Discover millions of ebooks, audiobooks, and so much more with a free trial

Only $11.99/month after trial. Cancel anytime.

Theory of Viscoelasticity: Second Edition
Theory of Viscoelasticity: Second Edition
Theory of Viscoelasticity: Second Edition
Ebook642 pages49 hours

Theory of Viscoelasticity: Second Edition

Rating: 0 out of 5 stars

()

Read preview

About this ebook

This comprehensive text, featuring the integration of numerous theoretical developments, offers a complete, consistent description of the linear theory of the viscoelastic behavior of materials. Relevant theoretical formulations are derived from a continuum mechanics viewpoint, followed by discussions of problem-solving techniques.
Topics cover viscoelastic stress strain constitutive relations, isothermal boundary value problems, thermoviscoelasticity, mechanical properties and approximate transform inversion, problems of a nontransform type, wave propagation, general theorems and formulations, nonlinear viscoelasticity, and nonlinear mechanical behavior. The text, which is ideal for graduate-level students, also includes Appendixes and an Index.
LanguageEnglish
Release dateApr 26, 2013
ISBN9780486318967
Theory of Viscoelasticity: Second Edition

Related to Theory of Viscoelasticity

Related ebooks

Civil Engineering For You

View More

Related articles

Reviews for Theory of Viscoelasticity

Rating: 0 out of 5 stars
0 ratings

0 ratings0 reviews

What did you think?

Tap to rate

Review must be at least 10 words

    Book preview

    Theory of Viscoelasticity - R. M. Christensen

    VISCOELASTICITY

    1.1. INTRODUCTION

    The general development and broad application of the linear theory of viscoelasticity is a relatively recent occurrence. In fact, the activity in this field has been primarily due to the large scale development and utilization of polymeric materials. Many of these newly developed materials exhibit mechanical response characteristics which are outside the scope of such theories of mechanical behavior as elasticity and viscosity; thus, the need for a more general theory is quite apparent.

    To be more specific, the theory of elasticity may account for materials which have a capacity to store mechanical energy with no dissipation of the energy. On the other hand, a Newtonian viscous fluid in a nonhydrostatic stress state implies a capacity for dissipating energy, but none for storing it. But, then, materials which must be outside the scope of these two theories are those for which some, but not all, of the work done to deform them, can be recovered. Such materials possess a capacity to both store and dissipate mechanical energy.

    A different way of characterizing these materials is through the nature of their response to a suddenly applied uniform distribution of surface tractions on a specimen. The term suddenly applied loading state or stress state as used in the present context does not imply rates sufficiently great to cause the excitation of a dynamic response in the specimen. An elastic material, when subjected to such a suddenly applied loading state held constant thereafter, responds instantaneously with a state of deformation which remains constant. A Newtonian viscous fluid responds to a suddenly applied state of uniform shear stress by a steady flow process. There are, however, materials for which a suddenly applied and maintained state of uniform stress induces an instantaneous deformation followed by a flow process which may or may not be limited in magnitude as time grows. A material which responds in this manner is said to exhibit both an instantaneous elasticity effect and creep characteristics. This behavior is clearly not described by either an elasticity or a viscosity theory but combines features of each.

    It is instructive to consider a situation which represents a generalization of the response to a single suddenly applied change in surface tractions. Suppose a material having the instantaneous elasticity and creep characteristics described above is subjected to two nonsimultaneously applied sudden changes in uniform stress, superimposed upon each other. After the first application of stress, but before the second, the material responds in some time dependent manner which depends upon the magnitude of the first stress state. But now consider the situation that exists at an arbitrarily small interval of time after the sudden application of the second stress state. The material not only experiences the instantaneous response to the second change in surface tractions but also it experiences a continuing time dependent response due to the first applied level of stress. An elastic material would respond only to the total stress level at every instant of time. Thus, this more general type of material possesses a characteristic which can be descriptively referred to as a memory effect. That is, the material response is not only determined by the current state of stress, but is also determined by all past states of stress, and in a general sense; the material has a memory for all past states of stress. A similar situation exists if one considers the deformation as being specified, and thus, the current stress depends upon the entire past history of deformation.

