Z
∂x ∂y ∂x ∂y ∂z

Remark 7.6. For some FEM formulations it will be found convenient to arrange the displacement gradients
of (7.11) as a 9-component vector, which is written below as row vector to save space:

gT = [ g1 g2 g3 g4 g5 g6 g7 g8 g9 ]
 ∂u  (7.13)
= X ∂u Y ∂u Z ∂u X ∂u Y ∂u Z ∂u X ∂u Y ∂u Z .
∂X ∂X ∂X ∂Y ∂Y ∂Y ∂Z ∂Z ∂Z

Remark 7.7. Displacement gradient matrices are connected by the relations

G = (I − 
G)−1 − I, 
G = I − (I + G)−1 . (7.14)
−1
For small deformations G ≈ 
G and 
G ≈ G−1 . For finite deformations 
G G = I.
Remark 7.8. The ratio between infinitesimal volume elements d V = d x d y dz and d V0 = d X dY d Z in the
current and reference configuration appears in many relations. Because of (7.10) this ratio may be expressed
as
dV ρ0
= = det F = J, (7.15)
d V0 ρ
where ρ and ρ0 denote the mass densities in the current and reference configuration, respectively. The
abbreviation J = det F is frequently used later. This equation expresses the law of conservation of mass.

§7.4.3. Stretch Tensors

The combinations
C R = FT F, C L = F FT . (7.16)
are two symmetric, positive definite matrices extensively used in following Chapters. When viewed
as second-order tensors, they are called the left and right stretch tensors, respectively. Alternative
names and notations in the literature are mentioned in §8.1. The matrices (7.16) naturally emerge
from quadratic forms involving dX and dx as follows:

dxT dx = dXT FT F dX = dXT C R dX,

(7.17)
dXT dX = dxT F−T F−1 dx = dxT C−1
L dx.

From (7.17), finite strain measures can be constructed as described in the next Chapter. But an
important property should be prenoted: if the body motion is rigid, dxT dx = dXT dX identically,
whence both C R and C L must reduce to the identity I.
Because of (7.17), deformation and strain measures that involve C R are called Lagrangian or
√ involving C L are called Eulerian or spatial. The linkage may be direct or
material, whereas those
indirect; e.g., if U = C R , a measure that is a function of U is classified as Lagrangian.

7–13
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

The foregoing definitions are illustrated next with simple motion examples. Examples 7.5 through
7.7 below illustrate motions associated with homogeneous deformations, whereas Example 7.8
deals with the simple nonhomogeneous extension of a bar member.
Y, y
B0 0
B
X, x
Z, z
A0 A
L0 L
Figure 7.9. Simple extension of bar. Cross sections A–B in C and A0 –B0 in C0 actually
coincide; drawn offset for visualization convenience. Z axis normal to paper.

Example 7.5. Simple extension. The prototype case for this motion is a prismatic, homogeneous bar of
reference length L 0 stretched to a current length L. See Figure 7.9. Select X ≡ x along the bar axis, with
Y ≡ y and Z ≡ z normal to it. The origin is at the left cross-section of the bar, which is X -fixed in the motion.
Call λ1 = L/L 0 the axial stretch ratio along X , while λ2 and λ3 denote stretches along Y and Z , respectively.
The reference-to-current motion is
x = λ1 X, y = λ2 Y, z = λ3 Z . (7.18)
The displacements are u X = x − X = (λ1 − 1)X , u Y = y − Y = (λ2 − 1)Y , and u Z = z − Z = (λ1 − 1)Z .
The deformation and displacement gradients are
   
λ1 0 0 λ1 − 1 0 0
F= 0 λ2 0 , G= 0 λ2 − 1 0 . (7.19)
0 0 λ3 0 0 λ3 − 1
Since these are diagonal matrices, F−1 and 
G are easily obtained.
If the materially is transversally isotropic, λ2 = λ3 . If so, for small deformations −(λ2 − 1)/(λ1 − 1) becomes
the classical Poisson’s ratio ν. If the material is incompressible, the motion is isochoric (bar does not change
√
volume), whence λ1 λ2 λ3 = 1. If the material is incompressible and transversally isotropic, λ2 = λ3 = 1/ λ1 .
The case of pure dilatation is obtained by making λ1 = λ2 = λ3 in the foregoing equations.
This analysis is continued in Example 8.1 of next Chapter.

