Вы находитесь на странице: 1из 49

Mechanics of Time-Dependent Materials 6: 351, 2002. 2002 Kluwer Academic Publishers. Printed in the Netherlands.

Poissons Ratio in Linear Viscoelasticity A Critical Review


N.W. TSCHOEGL
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, CA 91125, U.S.A.; E-mail: nwt@caltech.edu

WOLFGANG G. KNAUSS and IGOR EMRI


Division of Engineering and Applied Science, California Institute of Technology, Pasadena, CA 91125, U.S.A.; E-mail: {wgk,igoremri}@caltech.edu (Received 1 June 2001) Abstract. Poissons ratio is an elastic constant dened as the ratio of the lateral contraction to the elongation in the innitesimal uniaxial extension of a homogeneous isotropic body. In a viscoelastic material Poissons ratio is a function of time (or frequency) that depends on the time regime chosen to elicit it. It is important as one of the material functions that characterize bulk behavior. This paper develops the linear theory of the time- or frequency-dependent Poissons ratio, and it reviews work on its experimental determination. The latter poses severe difculties in view of the high accuracy required. Thus, reliable information on the viscoelastic Poissons ratio is as yet rather scanty. The paper also reports on attempts to measure the Poissons ratio of a viscoelastic material as a function of temperature. Lateral contraction in creep and at constant rate of extension receives attention as well. Key words: bulk behavior, frequency dependence, lateral contraction, material characterization, Poisson strain and stress, Poissons ratio, time dependence, temperature dependence, viscoelastic behavior

1. Introduction In the innitesimal deformation of an idealized purely elastic compressible material one may dene a time-independent material constant, called Poissons ratio, as the ratio of the lateral contraction to the elongation in an innitesimally small uniaxial extension (Eringen, 1962). In the innitesimal deformation of any real, i.e., viscoelastic, material, the lateral contraction is a time- or (equivalently) frequencydependent material function (Tschoegl, 1989, p. 528). Poissons ratio generated in response to a strain as a step function of time is then one of several viscoelastic response functions. This paper is concerned with the theory of this ratio, and with
Permanent address: Center of Experimental Mechanics, Faculty of Mechanical Engineering, University of Ljubljana, 1125 Ljubljana, Slovenia.

N.W. TSCHOEGL ET AL.

its experimental determination. It considers the viscoelastic Poissons ratio for the innitesimal deformation of a homogeneous isotropic material. Because of considerable experimental difculties, information on this function is still rather scanty. In addition, even this scant literature is replete with reports of work based on incorrect equations. One of the aims of this article is, therefore, to provide a denitive discussion of the theory. There are practical as well as theoretical aspects that make it highly desirable to stimulate the experimental determination of the viscoelastic Poissons ratio on a secure analytical foundation. 1.1. P RACTICAL A SPECTS Multicomponent injection molding is becoming an increasingly important technological process for manufacturing large structural elements. This technology allows the integration of several structural components, and thus the shortening of the manufacturing process, by excluding the time-consuming assembly phase. One of the main problems related to the integration of structural elements is the quality (the degree) of adhesion between two particular components. In injection molded objects the boundaries between two components are known as weld lines. Weld lines pose problems when two materials do not adhere well to each other (Fellahi et al., 1995; Karger-Kocsis and Csikai, 1987; Fisa et al., 1990). Such imperfections may seriously affect the objects performance. Examination in shear and in uniaxial tension of the degree of adhesion in weld lines formed between different soft and hard materials shows that the quality of adherence is satisfactory if the moduli of the two materials are identical or closely similar. In attempts to bond a soft and a hard material, this condition is not fullled. In this case one witnesses the development of the phenomenon known as Poisson strain, resulting from the difference in the lateral contraction of the two materials (Kinloch, 1990). The shear stress generated by this strain has its maximum on the boundary of the weld line (Adams and Pepiatt, 1973; Harris and Adams, 1984) causing crack initiation at or near that point (see also Gent and Hwang, 1988). Thus, evaluation of the strength of adhesion between a soft and a hard material and, consequently, the likelihood of the success of a multicomponent extrusion operation, requires knowledge of the lateral contraction of the two materials on both sides of the weld line. 1.2. T HEORETICAL A SPECTS Besides the eminently important practical considerations brought up above, there are other, theoretical, aspects to the desirability of determining the viscoelastic Poissons ratio. If it is known together with one other of the linear viscoelastic response functions of an isotropic material, any other such function can, at least in principle, be obtained indirectly from these two. In this way it is then possible again in principle to determine the time-dependent bulk modulus, K(t), and hence also the time-dependent bulk compliance, B(t) (see Tschoegl and