    It is this latter observation that materials can have a capacity for memory which will be given a simple but fundamental mathematical characterization in the next section. Thereafter a representation theorem will be used to form the linear viscoelastic stress strain constitutive relation, all under isothermal conditions. In so doing, the use of the term memory will be made more precise; however, at this point, it is well to take note of the fact that there are other theories of mechanical behavior of materials which have a memory of deformation but which are different in some fundamental way from what will be considered here. For example, the incremental theory of plasticity has memory effects in as much as a final state of deformation depends not only upon the final state of stress, but also upon the path in stress space traversed to reach this final state. However, the underlying difference between these two theories is that the plasticity theory is independent of the time scale involved in loading and unloading programs while the viscoelastic theory has a specific time or rate dependence.

    All derivations and applications in this and the succeeding chapters assume conditions of homogeneity. In many cases the extension to certain types of inhomogeneous materials is easily obtained, while in other cases the extension is difficult if not impossible. Thus, it is necessary to consider individually the extension of methods and analyses for homogeneous viscoelastic materials to inhomogeneous materials.

    Finally, note should be taken of the fact that even though most of the developments in viscoelasticity theory are recent the basic linear and isothermal field theory formulation has been available for a much longer time. While there were several early contributors, such as Maxwell, Kelvin, and Voigt, Boltzmann [1.1] in 1874 apparently supplied the first formulation of a three-dimensional theory of isotropic viscoelasticity, while Volterra [1.2] obtained comparable forms for anisotropic solids in 1909.

    1.2. INTEGRAL FORM OF STRESS STRAIN CONSTITUTIVE RELATIONS, STIELTJES CONVOLUTION NOTATION

    The formulation of the isothermal viscoelastic stress strain constitutive relations will now be considered. The statement of the other relevant field equations needed to form a complete theory will be deferred until the next chapter. First, the definitions of stress and strain will be briefly stated. For more detailed information reference should be made to elasticity texts, such as Sokolnikoff [1.3].

    The usual Cartesian tensor notation is employed with Latin indices having the range 1, 2, 3 and repeated indices imply the summation convention. Let the coordinates of a point in the body in a fixed reference configuration be denoted by Xi, referred to rectangular Cartesian axes. In the case of solids the fixed reference configuration is taken to be the undeformed configuration. Let xi denote the deformed configuration coordinates of the same point. The complete history of the motion of the continuum is specified by

    Xi(τ) = xi(Xjτ), — ∞< τ t

    where τ is the time variable and t denotes current time.

    The components of the displacement vector are defined by

    ui(τ) = Xi(τ) —Xi.

    A measure of the deformation is supplied by

    where &ij is the Kronecker symbol. Let be defined by

    where the notation ui,j = ∂i,j/Xj 1 and if the displacements of the body of continuum are small compared with all characteristic dimensions of the body. Under these conditions the infinitesimal strain tensor eij is defined by

    where it is immaterial whether the differentiation is with respect to the xi or Xi coordinates because the displacements are infinitesimal. Henceforth, in the infinitesimal theory, the derivatives will be written with respect to the xi coordinates.

    Stress is defined in the following way. Let δa be the area of an element in a surface which may either designate the boundary of the body, or may simply be a surface inside the body. Let n denote a unit vector normal to the infinitesimal surface element δa. The stress vector σ is defined in terms of the resultant force G acting on the surface element δa through

    This stress vector is defined on the side of δa corresponding to the positive direction of n. For each different orientation of the surface element δa, there will be a different stress vector σ. The stress tensor σij is defined through the transformation which relates the components of the stress vector to the orientation of the surface element; that is,

    σi= σjinj.

    This relationship is obtained by the application of the balance of linear momentum to a small tetrahedron. Symmetry of the stress tensor σij = σji follows from the assumed balance of angular momentum applied to a small volume element. In the present infinitesimal theory, the area δa may be taken relative to either the reference or the deformed configuration with only negligible differences, since the displacements between these two states are of infinitesimal order. This is in contrast to the general theory presented in Chapters 8 and 9, in which none of the preceding smallness assumptions are involved.

    ij which is being sought. This relationship is assumed to be linear, which is consistent with the smallness assumptions already introduced. The infinitesimal theory is said to be linear, when in addition to the linear relations already noted, the additional field equations, which are needed to comprise a complete theory, are also linear. These additional relations are stated in Chapter 2.