Example 7.6. Simple shear. Here a block of material is sheared by an angle θ as shown in Figure 7.10.
Material fibers aligned with X move horizontally and do not change length. As a result the block undergoes
the motion
x = X + γ Y, y = Y, z = Z. (7.20)
in which γ = tan θ is called the amount of shear. Material fibers aligned with X translate horizontally and do
not change length, so the motion is isochoric. The displacements are u X = γ Y , and u Y = u Z = 0.
Y, y
0 γY

θ
γ = tan θ

Z, z X, x
A0 B0 A B

Figure 7.10. Simple shear of a square block. Points A and B in C actually coincide with A0
and B0 in C0 ; drawn offset for visualization convenience. Z axis normal to paper.

7–14
 §7.4 DEFORMATION MEASURES

The deformation and displacement gradients are

   
1 γ 0 0 γ 0
F= 0 1 0 , G=F−I= 0 0 0 . (7.21)
0 0 1 0 0 0
Since these are sparse the polar decomposition is easily found. It is worked out in Example 8.2 of next Chapter.

L
ψ

C
(1) rigid translation (3) plane rigid rotation by
by uXC , uYC angle ψ (positive CCW)
HY uYC
Y, y (2) stretching by λ 1 = L /L 0 ,
0 λ 2 =HY /HY0 , λ 3 =1
Z, z X, x
HY0 C0
L0 uXC

Figure 7.11. Rectangular block undergoing translational rigid motion, stretching and rigid rotation, in that order.

Example 7.7. Combined stretch, rotation and rigid translation. This example shows how to combine homoge-
neous motions. A rectangular, bar-like block of material is subject to the three motions pictured in Figure 7.11.
They are applied in the following order:
1. A translational rigid motion that displaces the axes origin C0 to C by u XC and u Y C along X and Y ,
respectively. No motion along Z .
2. Stretchings of λ1 = L/L 0 , λ2 = HY /HY 0 and λ3 = 1 along X , Y , and Z , respectively. (Z lengths are not
pictured in the figure.)
3. A rigid rotation ψ (positive CCW) in the X, Y plane about C.
The combined motion can be expressed in the matrix form
     
x c −s 0 λ1 0 0 X − u XC
y = s c 0 0 λ2 0 Y − uY C . (7.22)
z 0 0 1 0 0 1 Z
in which c = cos ψ and s = sin ψ. The combination of stretching and rigid rotation by a matrix product is
called a multiplicative composition. Expanding gives
x = λ1 c (X − u XC ) − λ2 s (Y − u Y C ), y = λ1 s (X − u XC ) + λ2 c (Y − u Y C ), z = Z. (7.23)

Since matrix products do not necessarily commute, the order in which stretch and rotation are applied is
important. If the sequence were reversed: rotate-then-stretch, the motion would be different unless λ1 = λ2 .
By inspection, the deformation and displacement gradients are
   
λ1 c −λ2 s 0 λ1 c − 1 −λ2 s 0
F= λ1 s λ2 c 0 , G=F−I= λ1 s λ2 c − 1 0 . (7.24)
0 0 1 0 0 0
Note that the translational rigid motion does not affect these tensors, but the rigid rotation does. The polar
decomposition and strains are worked out in Example 8.4 of next Chapter.

7–15
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

Y, y
0
X, x
10 C0=30 Z, z 20
1 C 3 2

L 0 /2 L 0 /2 L /2 L /2
L0 L
u1 L 1 /2 L 2 /2
u3
u2

Figure 7.12. Rectangular block undergoing Nonhomogeneous axial motion.

