Journal of Elasticity (2006) 83: 291–307 # Springer 2006

DOI: 10.1007/s10659-005-9046-9

A Frame-Free Formulation of Elasticityj

WALTER NOLL
Carnegie Mellon University, Pittsburgh, PA, USA. E-mail: [email protected]

Received 13 November 2005; in revised form 9 December 2005

Abstract. As I pointed out at the end of Sect. 4 in [N2] of my booklet Five Contributions to
Natural Philosophy, it should be possible to make the principle of material frame-indifference
vacuously satisfied by using an intrinsic mathematical frame-work that does not use an external
frame-space at all when describing the internal interactions of a physical system. Here I will do just
that for the classical theory of elasticity and also for the theory of hyperelasticity, i.e., elasticity
based on a strain-energy function. I will also comment on possible restrictions on the
corresponding intrinsic response functions.

Mathematics Subject Classifications (2000): M13003, P21018, T1501X.

Key words: frame-indifference, elasticity, hyperelasticity, material symmetry.

0. Introduction
The term Fprinciple of material frame-indifference_ was introduced in 1965 by C.
Truesdell and me in our contribution [NLFT] to the Encyclopedia of Physics.
Earlier, I had used the term Fprinciple of objectivity,_ and some people use this
term to this day. I meant Fobjectivity_ to express independence of the Fobserver,_
but Truesdell disliked the term as being too easily misinterpreted. In fact, I now
believe that the principle has nothing to do with Bobserver,’’ who is defined in
the dictionary to be a person. What matters is the use of frames of reference as a
means to make precise the concept of location, as explained in Sect. 4 of [N2] of
[FC].jj
Here is a full statement of the principle of material frame-indifference, as it
applies to any physical system:

The constitutive laws governing the internal interactions between the

parts of the system should not depend on whatever external frame of
reference is used to describe them.
j
This paper is based, in part, on lectures that I gave on June 29, 2005 at the meeting in Reggio-
Calabria in honor of the 65th birthday of Gianpietro Del Piero and on July 6, 2005, at the
University of Messina.
jj
I regret that I ever used the term Fprinciple of objectivity_ and thereby misled a lot of people. I
came to a complete understanding of the principle of frame-indifference only in the past 15 years
or so.
292 W. NOLL

The principle applies only to external frames of reference, not to frames that
are constructed from the system itself. It is important to note that the principle
applies only to internal interactions, not to actions of the environment on the
system and its parts, because usually the frame of reference employed is actively
connected with the environment. For example, if one considers the motion of a
fluid in a container, one usually uses the frame of reference determined by the
container, which certainly affects the fluid. Inertia should always be considered
as an action of the environment on the given system and its parts, and hence its
description does depend on the frame of reference used.

REMARK. It has been pointed out by Ingo Müllerj that some of the equations
of the thermo-mechanics of rarified gases derived from statistical mechanics do
not satisfy the principle of frame-indifference. These formulas involve inertial
effects not captured by the standard formula, in which all the inertial forces are
propertional to the mass density. This failure of frame-indifference is not
surprising because inertial actions are external actions and not internal
interactions. Since the mean free path between collisions of molecules in rarified
gases can be very large, it is doubtful that interactions such as forces and heat-
transfer can be described as pure surface interactions as in standard continuum
thermo-mechanics. (For a rigorous explanation, see [N6].) Therefore, formulas
that formally look like internal constitutive equations may actually have a non-
standard conceptual interpretation.

First, in Section 1 below, I will describe the conventional formulation of the

constitutive equation of elasticity and the effect of the principle of frame-
indifference on it. I believe that this description is more transparent than the one
given in [NLFT] and the standard textbooks because it makes explicit use of
different frame-spaces.
As I pointed out at the end of Sect. 4 in [N2], it should be possible to make
the principle of material frame-indifference vacuously satisfied by using an
intrinsic mathematical frame-work that does not use an external frame-space at
all when describing the internal interactions of a physical system. In Sections 2
and 3 below I will do just that for the classical theory of elasticity by specializing
the treatment given in Sects. 7, 9, and 10 of [N7]. In Section 4 I will do the same
for hyperelasticity, i.e., elasticity based on a strain-energy function. (See Sect. 82
of [NLFT].) In Section 5 I will comment on possible restrictions on the intrinsic
response functions defined in Sections 2 and 4.
As I pointed out in [NO] of [FC], I believe that [NLFT] is now in many
respects obsolete and should be updated. In the third paper [N3] of [FC] I present
some guidelines for such an update. The present paper could serve as a blueprint
for updating Chapter D of [NLFT] and also for the beginning chapters of future
textbooks on elasticity.
j
In a lecture at the meeting in Reggio-Calabria mentioned above.
A FRAME-FREE FORMULATION OF ELASTICITY 293