    The hypothesis that the current value of the stress tensor depends upon the complete past history of the components of the strain tensor is formally expressed as

    into a corresponding stress history σij(tkl(t)y which corresponds to the instantaneous elasticity effect mentioned in the introduction. In general, all field variables are not only functions of time, but also are functions of position xiij upon xi is suppressed.

    ij(t) is assumed to be continuous and the functional is assumed to be linear, the Riesz representation theorem [1.4] may be employed to write the functional in (1.1) as a Stieltjes integral, giving

    where the integrating functions Gijkl(t) comprise a fourth order tensor such that Gijkl(t) = 0 for — ∞ < t < 0 and each component is of bounded variation in every closed subinterval of —∞ <t < ∞. Actually, the integral in (1.2) is the convolution form of a Stieltjes integral. This form implies that the constitutive relation (1.2) is independent of any shifts in the time scale. Such behavior is known as time translation invariance, and all results in this book assume this condition. The symmetry of the stress and strain tensors imply the following relations:

    ij(t) = 0 for t < 0 and assuming that Gijkl(t) t < ∞, then (1.2) can be written in the following form:

    It is seen that (1.4) can be thought of as being obtained from (1.2) through the integration of the Dirac delta function involved in the differential of the integrating functions Gijkl(t) at t=0. Alternatively, (1.4) can be obtained from (1.2) through the connection between Stieltjes and Riemann convolutions. Gurtin and Sternberg [1.5] recognize this latter line of reasoning to be the more rigorous and follow it in their derivation.

    Another form of the stress constitutive relation can be obtained from (1.4) through the change of variable τ = t s, and an integration by parts to form

    kl(t) ij(t) which has a step discontinuity at t = 0 into a uniformly convergent sequence of continuous functions with the resulting generalization of (1.5) as

    The derivation given in Ref. [1.5] thus avoids involvement with delta functions; otherwise, ij(t) = 0 for t ij(t) → 0 as t → ∞.

    The stress strain relation (1.5) is one form of the general viscoelastic constitutive relations. The integrating functions Gijkl(t) are mechanical properties of the material and are termed relaxation functions. The determination of relaxation functions will be considered in Section 1.5 and Chapter 4. The stress strain relation (1.5) is seen to follow from nothing more than the memory hypothesis, the smoothness assumptions, and the mathematical representation theorem. It has not been necessary to appeal to physical intuition or model representations in any way. This derivation will suggest the means of proceeding in the more complicated nonisothermal case given in Chapter 3. Having obtained the stress strain relation (1.5), it now can be seen to have a very simple physical interpretation. It can be considered to be the formulation of Boltzmann’s superposition principle such that the current stress is determined by the superposition of the responses to the complete spectrum of increments of strain. This point of view has been discussed extensively by Staverman and Schwarzl [1.6].

    An alternative form of the stress strain relation may be obtained by reversing the roles of stress and strain in the preceding derivation in such a way that current strain is determined by the current value and past history of stress. It is then found that

    where

    and Jijkl(t) = 0 for — ∞ < t < 0, and Jijkl(t) t < ∞. The functions Jijkl(t) are termed creep functions, and, as with relaxation functions, they represent mechanical properties of the material.

    It will be of great practical interest to have the isotropic forms of the visco-elastic stress strain relations. Consider first the tensor of relaxation functions. The most general isotropic representation of a fourth order tensor is given by the form

    where G1(t) and G2(t) are independent relaxation functions and δij is the Kronecker symbol. Following a procedure parallel to that in elasticity whereby the deviatoric components of stress sij and strain eij are introduced, then (1.5), using (1.9), reduces to

    and

    where

    In a similar way, the creep integral form of the stress strain relations has the isotropic form

    where J1(t) and J2(t) are the two independent isotropic creep functions. It is seen that G1(t) and J1(t) are the relaxation and creep functions appropriate to states of shear, while G2(t) and J2(t) are defined relative to states of dilatation.