POISSONS RATIO IN LINEAR VISCOELASTICITY

Emri, 1992). The experimental determination of the time-dependent bulk response functions is fraught with considerable difculty. Since both functions vary only relatively little with time, it is common practice to resort to the shortcut of simply substituting numerical constants. Thus, instead of determining the functions, K(t) or B(t), one then establishes only the glassy and the equilibrium bulk moduli, Kg and Ke , or the bulk compliances, Bg and Be . Even the determination of these numerical constants experimentally can be avoided by formulating the stress-strain relations (see Section 3.1.2) in terms of the time-dependent Poissons ratio, (t). One may then simply substitute numerical constants such as 1/3 for g and 1/2 for e . However, such short cuts neglect the time dependence of either bulk response function and this may introduce potentially severe inaccuracies into the sought-for solution (Knauss and Emri, 1981). Thus, satisfactory experimental determination of the principal time-dependent bulk functions, the bulk modulus, K(t), the bulk compliance, B(t) the longitudinal bulk modulus, M(t), or Poissons ratio, (t) remains highly desirable. Because of the experimental difculties involved in determining the bulk response functions directly, numerous attempts have been made to obtain them indirectly from two other, more easily accessible ones, such as, e.g., the time-dependent shear and stretch moduli, G(t) and E(t), or the shear and stretch compliances, J (t) and D(t). Our review of the relevant literature (cf. Section 4) reveals, unfortunately, that such attempts have never been successful. This failure appears to be due ultimately to the relatively small change in value of the bulk functions over the complete gamut from zero to innite time or frequency. While the magnitude of the shear and tensile moduli, G(t) and E(t), and compliances, J (t) and D(t), generally varies in magnitude over three to four and even ve decades, the bulk functions, K(t) and B(t), vary only by about a factor of two or three. Figure 1 shows plots of the logarithms of G(t), E(t), K(t) and M(t) for the Standard Linear Solid (Tschoegl, 1989, p. 531). These plots are typical of the characteristic behavior of any amorphous homopolymer or random copolymer. The determination of K(t) from E(t) and G(t) thus resembles the attempt to determine the weight of the captain by weighing the ship with and without him. Clearly, highly accurate measurements of the source functions are needed to calculate K(t) from E(t) and G(t), or B(t) from D(t) and J (t) successfully. Even worse, the entire variation of (t) is compressed effectively between the values of about 0.25 and less than 0.5. Its determination by calculation from any two other viscoelastic functions appears therefore to be even more exacting. The successful, highly accurate, simultaneous determination of two viscoelastic response functions is of great theoretical interest. The data could be stored in a convenient way and from them any desired linear response could be generated with ease because it would make possible the interconversion of any two material functions into any other. It would therefore allow the complete characterization of the linear viscoelastic behavior of a given isotropic material in a single experiment.

N.W. TSCHOEGL ET AL.

Figure 1. Comparison of K(t), M(t), G(t), and E (t) for the standard linear solid.

In addition, this would open further intriguing vistas. Thus, it would nally become possible to compare bulk and shear relaxation or retardation behavior unambiguously. Currently, there are few measurements of the response in shear and in bulk on the same material under identical conditions. Earlier data by Kono (1960, 1961a) on polystyrene and on poly(methyl methacrylate) were reported as functions of temperature, not time or frequency. Morita et al. (1968) reported similar work on poly(isobutyl methacrylate). They found that the bulk relaxation showed a relatively narrow distribution of relaxation times compared with that of the shear relaxation. Kono and Yoshizaki (1973) observed essentially the same on poly(isobutyl ether). Crowson and Arridge (1979a, b) determined J (t) and B(t) on the same epoxy but in separate experiments and then crossplotted the data to obtain J (T ) and B(T ). Concerning the relationships between shear and bulk relaxation time or retardation time distributions, Deng and Knauss (1997), and Sane and Knauss (2001) have more recently reported measurements of the bulk compliance of a PVAc (poly(vinyl acetate)) sample. The authors comparison of these with shear compliance measurements on a different specimen of the same PVAc sample produced an indication that the breadths of the transition zones in bulk and in shear are closely similar but that the bulk compliance transition is shifted to frequencies about three decades higher than those of the shear compliance transition. The former thus appears at shorter retardation times than the latter. These ndings are at variance with those listed in the preceding paragraph. The issue clearly requires further examination.