In the remainder of this paper, we assume that a continuous body-system B as

described in Sect. 3 of [N3] is given. The tangent space of B at a material point
X 2 B is denoted by T X and is called the body element of B at X. We say that the
body system consists of a simple material if the constitutive laws for every
material point X involve only the body element T X . A precise way of
formulating such laws is to put them in the form of a mathematical structure
involving T X . Most of the time we single out a particular material point X and
simply write T instead of T X . We emphasize that T is a linear space but not an
inner-product space.
We use the mathematical infrastructure, notation, and terminology of [FDS].
In particular, we use Flineon_ as a contraction of Flinear transformation from a
linear space to itself._j

1. Conventional Elasticity
The conventional formulation of the theory of elasticity involves a frame
space F (usually just called Fspace_), a reference placement
of B whose frame-
space is F , and, for each material point X, a response function g : L
! Sym V,
where L is a suitable open subset of the linear group Lis V of all automorphisms
of the translation space V of F , i.e., all invertible lineons on V. It is assumed that
1V , the identity mapping of V and the unity of the group Lis V, belong to L. The
value
T :¼ gðFÞ ð1:1Þ
is the Cauchy-stress at the present when F :¼ r
ðX Þ t is the gradient at
ðX Þ of
the transplacement t from the reference placement
to the present placement
t :¼ t 
: ð1:2Þ
The dependence of g on the material point X and the dependence of F and T on
time need not be made explicit. (See Sect. 43 of [NLFT].)
Now consider two frames of reference with corresponding frame-spaces F
0
and F and denote their translation spaces by V and V 0 , respectively. If x is the
location of a material point at a given time t in the frame-space F , then the
location of the same material point at the same time t in the frame-space F 0 will
be given by x0 ¼ t ðxÞ where t : F
! F 0 is an isometry and hence a
Euclidean isomorphism (see Sect. 45 of [FDS]). The mapping t 7! t describes
the motion of the frame F relative to the frame F 0 . The present placement
relative to F 0 corresponding to the placement t relative to F is given by
0t ¼ t  t : ð1:3Þ

j
In [NLFT] the term Ftensor_ is used instead of Flineon._ I pointed out in [N8] that Ftensor_ has a
much more general meaning and lineon is just a special case.
294 W. NOLL

The gradient At :¼ rt 2 Lis ðV; V 0 Þ (see Sect. 33 of [FDS]) is an inner-product

preserving linear isomorphism from V onto V 0 . A vector u and a lineon L relative
to the frame-space F at time t will appear as

u0 ¼ At u and L0 ¼ At LA
1
t ; ð1:4Þ

respectively, relative to the frame-space F 0 . In particular, the present Cauchy-

stress T0 relative to F 0 is related to the present Cauchy-stress T relative to F by

T0 ¼ At TA
1
t ; where At ¼ rt : ð1:5Þ

To say that descriptions in two frame spaces F and F 0 describe the same
physical process means that these descriptions must be isomorphic, i.e., the
description in the frame F can be transported to the description in the frame F 0
by a fixed Euclidean isomorphism  : F
! F 0 . Its gradient B :¼ r 2
Lis ðV; V 0 Þ is an inner-product preserving linear isomorphism from V to V 0 .
Now let the same elastic response at a material point desribed by (1.1) in the
frame F be described by the response function g0 : T 0
! Sym V 0 in the frame
F 0 . The requirement that this response function is obtained from the original one
by the Euclidean isomorphism  has the following consequences:

(a) The reference placement
0 used for g0 must be given by

0 ¼  
: ð1:6Þ

(b) The domain L0 of response function g0 must be related to the domain of the
response function g by

L0 ¼ fBLB
1 j L 2 Lg  Lis V 0 ð1:7Þ

and we must have

g0 ðBLB
1 Þ ¼ BgðLÞB
1 for all L 2 L: ð1:8Þ

Now, in the frame F 0 , the transplacement 0t from the reference placement
0
to the present placement 0t is characterized by
0t ¼ 0t 
0 : ð1:9Þ

Using (1.2), (1.3), and (1.6), it follows from (1.9) that t  t ¼ t  t 
¼

0t   
and hence that t  t ¼ 0t  . Taking the gradient at
ðX Þ, we
obtain

At F ¼ F 0 B ; ð1:10Þ

where F0 is the gradient at
0 ðX Þ ¼ ð
ðX ÞÞ of the transplacement 0t .