    Obviously (1.14) and (1.15) are not relations which are independent of (1.10) and (1.11), respectively. It follows that the creep and relaxation functions J(t) and Gα(t) t < ∞ and let it be of exponential order as t —> ∞. Then the Laplace transform f(s) of f(t) is defined as

    The Laplace transform of ij=0 for t < 0.

    It follows from (1.17) that

    The comparable elasticity relationship between moduli and compliances is Jα — Gα-¹. The intuitive extension of this elasticity result to viscoelasticity would suggest Jα(t) — [Gα(t)]-¹ which is seen from (1.18) to be incorrect. However, the initial and final value theorems of the Laplace transformation can be used to show

    and for solids

    The Stieltjes convolution forms mentioned earlier in this section provide a notational convention that can be used profitably in many situations. This possibility will be discussed here and used in Sections 2.12, 5.2, and 5.3.

    Following the treatment of Gurtin and Sternberg [1.5] the stress strain relation (1.2) can be written in the form

    where the Stieltjes convolution, φ * d (t) (t) is defined by

    (t)→0 for t → — ∞ and φ(t) t < ∞. Under the further assumption that φ(t) = 0 for t < 0 then the form (1.20) can be shown to obey the commutivity relation

    Using this property, the stress strain relation (1.2) can be equivalently written as

    σij= Gijkl* d∈kl

    which of course symbolizes the form (1.5). In a similar manner the isotropic form of the stress strain relations can be written as

    The associativity and distributivity properties of Stieltjes convolutions will be useful in the future applications of this notation. These properties, under the above assumptions on φ(t) (t) state the following:

    and

    The proofs of these identities are straightforward, and can be found in Ref. [1.5].

    1.3. CONSEQUENCES OF FADING MEMORY AND THE DISTINCTION BETWEEN VISCOELASTIC SOLIDS AND FLUIDS

    As was mentioned in Section 1.2, the stress strain relations derived there were based only upon the memory hypothesis, the smoothness assumptions, and a mathematical representation theorem. It is now of some interest to examine the consequences of a further possible physical hypothesis. Specifically we seek to determine any restrictions which may be imposed upon the stress strain relations for the memory effect being restricted to be that of a particular type. The motivation for doing this comes from the extensive theoretical work which has been done upon the type of memory effect known as fading memory. An early mathematical statement of the concept of fading memory was given by Volterra [1.7]. A more recent and more complete statement of a fading memory hypothesis has been given by Coleman and Noll in several references which are mentioned in the discussion by Coleman and Mizel [1.8]. This fading memory hypothesis has been used to great advantage in several works such as Coleman’s thermodynamical formulation [1.9] of a nonlinear theory of viscoelasticity. The formal definition of this type of memory effect is more involved than is needed here for application to the infinitesimal deformation theory. Such a formal definition will be profitably used in Chapter 8 for the derivation of a nonlinear theory of viscoelasticity.

    For the present purposes of a linear theory, a simple definition of fading memory of the following type will suffice. If the current value of one field variable has a linear functional type dependence upon the complete past history of another field variable, then a fading memory hypothesis implies that current value of the first variable depends more strongly upon the recent history than upon the distant history of the second variable. More specifically, the dependence of the current value of the first variable upon the values of the second variable at previous times is determined through a weighting function which must assign a continuously decreasing dependence upon past events which are continuously more distant from the current time. The use of the term weighting function is made clear in the following restriction of the viscoelastic stress strain.relations to satisfy the fading memory hypothesis. In the stress strain relation (1.2) or (1.4), the slopes of the relaxation functions are seen to be weighting functions which characterize the extent to which the current value of each component of stress is influenced by the values of the strain at past times. For each component of the viscoelastic stress strain relations to satisfy the fading memory type behavior, it is sufficient that the magnitude of the slope of each component of the relaxation function tensor be a continuously decreasing function of time; thus,

    This fading memory hypothesis is reasonable in the sense that it would be physically unrealistic to expect materials to have a growing memory for the more distant events. In fact, all relaxation function experimental measurements have produced results in accordance with the criterion (1.25). It is a simple matter to establish the fact that the fading memory hypothesis also is satisfied by the restrictions on the components of the creep function tensor as

    Further restrictions upon the forms of relaxation functions are given in Section 3.3.