POISSONS RATIO IN LINEAR VISCOELASTICITY

The viscoelastic Poissons ratio is characterized by a distribution of delay times (see Section 3.1.4) that cannot be assumed a priori to be identical with either the retardation or the relaxation times. At this time nothing is known concerning the distribution of Poissons ratio delay times in comparison with the distribution of retardation times in shear and/or in bulk. 1.3. T HE S TANDARD P ROTOCOL Our review of the literature has convinced us that the determination of any of the bulk functions by calculation from any other two requires that the source functions be obtained following a strict protocol. This protocol demands that the source functions be determined simultaneously on the same specimen, under the same conditions of the experimental environment. This guarantees identical initial and boundary conditions. To this we add the requirement of high accuracy and precision. We shall refer to this protocol as the Standard Protocol and, because of its importance, shall refer to it repeatedly in this paper. It is mandatory to avoid grave errors that are able to arise from even the smallest specimen-to-specimen variation in material properties, and even the smallest differences in the environmental variables, mainly temperature, pressure, and humidity. These requirements place severe but in all likelihood not insurmountable conditions on the experimental determinations of the source functions from which Poissons ratio is to be calculated. There are not many experimental possibilities for the simultaneous measurement of two time-dependent functions as required by the Standard Protocol. Measurements of the transmission of waves in the longitudinal and transverse direction appear to conform to it but are applicable only at high frequencies. This method has been used, for instance, by Kono (1960a, b), Waterman (1963a, b), Morita et al. (1968), Kono and Yoshizaki (1973), and by Husler et al. (1987). These authors necessarily reported isochronal measurements of Poissons ratio as functions of temperature (see Section 4.32). Simultaneous isothermal measurements are still lacking. The determination of the time-dependent Youngs modulus, E(t), simultaneously with the concomitant lateral contraction appears to offer one of the desired possibilities for determining the time-dependent Poissons ratio, (t). At this time promising research based on this approach is under way at the Center for Experimental Mechanics of the University of Ljubljana (Samarin et al., 2002). The Appendix outlines the determination of the bulk relaxation modulus, K(t), from measurements of E(t) and (t), and it also indicates the way in which any other of the response functions can be generated from the same data.

8 2. The Elastic Poissons Ratio

N.W. TSCHOEGL ET AL.

The purpose of this section is to present Poissons ratio as developed in elasticity theory since this forms the basis of any consideration of the linear viscoelastic Poissons ratio. Section 3 will extend the formulations to linear viscoelastic theory. 2.1. F UNDAMENTAL AND D ERIVED E LASTIC C ONSTANTS In the innitesimal deformation of an isotropic elastic material changes in size and changes in shape are neatly separated. The response of such a material to changes in size are expressed by the bulk modulus, K , while the response to changes in shape are represented by the shear modulus, G. The reciprocals of these moduli are the bulk compliance, B = 1/K , and the shear compliance, J = 1/G. These two moduli and compliances may therefore be recognized as the fundamental elastic constants of an isotropic elastic material in innitesimally small deformations. There are many other modes of deformation that comprise simultaneous changes in both size and shape. An elastic constant can be dened for any of these deformations. These elastic constants may be called derived elastic constants. Any elastic constant can, in principle, be obtained from any two others and, in particular, any derived elastic constant can be related to the fundamental ones. In the longitudinal bulk modulus, 4 M = K + G, 3 1 , 1 (2.1)

the size and shape contributions are simply additive. The modulus is dened as M= 2 , 3 = 0, (2.2)

where 1 is the stress applied in the 1-direction, and 1 is the extension in the same direction, deformation in the other two lateral dimensions being prevented. M is the modulus one measures in uniaxial compression when expansion in the two transverse directions is disallowed as it may be, e.g., in wave propagation, or, if the material is compressed in a thick-walled cylinder that prevents any lateral expansion. By contrast, the deformation in uniaxial tension is described by the stretch (or Youngs, or tensile) modulus, E , dened as E= 1 , 1 2 , 3 = 0. (2.3)

Here, no stress acts in the two transverse directions, and the deformations in these directions are thus free. In terms of the two fundamental moduli E becomes E= 9KG , 3K + G (2.4)

POISSONS RATIO IN LINEAR VISCOELASTICITY

where the size and shape contributions are not additive. We note however, that in its reciprocal, the stretch (or tensile) compliance, D = 1/E , the contributions of the shear and bulk compliances are additive since we have 1 1 D = J + B. 3 9 (2.5)