A FRAME-FREE FORMULATION OF ELASTICITY 295

For the frame F 0 , the stress relation (1.1) becomes

T0 :¼ g0 ðF0 Þ: ð1:11Þ
Using (1.1), (1.5) and (1.10), it follows from (1.11) that

At gðFÞA
1 0
1
t ¼ g ðAt FB Þ: ð1:12Þ
Using (1.8) with L :¼ B
1 At F, we obtain

g0 ðAt FB
1 Þ ¼ BgðB
1 At FÞB
1

and hence, by (1.12),

QgðFÞQ> ¼ gðQFÞ ; where Q :¼ B
1 At : ð1:13Þ

We note that Q is an inner-product automorphism of V, in other words, a
member of the orthogonal group Orth V, so that Q
1 ¼ Q> . The dependence
of F and Q on time need not be made explicit here. Now, (1.13) must hold for
arbitrary motions and hence for all F 2 L. Also, the frame F 0 can be chosen in
an arbitrary manner and hence (1.13) must also hold for all Q 2 Orth V. Thus,
the relation (1.13) coincides with (43.2) in [NLFT] and is the basis for all the
reduced constitutive equations stated in Sect. 43 of [NLFT].

2. Frame-free Elasticity
To describe the structure of an elastic element, we need a little bit of linear
algebra that is not presented in most textbooks. (However, a detailed treatment is
given in Chapter 2 of [FDS].)
Let a finite-dimensional linear space T be given. The dual T * :¼ Lin ðT ; IRÞ
of T consists of all linear forms on T . The dual T ** of the dual T * can be
identified with the original space T as follows: When an element t 2 T is
regarded as an element of T **, its action on T * is given by

 :¼  t
t for all  2 T *: ð2:1Þ
We express this identification by T ** ffi T .
Let V also be finite-dimensional linear space and let L 2 Lin ðT ; VÞ be a
linear mapping from T to V. Then the transpose L> 2 Lin ðV*; T *Þ is defined
by

ðL>  Þt :¼  ðLtÞ for all t 2 T ;  2 V*: ð2:2Þ

296 W. NOLL

Let G 2 Lin ðT ; T *Þ be given. Then, since T ** ffi T , we can identify

Lin ðT **; T *Þ ffi Lin ðT ; T *Þ, so that G and its transpose G> belong to the
same space. We say that G is symmetric if G> ¼ G. The set

Sym ðT ; T *Þ :¼ fG 2 Lin ðT ; T *Þ j G> ¼ G g ð2:3Þ

is a subspace of Lin ðT ; T *Þ.

We say that G 2 Sym ðT ; T *Þ is positive if

ðGtÞt  0 for all t 2 T ð2:4Þ

and we say that it is strictly positive if, in addition, ðGtÞt is zero only when t is
zero. The set of all positive elements of Sym ðT ; T *Þ is denoted by PosðT ; T *Þ
and the set of all strictly positive elements by Posþ ðT ; T *Þ. Both PosðT ; T *Þ
and Posþ ðT ; T *Þ are linear cones, i.e., they are stable under addition and scalar
multiplication with strictly positive real numbers.j All the elements of Posþ
ðT ; T *Þ are invertible and hence linear isomorphisms, i.e., we have

Posþ ðT ; T *Þ  Lis ðT ; T *Þ: ð2:5Þ

An inner-product space is a finite-dimensional linear space V with an

additional structural element given by singling out a particular member
ip 2 PosðT ; T *Þ. Since ip 2 Lis ðV; V*Þ, it then can be and is used to identify
V* ffi V, which justifies the standard notation v  w :¼ ipðvÞw for all v; w 2 V.
As explained in Sects. 3 and 4 of [N3], a configuration of the given body-
system B is a metric d : B  B
! IP that is isometric to a fit regionjj in a
Euclidean space. Such a configuration induces an element G 2 Posþ ðT ; T *Þ for
the body element T . We call G the configuration of the element induced by the
configuration d of the whole system.
The structure of an elastic material element is described by the following
ingredients:

1) A body element T ,
2) A convex open subset G of Posþ ðT ; T *Þ, whose members are called the
configurations of the element.
3) A mapping h : G
! Sym ðT *; T Þ, called the intrinsic stress–response
function of the element. Its value

S ¼ hðGÞ ð2:6Þ

is the intrinsic stress determined by the configuration G.

j
A detailed analysis of such linear cones is given in [NS].
jj
The definition and a detailed analysis of the concept of a fit region is given in [NV].
A FRAME-FREE FORMULATION OF ELASTICITY 297

A deformation of the elastic element is simply a change of configurations, and

a deformation process is a one-parameter family of configurations as described in
Sect. 4 of [N3].
As I pointed out in Sect. 6 of [N3], a frame of reference is usually employed
to describe the external actions of the environment on the body-system. Thus, we
now assume, as in Section 1, that a frame space F with translation space V and
a reference placement
of B in F are given. We denote the present placement
by t and the transplacement from the reference placement
to the present
placement by t , so that (1.2) holds. Taking the gradient of (1.2) at the given
material point X 2 B we obtain