    The other item to be discussed in this section concerns the distinction between viscoelastic solids and fluids. While it may be intuitively obvious that an elastic medium is a solid, and a viscous medium is a fluid, the situation for viscoelastic materials is less clear since it has some elements of both elastic and viscous behaviour. It is not obvious whether a viscoelastic material should be a solid or a fluid or could be either. In fact it can be either, as will be shown. In a general sense the distinction between solids and fluids is not as simple as it might at first seem to be. The rigorous definition of each, as given by Truesdell and Noll [1.10] is fairly involved and will be considered in Chapter 8. A consequence of the formal definition of a fluid is that a fluid must necessarily be isotropic, in contrast to solids. For our present purposes we will distinguish between isotropic solids and fluids in the following simple and nonrigorous physical manner. A viscoelastic fluid, when subjected to a fixed simple shear state of stress, will respond with a steady state of flow, after transient effects have died out. Also, a viscoelastic fluid, when subjected to a fixed simple shear state of deformation, will produce a stress state which will eventually decay to zero. Contrary to this, an isotropic viscoelastic solid when subjected to a simple shear state of deformation, will have a corresponding component of stress which remains nonzero as long as the state of deformation is maintained. In other words, the viscoelastic fluid has an unlimited number of undeformed configurations while the viscoelastic solid may have only one.

    The implications relative to the mechanical properties of the distinction between viscoelastic solids and fluids will now be determined. In the remainder of this section it will be convenient to distinguish between the coordinates of the particles in a particular fixed reference state, and the coordinates of the particles in any state of possible deformation. This is similar to the situation outlined at the beginning of Section 1.2 where the coordinates of a particle in the reference and deformed configurations were denoted by Xi and xi = Xi(τ, Xj), respectively. This convention is in contrast to our usual infinitesimal theory notation that xi are the coordinates in a fixed reference state.

    Consider a viscoelastic material which is subjected to a simple shear state of deformation specified by the displacement components from the fixed reference configuration as

    where h(i) is the unit step function. Using this in the strain displacement relations

    then (1.10) gives the only nonzero relation between the components of stress and strain as

    where it is recalled that

    G1(t) = 0, t <0.

    It follows from the definition of an isotropic viscoelastic solid that for an isotropic viscoelastic material to be a solid it is necessary and sufficient that

    nonzero constant (isotropic solid).

    For a viscoelastic material to be a fluid it is necessary that

    (fluid).

    However, this requirement on the limiting behavior of the relaxation function in shear is not sufficient for a material to be a fluid. As we will next show, sufficiency requires that the material satisfy the steady state flow requirement.

    Since a viscoelastic fluid has the property to allow flow in a steady state condition, we should be able to derive a relevant coefficient of viscosity. The existence of such a coefficient of viscosity, in steady flow, ensures that the material is a fluid. In obtaining this coefficient of viscosity, we begin by first recalling the appropriate stress constitutive relation for a Newtonian viscous fluid, given by

    and where xi(t) are the coordinates at current time t and xi(t) are the components of velocity. Note the distinction between the reference configuration used here and the fixed reference configuration used in an infinitesimal theory. Here the reference configuration is the current configuration xi(t), while in an infinitesimal theory the reference configuration, given by coordinates Xi , is fixed and for a solid is normally taken to be the undeformed configuration. In a general (non-infinitesimal) state of deformation, the rate of deformation tensor dij given by (1.27), and the rate of the infinitesimal strain tensor,

    are completely different. But, in the special case of simple shear flow, they can be shown to be the same. This state of flow is specified by

    To specify the state of simple shear flow in terms of displacements requires the form

    ui(t) = xi(t) –Xi

    which defines the components of displacement in terms of the coordinates. Then the flow state is equivalently defined by

    With these relations and ij and dtj are identical. Thus, the state of simple shear flow will be used to determine the equivalent coefficient of viscosity for a viscoelastic fluid, in terms of its relaxation function.