We illustrate the point that an elastic constant can be dened for any mode of deformation, by dening a modulus in equibiaxial tension, H , as the ratio of the stress to the strain in the same direction when a stress of equal magnitude is applied in the transverse direction also. We then have H = 18KG , 3K + 4G 1 = 2 , 3 = 0. (2.6)

While there does not appear to be any need for this modulus in practical applications, it is worth pointing out that for an incompressible rubber (i.e., one for which K is very much larger than G) equibiaxial extension is equivalent to uniaxial compression. This is intuitively obvious. In the rst case, the extensions are prescribed and no force is applied in the direction normal to them. In the second, a compressive force is applied in one direction and deformations are allowed in the transverse directions. 2.2. T HE L ATERAL C ONTRACTION R ATIO When measuring the stretch modulus, the uniaxial extension is accompanied by a contraction in the two transverse directions since the lateral expansion is free. The ratio between the lateral contraction, 2 = 3 , and the uniaxial stretch, 1 , is referred to as Poissons ratio, , which is thus dened by = 3 2 = . 1 1 (2.7)

In terms of the fundamental moduli, K and G, and in terms of the fundamental compliances, B and J , it is given by = 3J 2B 3K 2G = . 6K + 2G 6J + 2B (2.8)

Alternatively, we have = and = J E 1= 1. 2G 2D (2.10) E 1 B 1 = 2 6K 2 6D (2.9)

10

N.W. TSCHOEGL ET AL.

For good measure we add the expression for the Poissons ratio in terms of the longitudinal bulk modulus, M , and the shear modulus, G. = M 2G . 2(M G) (2.11)

If the material may be deemed incompressible, i.e., in the limit as K approaches innity, or B approaches zero, Poissons ratio becomes 0.5 as is easily ascertained from Equation (2.8). Thus, we have
K

lim = 0.5,

(2.12)

and this is the maximum value that the ratio can ever assume in an innitesimal uniaxial stretch whether the ratio is elastic or viscoelastic (time-dependent). A Poissons ratio of 0.5 corresponds to an innite bulk modulus and is thus never attained in practice. While no material is truly incompressible, K = may be a reasonable approximation when K G. It is shown in elasticity theory (see, e.g., Malvern, 1969) that in an isotropic material the range of Poissons ratio is given by 1 0.5. It can thus take on even negative values, representing an expansion, rather than a contraction, in the transverse directions. As pointed out before, the denition of derived elastic constants is a matter of convenience. There is nothing unique about the stretch modulus, E or the (elastic) Poissons ratio, . We may, for instance, dene a lateral contraction ratio in equibiaxial tension, H , as the ratio of unit stretch in biaxial tension to the contraction in the free direction, i.e., by H = 3 , 1 2 = 1 . (2.13)

In terms of the fundamental moduli we then have H = 6K 4G , 3K + 4G (2.14)

and, for an incompressible material, this would give a biaxial Poissons ratio of 2. Finally, because we will need their viscoelastic counterparts in the Appendix, we list the relations linking the two fundamental moduli, the shear modulus, G, and the bulk modulus, K , to the tensile modulus, E , and the Poissons ratio, . We have K= E 3(1 2) and G= E 2(1 + ) (2.15)

[cf. Equations (3.22)].