M ¼ FK ; where K :¼ rX
; M :¼ rX t ; and F :¼ r
ðX Þ t : ð2:7Þ

We call K 2 Lis ðT ; VÞ the reference placement of the body element T , M 2

Lis ðT ; VÞ the present placement of this body element, and F 2 Lis ðVÞ the
transplacement of this body element from K to M. The dependence of M and F
on time need not be made explicit.
As explained in Sect. 4 of [N3], the configuration

GR :¼ K> K 2 Posþ ðT ; T *Þ ð2:8Þ

of the element T is induced by the reference placement K and will be called the
reference configuration. The present configuration

G :¼ M> M ¼ K> F> FK 2 Posþ ðT ; T *Þ ð2:9Þ

of the element T is induced by the present placement M. The requirement that

GR and G belong to the prescribed subset G of Posþ ðT ; T *Þ is equivalent to the
conditions

F 2 L :¼ f F 2 Lis V j K> F> FK 2 Gg and 1V 2 L: ð2:10Þ

Now, by (6.4) of [N3], the present Cauchy-stress T is related to the present

intrinsic stress S by

1
S ¼ M
1 TM> : ð2:11Þ
Therefore, since S ¼ hðGÞ is the intrinsic stress determined by the present
configuration G, it follows from (2.11), (2.9) and (2.7) that the Cauchy stress
determined by F is given by

T ¼ gðFÞ :¼ FKhðK> F> FKÞK> F> : ð2:12Þ

The domain L of the response function g defined by (2.12) is given by (2.10).

Given F 2 L it easily follows from (2.12) that g does indeed satisfy the condition
(1.13) for all Q 2 Orth V.
298 W. NOLL

We note that the intrinsic response function h is related to the response func-
tion t for the second Piola–Kirchhoff tensor (see (43.10)2 in [NLFT]) as follows:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
~ ¼ tðCÞ ¼
T ðdet CÞKhðK> CKÞK> when C :¼ F> F: ð2:13Þ

We emphasize again that the body-element T is just a three-dimensional

linear space and not an inner product space. Let a configuration G of T be given.
It is a member of the linear cone Posþ ðT ; T *Þ. Since an inner product is just a
member of this cone that has been singled out, we can deal with G as if it were
an inner product and apply the concepts, notations, and theory of inner-product
spaces, but they must be used only relative to the given configuration. For
example, a basis b :¼ ðb1 ; b2 ; b3 Þ of T is G-orthonormal if

0 if i 6¼ j
ðGbi Þbj ¼ i; j :¼ for all i; j 2 f1; 2; 3g: ð2:14Þ
1 if i ¼ j

A basis that is G-orthonormal necessarily fails to be G0 -orthonormal when G0 is a

configuration different from G.
The orthogonal group of the configuration G is defined by

Orth ðGÞ :¼ fA 2 Lis T j A> GA ¼ Gg: ð2:15Þ

Orth ðGÞ is not only a subgroup of the linear group Lis T , but even of the
unimodular group

Unim T :¼ fA 2 Lis T j jdetAj ¼ 1g: ð2:16Þ

Unim T includes infinitely many orthogonal groups as subgroups, namely one

for each configuration G.
Let S be the intrinsic stress determined by the configuration G as in (2.6). The
spectral theory as treated, for example, in Chapter 8 of [FDS] can be applied to S
relative to G with the following result: There is a G-orthonormal basis b :¼
ðb1 ; b2 ; b3 Þ of T and a list ð1 ; 2 ; 3 Þ of real numbers such that

ðSGÞbi ¼ i bi for all i 2 f1:2:3g: ð2:17Þ

The terms in ð1 ; 2 ; 3 Þ are called the principal stresses of S relative to G.

If we introduce a frame-space F as before and consider the Cauchy-stress T
related to S by (2.11), it follows from (2.17), using (2.9), that

Tvi ¼ i vi ; where vi :¼ Mbi for all i 2 f1:2:3g: ð2:18Þ

Hence the principal stresses characterized by (2.17) coincide with the ones in
conventional elasticity. (See Sect. 48 of [NLFT].)
A FRAME-FREE FORMULATION OF ELASTICITY 299

3. Material Isomorphisms, Material Symmetry

Isomorphisms between elastic material elements are defined in accordance with
the general notion of isomorphism for arbitrary mathematical structures. To be
specific, let ðT 1 ; G1 ; h1 Þ and ðT 2 ; G2 ; h2 Þ be the ingredients that define two elastic
material elements as described by 1), 2), and 3) in the previous section. A linear
isomorphism A 2 Lis ðT 1 ; T 2 Þ is a material isomorphism if

G1 ¼ fA> GA j G 2 G2 g ð3:1Þ

and

Ah1 ðA> GAÞA> ¼ h2 ðGÞ for all G 2 G2 : ð3:2Þ

We say that an elastic body system B is materially uniform if, for any two
material points X ; Y 2 B, the corresponding elastic elements ðT X ; GX ; hX Þ and
ðT Y ; GY ; hY Þ are materially isomorphic. We say that the body system B is
homogeneous if there is a placement
of B in a frame-space F with translation
space V such that, for any two material points X ; Y 2 B,