Example 7.8. Nonhomogeneous bar extension. This example, depicted in Figure 7.12, illustrates the simplest
Nonhomogeneous motion: a bar undergoing a linearly varying axial deformation gradient. It has more finite
element flavor than the previous ones, and in fact can be taken as departure point for a 3-node bar element.
The bar has length L 0 in C0 and is aligned with X ≡ x. The frame origin is placed at the bar center C0 .
Three points called nodes are marked as shown in Figure 7.12: 10 and 20 at the ends X = ±L 0 /2 and 30
at the center X = 0, coincident with C0 . (Transition from letter to number labels in this example is linked
to future interpretation as a 3-node bar finite element.) The motion from C0 to C is defined by the 3 node
displacements: u 1 , u 2 and u 3 . The current configuration nodes are labeled 1, 2 and 3. The total deformed
length is L = L 0 + u 2 − u 1 = L 1 /2 + L 2 /2, in which L 1 /2 = L 0 /2 + u 3 − u 1 and L 2 /2 = L 0 /2 + u 2 − u 3 ,
as shown in the figure.
The displacement field can be expressed in FEM fashion as

u = u 1 N1 (ξ ) + u 2 N2 (ξ ) + u 3 N3 (ξ ), (7.25)

in which N1 (ξ ) = −ξ(1 − ξ )/2, N2 = ξ(1 + ξ )/2, N3 (ξ ) = 1 − ξ 2 are shape functions, and ξ = 2X/L 0 is
the isoparametric (natural) coordinate. The ξ coordinates of 10 , 20 and 30 are −1, +1 and 0, respectively. The
X → x mapping is x = X + u = X + u 1 N1 + u 2 N2 + u 3 N3 . The displacement gradient g = ∂ x/∂ X − 1 is
1  
g= u 1 (2ξ − 1) + u 2 (2ξ + 1) − u 3 4ξ . (7.26)
L0
This varies linearly in ξ , whence g is fully specified if we know its value at two distinct points on the bar axis.
Most convenient are those at which g can be readily evaluated in terms of the lengths L 1 , L 2 and L 0 . Those
are called gage points, which were introduced in the 1966 thesis [221] under the name strain nodes. A simple
calculation show that the quarter points in C0 , located at ξ = − 12 and ξ = 12 , are gage points labeled G 1 and
G 2 , respectively,9 whence
2(u 3 − u 1 ) L1 − L0 2(u 2 − u 3 ) L2 − L0
gG1 = = , gG2 = = . (7.27)
L0 L0 L0 L0
At any other location, g(ξ ) = 12 (gG1 + gG2 ) + (gG2 − gG1 )ξ , which in terms of lengths gives
L − L0 L2 − L1
g(ξ ) = + ξ. (7.28)
L0 L0
Why is it so important to express the axial displacement gradient in terms of original and deformed lengths?
Because those are invariant with respect to arbitrary rigid rotations in 2D or 3D. In practical FEM analysis the

9 In C0 , G 1 and G 2 are midway between 1–3 and 3–2, respectively: X G1 = −L 0 /4 and X G2 = L 0 /4, whence the “quarter
point” label. In C their position may be obtained through the mapping x G1 = X G1 +u(− 12 ) = −L 0 /4+(3u 1 −u 2 +6u 3 )/8
and x G2 = X G2 + u( 12 ) = L 0 /4 + (−u 1 + 3u 2 + 6u 3 )/8. Those are no longer quarter points in the deformed bar unless
the displacement is linear so u 3 = (u 1 + u 2 )/2.

7–16
 §7.5 *USEFUL VECTOR AND MATRIX FORMULAS

reference and current configuration of an element are not global-frame-aligned as in Figure 7.12. But if the
internal energy is expressed in terms of lengths, misalignment is irrelevant.
The next chapter continues the review of continuum mechanics by focusing on field equations used
in geometrically nonlinear FEM analysis.
§7.5. *Useful Vector and Matrix Formulas
This section collects vector and matrix identities used in subsequent chapters. It also provides a quick
“refresher” on matrix calculus operations.
§7.5.1. *Vector Identities
Below are some well known vector identities. Here a and b denote scalars, whereas u and v are generic 3 × 1
column vectors. All of these are assumed to be fields, that is, functions of position coordinates {x, y, z}; these
arguments are usually omitted for brevity. Functions are assumed to be sufficiently smooth for differentiation
operations to make sense. Direct notation is used; translation to matrix and indicial form is immediate. For
full form expressions in terms of components, see [727].