POISSONS RATIO IN LINEAR VISCOELASTICITY

11

2.3. L ATERAL C ONTRACTION IN I NHOMOGENEOUS AND IN A NISOTROPIC M ATERIALS The foregoing discussion of the elastic Poissons ratio dealt exclusively with homogeneous isotropic materials in innitesimally small deformation. In inhomogeneous isotropic materials, such as polymers containing particulate llers, Poissons ratio may be greater than 0.5. Below the melting region semi-crystalline polymers by and large behave like lled ones, the crystalline regions acting as llers and the amorphous zones constituting the matrix. On the other hand, in foams the lateral contraction may assume very low values (Blatz and Ko, 1962; Weber et al., 1997). Filled and reinforced materials may exhibit the phenomenon of dewetting (i.e., de-bonding between ller and matrix), resulting in the formation of vacuoles in any but the smallest deformations (see, e.g., Kimoto et al., 1990). In dewetted materials one may experience lateral contraction ratios that are not limited to 1 0.5. Values greater than 0.5 are possible. The latter generally arise from a structural instability in the material. Anisotropic materials (such as, e.g., ber-reinforced polymers, which are, of course, necessarily also inhomogeneous) require a larger number of elastic constants for a full characterization than the two required for isotropic materials (see, e.g., Tschoegl, 1989, pp. 13ff). More than a single lateral contraction ratio may be dened in such materials depending on the type of anisotropy. Thus, axisymmetric (also called transversely isotropic) materials such as drawn bers, exhibit an axis of symmetry normal to a plane of symmetry and require ve constants. Orthotropic materials such as, for instance, many polymer lms, possess three orthogonal planes of symmetry as their symmetry elements. These materials require nine independent elastic constants, for example, three stretch moduli, three shear moduli, and three Poissons ratios for a full characterization. It is not the purpose of this paper to discuss in detail this rather complicated topic on which a considerable literature exists. For a brief review see, e.g., Summerscales and Fry (1994). We mention a few representative publications. Darlington and Saunders (1970) simultaneously measured the longitudinal and lateral strain on highly oriented (axisymmetric) polyethylene. Clayton et al. (1973) made similar measurements. Ward and his collaborators (Ladizesky and Ward, 1971; Wilson et al., 1976a, b; Richardson and Ward, 1978; Zihlif et al., 1982) used a variety of methods to determine Poissons ratios on isotropic and anisotropic polyethylenes (LDPE and HDPE) and poly(ethylene terephthalate). Kimoto et al. (1990) used holographic interferometry to study Poissons ratio as a function of the orientation in ber-reinforced plastics.

12 3. The Viscoelastic Poissons Ratio

N.W. TSCHOEGL ET AL.

Poissons ratio as dened in Section 2 is an elastic constant, as are all the other quantities mentioned there. They all are, therefore, dened for a (hypothetical) material whose mechanical properties show no time dependence. Such a material is hypothetical because in the strict sense of the word it does not exist. In any mechanical deformation the deformational energy is not only stored elastically but part of it is invariably dissipated by viscous forces in accordance with the second law of thermodynamics. This dissipation is responsible for the time dependence of the mechanical properties in any real material. Real materials are therefore viscoelastic. The theory of elasticity can be applied to a real material only on the assumption that the time dependence of its mechanical properties can be neglected. This is often the case in practice because in certain materials (e.g., metals and most elastomers) the time dependence may not be marked. This is the reason for the widespread use of elastic constants to describe mechanical material behavior, and this is the case particularly with the behavior in bulk. In many applications, however, the inherent time dependence of a material cannot be disregarded. When measured on a real, i.e., viscoelastic material, all elastic constants (i.e., the moduli, the compliances, and also Poissons ratio), become material functions of time or, equivalently, functions of frequency (log t log f ). We reserve the term time-dependent Poissons ratio for the lateral contraction ratio measured in an innitesimally small uniaxial deformation of a viscoelastic material in response to a strain, 1 (t) = 0 h(t), as a unit step function, h(t), of time. Thus, (t) = 2 (t)/0 where 0 is the imposed constant strain and 2 (t) is the time-dependent lateral contraction in the transverse direction. The proper derivation of this relation will be given in Section 3.2 [cf. Equation (3.35)]. In Sections 3.7 and 3.6 we will show how (t) can be obtained from experiments in creep and at constant rates of deformation. It needs emphasis that, as the stress in uniaxial extension gradually relaxes the development of the nal value of the contraction is delayed (retarded). Consequently, the time-dependent Poissons ratio has the character of a compliance. In its behavior, therefore, a relaxation process in the axial direction is accompanied by retardation processes in the transverse direction. In an analogous manner we refer to a (complex) frequency-dependent Poissons ratio as the lateral contraction ratio measured in an innitesimally small uniaxial deformation of a viscoelastic material in response to a steady-state sinusoidally oscillating strain. Ratios not conforming to either of these two descriptions will be referred to simply as lateral contraction ratios. 3.1. R ELATIONS IN THE T RANSFORM P LANE Because the linearly viscoelastic relations are functions of time or, equivalently, of frequency, their explicit forms depend on the type of excitation under consideration. When we are concerned, e.g., with material functions of time, we may have to deal with impulse, step, ramp, or other excitations (Tschoegl, 1989, chapter A2).