AXY :¼ ðrY
Þ
1 rX
2 Lis ðT X ; T Y Þ ð3:2Þ

is a material isomorphism.
A body system can be materially uniform without being homogeneous,
and the deviation from homogeneity is related to what are often called
Fcontinuous distributions of dislocations._ (See also the Remark at the end of
Sect. 9 in [N7].)
Now let an elastic material element ðT ; G; hÞ be given as described in the
previous section. An automorphism of this element, i.e., an isomorphism from
the element to itself, is called a symmetry of the element. We assume that all
symmetries are unimodular because a change of volume cannot reasonably leave
material properties unchanged. They form the symmetry group

G :¼ fA 2 Unim T j AhðA> GAÞA> ¼ hðGÞ for all G 2 Gg: ð3:3Þ

Of course, it must be assumed that, for each A 2 G, we must have A> GA 2 G

whenever G 2 G.
The symmetry group of a configuration G 2 G is defined by

GG :¼ G \ Orth ðGÞ: ð3:4Þ

It is easily seen that

AhðGÞG ¼ hðGÞGA for all A 2 GG : ð3:5Þ

300 W. NOLL

We say that a given configuration G 2 G is isotropic if GG :¼ Orth ðGÞ , i.e.,

if Orth ðGÞ  G. We denote the set of all isotropic configurations by
Itr G :¼ fG 2 G j Orth ðGÞ  Gg: ð3:6Þ
If G 2 G is isotropic, it follows from (3.5) that hðGÞG commutes with all
A 2 Orth ðGÞ and hence is proportional to the identity 1T of T . Therefore, there
is a function ^ : Itr G
! IR such that
hðGÞG ¼
^ðGÞ1T for all G 2 Itr G: ð3:7Þ
It easily follows from (3.3), (3.6) and (3.7) that
^ðA> GAÞ ¼ ^ðGÞ for all G 2 Itr G ; A 2 G: ð3:8Þ
We need the following

PROPOSITION. Let G; G0 2 Posþ ðT ; T *Þ be given. Then

ðG0 ¼ A> GA for some A 2 Unim T Þ () detðG
1 G0 Þ ¼ 1: ð3:9Þ

The proof of this proposition is essentially the same as that of Prop. 19.1 in
[N7]. Using this Proposition and (3.8), we see that
 
detðG
1 G0 Þ ¼ 1 ¼) ^ðG0 Þ ¼ ^ðGÞ for all G; G0 2 Itr G: ð3:10Þ
We now choose a reference configuration GR 2 G and define the function
^ : G
! IP by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
^ðGÞ ¼ detðG
1 GR Þ for all G 2 G: ð3:11Þ

If the units for mass and volume are chosen such that the mass density (mass per
unit volume) in the reference configuration is 1, then the value ^ðGÞ is the mass-
density determined by the configuration G. Here, mass can be taken as a proxy
for volume in the reference-configuration, and hence need not be connected with
inertial or gravitational mass, which should not enter internal constitutive laws. It
is easily seen from (3.11) that

detðG
1 G0 Þ ¼ 1 () ^ðG0 Þ ¼ ^ðGÞ for all G; G0 2 G: ð3:12Þ

Therefore, by (3.10), the value ^ðGÞ depends only on ^ðGÞ. Hence there is a
function

: P
! IR ; where P :¼ ^> ðItr GÞ; ð3:13Þ
such that ^ðGÞ ¼
ð^ðGÞÞ for all G 2 Itr G. Hence (3.7) yields

hðGÞ ¼

ð^ðGÞÞG
1 for all G 2 Itr G: ð3:14Þ

A FRAME-FREE FORMULATION OF ELASTICITY 301

The function
is called the pressure function for the set Itr G of all isotropic
configurations of the elastic material element.
We now assume, as at the end of Section 2, that a frame space F with
translation space V and a reference placement
of B in F are given and we use
again the notation (2.7). It easily follows from (2.8) and (2.9) that

detððGR Þ
1 GÞ ¼ detðFF> Þ ¼ jdet F j2

and hence, by (3.11), ^ðGÞ ¼ jdet F j
1 . Therefore, the value
ðFÞ of the func-
tion
: L
! IP defined by

ðFÞ :¼ jdet F j
1 for all F 2 L ð3:15Þ

gives the mass-density of the body element when its configuration is determined
by F according to (2.9). Using (2.12), (3.14), and (2.9), we easily conclude that

T ¼ gðFÞ ¼

ð
ðFÞ1V ð3:16Þ

when the configuration determined by F is isotropic. This formula (3.16) is con-

sistent with conventional formulas such as (50.2)1 in [NLFT].
We say that the element is fluid if its symmetry group is the full unimodular
group, i.e., if