∇(a b) = a ∇b + b ∇a. (7.29)

∇(u . v) = (u . ∇) v + (v . ∇) u + u × (∇ × v) + v × (∇ × u). (7.30)

∇ × (∇a) = 0. (7.31)
∇ . (∇ . a) = 0. (7.32)
∇ × (∇ × u) = ∇ (∇ . u) − ∇ u. 2
(7.33)
∇ . (a u) = a (∇ . u) + (u.∇) a. (7.34)
∇ . (u × v) = v . (∇ × u) − u . (∇ × v). (7.35)
∇ × (a u) = a(∇ × u) + (∇a) × u. (7.36)
∇ × (u × v) = u (∇ . v) − v (∇ . u) + (v . ∇) u − (u . ∇) v. (7.37)
Note that ∇ . u = u i,i is the divergence of u (a scalar). On the other hand, ∇ . a is the gradient of a(x, y, z)
stored as a column vector whereas a . ∇ is the same gradient stored as a row vector.

§7.5.2. *Integral Identities

In the following integral identities, V denotes the volume of a body, S is its surface, n the exterior unit normal,
C is a closed curve (circuit) on the surface, and t the unit tangent to C. The differentials of V , S and C are
denoted as d V , d S and ds, respectively. Apropriate smoothness of S and C is assumed.
Divergence theorem for scalar field a = a(x, y, z):
 
a n dS = ∇a d V. (7.38)
S S

Divergence theorem for vector field u(x, y, z):

 
u . n dS = ∇ . u d S. (7.39)
S V

Divergence theorem for vector curl n(x, y, z) × u(x, y, z):

 
(n × u) d S = (∇ × u) d V. (7.40)
S V

7–17
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

Stokes’ theorem for vector field u(x, y, z):

 
u . t ds = (∇ × u) . n d S. (7.41)
C S

Stokes’ theorem for dot product u(x, y, z) . v(x, y, z):

 
 
(u . v) t ds = ∇(u . v) × n d S. (7.42)
C S

§7.5.3. *Quick Refresher on Matrix Calculus

The material in this subsection is intended to help readers unfamiliar (or rusty) with matrix calculus as regards
some finite element derivations. These involve taking derivatives of a scalar or vector with respect to a vector.
The exposition is done using full-form examples. In what follows a, b, and c denote scalars, x and y are 2 × 1
column vectors, A is an arbitrary 2 × 2 matrix, and S is a symmetric 2 × 2 matrix:
       
x1 y1 A11 A12 S11 S12
x= , y= , A= , S= . (7.43)
x2 y2 A21 A22 S12 S22
Matrix, full and indicial notations are used as appropriate; for the latter indices i, j run over 1,2.
Derivative of a scalar with respect to a vector:
 ∂c 
∂c   ∂ x1
= ∂∂c
xi = (7.44)
∂x ∂c
∂ x2
Derivative of a vector with respect to a vector:
 ∂ x1 ∂ x1 
∂x  ∂ xi  ∂ y1 ∂ y2
= ∂y = . (7.45)
∂y j ∂ x2 ∂ x2
∂ y1 ∂ y2
Second derivative of a scalar with respect to two vectors:
 
  ∂ 2c ∂ 2c
∂ 2c ∂ 2c ∂x ∂y ∂ x1 ∂ y2 
= ∂ xi ∂ y j =  12 1 (7.46)
∂x ∂y ∂ c ∂ 2c
∂ x2 ∂ y1 ∂ x2 ∂ y2
This yields a symmetric matrix if x = y, as in the second derivative of the GL strain given in (9.14)
Dyadic product (a.k.a. outer product) of two scalar-with-respect-to-vector derivatives:
 ∂a ∂b ∂a ∂b   ∂a   T
∂a ∂b  ∂a ∂b  ∂ x1 ∂ y1 ∂ x1 ∂ y2 ∂ x1  ∂b ∂b  = ∂a ∂b
⊗ = ∂x ∂y = ∂a ∂b ∂a ∂b = ∂ y1 ∂ y2 . (7.47)
∂x ∂y i j ∂a ∂x ∂y
∂ x2 ∂ y1 ∂ x2 ∂ y2 ∂ x2
In particular, if y = x and a = b = c, the resulting matrix is symmetric:
 ∂c ∂c ∂c ∂c
  ∂c   T
∂c ∂c  ∂c ∂c  ∂ x1 ∂ x1 ∂ x1 ∂ x2 ∂ x1  ∂c ∂c  = ∂c ∂c
⊗ = ∂x ∂x = ∂c ∂c ∂c ∂c = ∂ x1 ∂ x2 . (7.48)
∂x ∂x i j ∂c ∂x ∂x
∂ x2 ∂ x1 ∂ x2 ∂ x2 ∂ x2
Some consequences of above formulas for dot product forms:
√
∂(xT y) ∂(xT y) ∂(xT x) ∂ xT x x
= y, = x, = 2x, = √ . (7.49)
∂x ∂y ∂x ∂x xT x