    Using the specified displacements (1.29) in (1.10) gives

    Using the same state of deformation (1.28) in (1.27) for a Newtonian viscous fluid gives

    The steady state stress will be achieved by the viscoelastic fluid at large values of time. Thus, a comparison of (1.30) and (1.31) at large values of time gives

    It follows that if a viscoelastic fluid, in a state of simple shear flow, is subjected only to steady rates of deformation, it will behave exactly as a Newtonian viscous fluid with a coefficient of viscosity given by (1.32), where the integral is assumed to exist. Actually, the use of a coefficient of viscosity defined by (1.32) not only implies a steady rate of deformation of the viscoelastic fluid, but it also implies a vanishingly small rate of deformation. This is necessary for the relation (1.30), which is based upon the infinitesimal theory, to be true at large values of time, as is implied by (1.32). The viscosity determined by (1.32) is called the zero shear rate viscosity. This nonrigorous examination of the flow characteristics of viscoelastic fluids under steady flow conditions will be superseded by the general nonlinear viscoelastic fluid treatments given in Chapters 8 and 9.

    For a viscoelastic fluid the creep function in shear has the form

    where J1(t) approaches a finite positive asymptote for large values of time and n is the zero shear rate viscosity defined in (1.32). The derivation of this form is included as an exercise.

    For polymeric materials, the distinction between solids and fluids is very simple when considered on a molecular scale. In the case of fluids, the individual long chain molecules are completely unconnected and over long time spans have unlimited mobility with respect to each other. But for solids, there are discrete chemical bonds between adjacent molecules which are called cross links and which prevent unlimited flow. However, in the case of a very lightly cross linked material, the distinction between a solid and a fluid may be almost unnoticeable insofar as the methods and techniques suggested by an infinitesimal theory are concerned.

    We now have the means to distinguish between viscoelastic solids and fluids. In the later applications, both types of materials will be considered. However, it is necessary to remember that in applying stress strain relations of the type discussed in this chapter, whether they represent solids or fluids, their application must not in general violate the assumptions of infinitesimal deformation, measured with respect to a fixed reference configuration. In this sense, the distinction between solids and fluids for application in an infinitesimal deformation theory is not involved with differences in methods and techniques of solving problems; rather, the distinction is more important in assessing the appropriateness of applying the theory to particular types of problems. Thus, since the linear theory cannot be used to solve general (noninfinitesimal deformation) flow problems for viscoelastic fluids, its main usefulness in the case of fluids is in deriving the appropriate forms of the constitutive relations, which are widely determined and interpreted experimentally.

    1.4. DIFFERENTIAL OPERATOR FORM OF STRESS STRAIN CONSTITUTIVE RELATIONS

    The creep and relaxation integral forms of the stress strain constitutive relations are by no means the only possible forms. Two other forms will be given; the first given here involves differential operators, the second will be given in Section 1.6.

    Relevant to isotropic materials, we consider the following differential operator form of a relation between the deviatoric components of stress and strain:

    Or, more compactly, we write this as

    where

    with D designating the operator d/dt. Relation (1.34) is certainly a possible relationship between stress and strain, but at this point it is not clear whether it has any significance for viscoelasticity. To investigate this possibility it is helpful to see if any connection between (1.34) and (1.10) can be established. To this end we take the Laplace transform of (1.34), using the relationships on the transforms of derivatives, then (1.34) becomes

    where

    and sij(k-r)(0) designates the k r order derivative of sij(t) evaluated at t = 0, with a similar relationship for eij(t). The Laplace transform of (1.10) is given by the first equation in (1.17). This is seen to be equivalent to (1.36) if

    and

    Equation (1.39) supplies a requirement upon the initial conditions; thus, the initial conditions upon stress and strain are not completely independent, and relations such as (1.39) must be satisfied.

    Relation (1.38) gives the conditions under which the relaxation integral and differential operator forms of the deviatoric stress strain relation are equivalent. In an entirely similar manner, the independent dilatational part of the isotropic stress strain relations can be written in

    Enjoying the preview?
    Page 1 of 1