G ¼ Unim T : ð3:17Þ

The element is fluid if and only if all of its configurations are isotropic, i.e., we
have Itr G ¼ G. Hence, for a fluid element, the relation (3.14) is valid for all
configurations G 2 G and (3.16) is valid for all F 2 L. In this case, the pressure
function
alone is enough to describe the elastic response.
We say that the element is solid if

G  Orth ðGÞ for some G 2 G: ð3:18Þ

If that is the case then the configurations G for which (3.18) holds are said to be
undistorted.
We say that the element is isotropic if it has isotropic configurations, i.e., if

Orth ðGÞ  G for some G 2 G: ð3:19Þ

Assume now on that the element is an isotropic solid element, which means,
by (3.18) and (3.19), that its undistorted configurations G are characterized by
the property that

Orth ðGÞ ¼ G: ð3:20Þ

302 W. NOLL

In fact, for isotropic solids, the set of undistorted configurations coincides with
the set Itr G of isotropic configurations. Moreover, if GR is an undistorted
reference configuration, we have
Orth ðGR Þ ¼ G ¼ Orth ðGÞ for all G 2 Itr G: ð3:21Þ
It easily follows from (3.21) that all undistorted configurations are proportional
to GR , i.e., that

Itr G  f cGR j c 2 IP g: ð3:22Þ

4. Frame-free Hyperelasticity
The structure of a hyperelastic material element is obtained from that of an
elastic material element defined in Section 2 by adding two more ingredients and
a basic axiom. The ingredients are:

4) A reference configuration GR 2 G.
5) A function ^ : G
! IP , of class C1, called the intrinsic energy-response
function.

Its value
 ¼ ^ðGÞ ð4:1Þ
is the strain-energy per unit mass determined by the configuration G.
As in the previous section, mass can be taken as a proxy for volume in the
reference configuration.
To formulate the basic axiom of hyperelasticity, we need the following result
from linear algebra, which is not in most textbooks:
Every member M
of ðSym ðT ; T *ÞÞ* :¼ Lin ðSym ðT ; T *Þ; IR Þ has exactly
one representation by a member M of Sym ðT *; T Þ such that

ðLÞ ¼ tr ðMLÞ
M for all L 2 Sym ðT ; T *Þ: ð4:2Þ
(This result is a consequence of the Representation Theorem for Linear Forms
on a Space of Linear Mappigs in Sect. 26 of [FDS].)
We use (4.2) to identify ðSym ðT ; T *ÞÞ* ffi Sym ðT *; T Þ by omitting, on the
right side of (4.2), the bar atop M.
To formulate the basic axiom, we also make use of the mass-density function
^ defined by (3.11).

BASIC AXIOM. The intrinsic stress–response function is related to the intrinsic

energy response function by

hðGÞ ¼ 2^ðGÞrG ^ for all G 2 G: ð4:3Þ

A FRAME-FREE FORMULATION OF ELASTICITY 303

Note that (4.3) makes sense only because of the identification mediated by
(4.2) when the bar over M is omitted because the right side belongs to
ðSym ðT ; T *ÞÞ* while the left side belongs to Sym ðT *; T Þ.
We now assume that a frame-vector-space V is given as in previous sections.
We choose a reference placement K 2 Lis ðT ; VÞ of the element in such a way
that
GR :¼ K> K; ð4:4Þ
the configuration induced by K, is the reference configuration given as ingredient
in 4). Every other placement of the element in the space V then has the form
FK 2 Lis ðT ; VÞ where F 2 Lis V is a transplacement of the element. The con-
figuration G induced by the placement FK is

G ¼ ðFKÞ> ðFKÞ ¼ K> CK; where C :¼ F> F : ð4:5Þ

Recall that C is called the right Cauchy–Green tensor in [NLFT].

We now put
G
:¼ fC 2 Posþ V j K> CK 2 Gg ð4:6Þ

and define 
: G

! IP by


ðCÞ :¼ ^ðK> CKÞ for all C 2 G
: ð4:7Þ

It easily follows from (4.7) that

rC 
¼ KðrK> CK ^ÞK> for all C 2 G
: ð4:8Þ

In analogy to (4.7) define
: G

! IP by

ðCÞ :¼ ^ðK> CKÞ for all C 2 G
: ð4:9Þ

Now put

L :¼ fF 2 Lis V j F> F 2 G
g: ð4:10Þ

In view of (2.12) we have

gðFÞ :¼ ðFKÞhðK> CKÞðFKÞ> ; where C :¼ F> F; for all F 2 L:

ð4:11Þ
Therefore, by (4.3), (4.7), (4.8), and (4.9), we conclude that
gðFÞ :¼
ðCÞFðrC 
ÞF> ; where C :¼ F> F; for all F 2 L: ð4:12Þ
This relation is identical to (84.12) in [NLFT] and hence shows that the basic
axiom is consistent with the classical frame-dependent theory of hyperelasticity.
304 W. NOLL

REMARK. The relation (4.12) is a reduced version of the stress relation (82.9)
in [NLFT]. In the context of thermodynamics, the latter can be derived from the
second law of thermodynamics by a procedure described in [CN] (see (5.4) in
[CN]). A frame-free version of this procedure should yield the relation (4.3) of
the Basic Axiom directly.