7–18
 §7. Notes and Bibliography

And for quadratic forms:

∂(xT A y) ∂(xT A y) ∂(xT A x) ∂(xT S x)

= A y, = AT x, = (A + AT ) x, = 2 S x. (7.50)
∂x ∂y ∂x ∂x

Some textbooks in continuum mechanics use directional derivatives to derive these expressions. This approach
introduces an additional parameter and requires taking limits as the parameter approaches zero. For the
foregoing expressions, the extra work does not yield new results. But it can pay off for more complicated
cases; e.g., if curvilinear coordinates are used.
Notes and Bibliography
Few textbooks collectively cover the kinematics of classical continuum mechanics as well as the extensions
required to formulate geometrically nonlinear finite elements in three different descriptions (TL, UL and CR).
In this regard the two volumes by Crisfield [181,182] may be cited. See Notes and Bibliography in next
Chapter for additional references on books treating continuum mechanics proper.

7–19
Chapter 7: REVIEW OF CONTINUUM MECHANICS: KINEMATICS

Homework Exercises for Chapter 7

Review of Continuum Mechanics: Kinematics

EXERCISE 7.1 [A:15] The classical Lagrangian finite strain measure (“classical” in the sense of being the
first one published and most popular) is denoted as Green-Lagrange (GL). It was introduced by (George) Green
in 1841. To derive it, consider Figure 7.8. From a material particle in C0 emanates an arbitrary differential
vector: dX, which is mapped to dx via F as per (7.10). The GL strain tensor, denoted as eG , is defined as
def 1    
dXT eG dX = 2
||dx||2 − ||dX||2 = 1
2
dxT dx − dXT dX . (E7.1)
Show that
eG = 12 (FT F − I) = 12 (C R − I). (E7.2)

EXERCISE 7.2 [A:15] The classical Eulerian finite strain measure is denoted as Almansi-Euler (AE) — also
often called Almansi-Hamel. To derive it, consider Figure 7.8. From a spatial particle in C emanates an
arbitrary differential vector: dx, which is mapped to dX via F−1 as per (7.10). The AE strain tensor, denoted
by 
e A , is defined as
   
dxT 
def 1
e A dx = 2
||dx||2 − ||dX||2 = 1
2
dxT dx − dXT dX . (E7.3)
Show that
e A = 12 (I − F−T F−1 ) = 12 (I − C−1
 L ). (E7.4)

Y, y

rigid 90o rotation about Z

L=L0

0
X, x
Z, z
L0

Figure E7.1. Rigid rotation of bar about Z .

EXERCISE 7.3 [A:20] A bar of length L 0 lies originally along the X axis in the reference configuration C0 .
The bar is rigidly rotated 90◦ about Z ≡ z to lie along the Y ≡ y axis while retaining the same length at the
current configuration C. See Figure E7.1.
(a) Verify that the motion from C0 to C is given by
x = −Y, y = X, z = Z. (E7.5)
(b) Compute the displacement field u, the deformation gradient matrix F, the right stretch tensor C R , the
displacement gradient matrix G and the Green-Lagrange (GL) strain tensor given in (E7.2) above. Show
that the GL measure correctly predicts zero axial strain e GX X whereas the infinitesimal strain measure
 X X = ∂u X /∂ X predicts the absurd value of −100% strain.
EXERCISE 7.4 [A:25] Repeat the previous Exercise for the Biot, Hencky, Swainger, Almansi, midpoint and
Bazant strains listed in Figure 8.3 of the next Chapter. Verify that e X X = 0 for all of those measures.

7–20