The symmetry group of the strain-energy is defined by

G :¼ fA 2 Unim T j ^ðA> GAÞ ¼ ^ðGÞ for all G 2 Gg: ð4:13Þ

THEOREM. We have G  G, i.e., every strain-energy symmetry is also a

symmetry of the elastic element. If the element is solid and if the reference
confiuration is undistorted we actually have G ¼ G.
Proof. Let A 2 G be given so that

^ðA> GAÞ ¼ ^ðGÞ for all G 2 G: ð4:14Þ

Taking the gradient with respect to G and using the identification (4.2), we find

tr ððrG ^ÞLÞ ¼ tr ððrA> GA ^ÞA> LAÞ

¼ tr ðAðrA> GA ^ÞA> LÞ for all L 2 Sym ðT ; T *Þ; ð4:15Þ

and hence

AðrA> GA^ ÞA> ¼ rG : ð4:16Þ

Noting, in view of (3.11), that

^ðA> GAÞ ¼ ^ðGÞ for all G 2 G; ð4:17Þ

it follows from (4.16) and the basic axiom (4.3) that AhðA> GAÞA> ¼ hðGÞ for
all G 2 G and hence, in view of (3.3), that A 2 G. Since A 2 G was arbitrary,
the inclusion G  G is proved.
Actually, the proof above shows that A 2 G if and only if the gradient of
(4.14) with respect to G is zero. That is the case if and only if the difference
^ðA> GAÞ
^ðGÞ does not depend on G. This means that we have A 2 G if and
only if

^ðA> GAÞ
^ðGÞ ¼ ^ðA> GR AÞ
^ðGR Þ for all G 2 G: ð4:18Þ

Now, if the element is solid and the reference configuration undistorted, it

follows from (3.16) with G :¼ GR that A 2 Orth ðGR Þ. Therefore, we have
A> GR A ¼ GR and the right side of (4.18) is zero. Since A 2 G was arbitrary we
conclude from (4.18) and (4.14) that, in this case, G ¼ G. Ì
A FRAME-FREE FORMULATION OF ELASTICITY 305

The Theorem just proved is the frame-free version of a result presented in

Sect. 58 of [NLFT].

5. Restrictions upon the Response Functions

The most common restrictions on the intrinsic response functions concern
smoothness. If the energy response function ^ is of class C1 then the stress
response function h is continuous; if ^ is of class C2, then h is of class C1. The
latter assumption is sufficient for most purposes.
The linear cone Posþ ðT ; T *Þ and the space Sym ðT ; T *Þ have natural
(partial) orders, denoted by  and defined by
G1  G2 () G2
G1 2 Posþ ðT ; T *Þ for all G1 ; G2 2 Posþ ðT ; T *Þ
ð5:1Þ
and
S1  S2 () S2
S1 2 Posþ ðT *; T Þ for all S1 ; S2 2 Sym ðT *; T Þ:
ð5:2Þ
(These definitions are special cases of (2.8) in [NS].)
Note: Given G 2 Posþ ðT ; T *Þ and c 2 IP we have G  cG () c > 1.
We postulate the following

AXIOM. The stress–response function h is strictly isotone in the sense that

G1  G2 ¼) hðG1 Þ  hðG2 Þ for all G1 ; G2 2 G: ð5:3Þ
Intuitively, the axiom states that in order to expand the elastic element in all
directions, one must increase the stresses in all directions.

PROPOSITION. For isotropic materials, the pressure function
characterized

by (3.13) and (3.14) is stricty antitone in the sense that

1 < 2 ¼)
ð 1 Þ >
ð 2 Þ for all 1; 2 2 P: ð5:4Þ

Proof. Let 1 2 P be given. By (3.13), we may choose G1 2 Itr G such that
1 ¼ ^ðG1 Þ. Let c 2 IP be given and put G2 :¼ cG1 2 G. Since Orth ðG2 Þ ¼
Orth ðG1 Þ  G we also have G2 2 Itr G. An easy calculation using (3.11) shows
that
3 3
^ðG2 Þ ¼ c
2 ^ðG1 Þ ¼ c
2 1 : ð5:5Þ

Now let 2 P be given. By (5.5) we then have

2
  23
1
^ðG2 Þ ¼ 2 () c :¼ : ð5:6Þ
2
306 W. NOLL

We assume now that (5.6) holds and use (3.14) with the result

1
ð 2 Þ
1
hðG2 Þ ¼

ð 2 ÞG
1
2 ¼

ð 2 ÞðcG1 Þ ¼ c hðG1 Þ: ð5:7Þ

ð 1 Þ
Using (5.6) we see that
G1  G2 () c > 1 () 1 > 2; ð5:8Þ
and using (5.7) we see that

ð 2 Þ
hðG1 Þ  hðG2 Þ () > c: ð5:9Þ

ð 1 Þ
Combining (5.8) and (5.9), we infer from the strict isotonicity of h that
1 > 2 ¼)
ð 2 Þ >
ð 1 Þ. Ì

The antitonicity of
implies that there is at most one 2 P such that

ð Þ ¼ 0. In view of (3.20) we conclude that, for isotropic solids, there is at most
one undistorted configuration GN such that hðGN Þ ¼ 0. If such GN exists, it is
called a natural configuration.
A response restriction that one might explore is the assumption that the
energy-response function ^ is convex. It is not hard to prove that this is the case
if and only if the stress–response function h satisfies
 
tr ð 1 hðG2 Þ
2 hðG1 ÞÞðG2
G1 Þ  0 for all G1 ; G2 2 G ð5:10Þ

when 1 :¼ ^ðG1 Þ ; 2 :¼ ^ðG2 Þ.

A large number of possible restrictions on the response functions have been
considered in the conventional literature on elasticity. The following names
appear in Sects. 51, 52, and 86 of [NLFT]: P–C inequality, T–E inequalities, IFS
condition, E–T inequalities, O–F inequalities, B–E inequalities, E–inequalities,
GCN condition, and C–N condition. In the later literature, conditions such as
rank-one convexity, quasi-convexity, and poly-convexity have been extensively
explored.
All these restrictions should be frame-indifferent in order to be relevant. I am
not sure whether this has been proved in all cases. If they are frame-indifferent, it
must be possible to give them a frame-free formulation. I hope that someone will
start the task of doing so in the future.

6. Phase Transitions
The domain G of the intrinsic response function h in Section 2 (or of the intrinsic
strain energy function ^ in Sect. 4) must be restricted in such a way that the
configurations in G are compatible with only a single phase of the material
element in question. If the material element undergoes a phase transition, the
A FRAME-FREE FORMULATION OF ELASTICITY 307

configurations of the new phase belong to a subset G0 of Posþ ðT ; T *Þ that is

disjoint from G. The intrinsic response function h0 for the new phase need not
have the same material symmetry as h. For example, the material can be fluid in
one phase and solid in the other. If a configuration belongs to a point on the
boundary of both G and G0 , one can expect that the element can describe two
coexistent phases.
Phase transitions are related to questions of stability, and these are related to
the restrictions mentioned in the previous section. I hope that someone will
develop a frame-free conceptual analysis of such phase transitions. This analysis
cannot be complete unless the entire matter is treated in the context of thermo-
mechanics.

Acknowledgement
I am grateful to Roger Fosdick for helpful suggestions and proofreading.

References
[FDS] W. Noll, Finite-Dimensional Spaces: Algebra, Geometry, and Analysis, Vol. I. Kluwer
Academic (1987) 393 pages.
[NLFT] C. Truesdell and W. Noll, The Non-Linear Field Theories of Mechanics, 3rd edn.
Springer (2004) 602 pages. (The first edition appeared in 1965.)
[FC] W. Noll, Five Contributions to Natural Pilosophy. Published electronically (2004) 73
pages. Posted on the website math.cmu.edu/~wn0g/noll.
[N0] W. Noll, Introduction to [FC]
[N2] W. Noll, On the Principle of Material Frame-Indifference. Second paper in [FC].
[N3] W. Noll, Updating the Non-Linear Field Theories of Mechanics. Third paper in [FC].
[N6] W. Noll, The Theory of Surface Interactions. (2005) 14 pages. Posted on the website
math.cmu.edu/~wn0g/noll.
[N7] W. Noll, A new mathematical theory of simple materials. Arch. Ration. Mech. Anal. 52
(1972) 1–50.
[N8] W. Noll, Isocategories and Tensor Functors. Dept. of Mathematics, Carnegie Mellon
University, Research Report No.92-142 (1992) 19 pages. Posted on the website math.
cmu.edu/~wn0g/noll.
[NS] W. Noll and J.J. Schäffer, Orders, gauge, and distance in faceless linear cones; with
examples relevant to continuum mechanics and relativity. Arch. Ration. Mech. Anal. 66
(1977) 345–377.
[NV] W. Noll and E. Virga, Fit regions and functions of bounded variation. Arch. Ration.
Mech. Anal. 102 (1988) 1–21.
[CN] B.D. Coleman and W. Noll, The thermodynamics of elastic materials with heat
conduction and viscosity. Arch. Ration. Mech. Anal. 13 (1963) 167–178.