POISSONS RATIO IN LINEAR VISCOELASTICITY

13

Manipulations involving the material functions of the linear theory of viscoelasticity are greatly facilitated by formulating problems in the complex Laplace transform plane. The transform, f(s), of the function f (t) is given by

f(s) =
0

f (t) exp(st) dt = L[f (t)],

(3.1)

where L[ ] denotes the Laplace transform operator . The transform variable, s , is complex in general, i.e., s = a + j where j = 1. We are, however, normally interested in either time-dependent or frequency-dependent viscoelastic response functions. In the rst case the functions lie along the real axis of the transform plane while they lie along the positive imaginary axis of the plane in the second case. The two cases are obtained essentially by letting 0 in the rst, and letting a 0 in the second case. After the appropriate manipulation we retransform the complex plane relations either to the real time axis, or to the imaginary frequency axis, whichever is desired. To work in the transform plane we invoke the so-called correspondence (or equivalence) principle (see Lee, 1960). It is certainly not admissible to simply substitute a time-dependent viscoelastic function in place of an elastic constant in a relation supplied by the theory of elasticity. While, e.g, G = 1/J , the time-dependent moduli and compliances are not one anothers reciprocals, i.e., G(t) = 1/J (t) However, the s-multiplied Laplace transforms, called the Carson transforms, are reciprocals of one another. We have s G(s) = or J(s) = 1 , G(s) s2 and retransformation yields the two equivalent convolution integrals
t t

1 , s J (s)

(3.2)

(3.3)

G(t u)J (u) du =


0 0

G(u)J (t u) du = t.

(3.4)

Functions like these are therefore sometimes referred to as being time reciprocals of one another. 3.1.1. The Correspondence Principle According to this principle, if an elastic solution to a stress analysis problem is known,1 substitution of the appropriate complex-plane transforms for the elastic quantities supplies the viscoelastic solution to the same problem in the transform

14

N.W. TSCHOEGL ET AL.

plane. Thus, the stress-strain relations of the theory of elasticity become relations between the Laplace transforms of the stress and strain in which the moduli are replaced by the corresponding relaxances, the Laplace transforms of the impulse response functions. The changes most relevant for our present purpose are i (s), i and G Q(s), (s), K P and (s), EY (3.5) (s) is the bulk, where s is the Laplace transform variable, and Q(s) is the shear, P (s) is the stretch relaxance. and Y When dealing with compliances instead of moduli, we substitute the retardances according to (s), J U B A(s), and D L(s), (3.6) (s) is the shear, A(s) where now U is the bulk, and L(s) is the stretch retardance. There is, in addition, the substitution (s), (3.7) i i (s),

where (s) may be called the Poisson retardance. Since in this article we are interested primarily in the response to the imposition of a unit step function of time, we must replace the relaxances and retardances by the s-multiplied Laplace transforms of the time-dependent functions using Q(s) = s G(s), (s) = s J(s), U (s) = s K(s), P A(s) = s B(s), (s) = s E(s), Y and L(s) = s D(s). (3.8)

In addition, we have s M(s) for the transform of the longitudinal bulk modulus, and we further have (s) = s (s) for the transform of Poissons ratio. To obtain step functions of time, therefore, we substitute the s-multiplied Laplace transforms (i.e., the Carson transforms) for the elastic constants. The reason for this becomes apparent upon recognizing that multiplication by s in the transform plane implies differentiation, and that the unit impulse function is , the derivative of the unit step function. Of course, the inverse transforms of G(s) K(s), E(s), J (s), B(s), D(s), and (s) are G(t), K(t), E(t), J (t), B(t), D(t), the shear, bulk, and stretch relaxation moduli and compliances, and the time-dependent Poissons ratio, (t). The Carson transform relations can often be simplied by striking the transform variable, s , on both sides of an equation, or in both the nominator and the denominator of a fraction. Thus, Equation (2.4) becomes E(s) = G(s) 9K(s) 3K(s) + G(s) (3.9)

POISSONS RATIO IN LINEAR VISCOELASTICITY

15

which looks as if the elastic constants in Equation (2.4) had been replaced by Laplace transforms while, in fact, they had been replaced by the Carson transforms, followed by simplication. The use of the Laplace transform instead of the Carson transform is the source of some incorrect formulae in the literature. When an expression contains s (s) , the transform variable s often cannot be eliminated. Poissons ratio may, in principle, be obtained from any two other linear viscoelastic response functions. Thus, in terms of the transforms of the two fundamental moduli and compliances (s) becomes (s) = 3J(s) 2B(s) 3K(s) 2G(s) = . ] ] 2s [3K(s) + G(s) 2s [3J(s) + B(s) (3.10)

The second of Equations (3.10) follows since 1/s K(s) = s B(s) and 1/s G(s) = s J(s). Below we list other useful relations between the s-multiplied Laplace transforms. Equations (2.9) yield (s) = 1 1 E(s) B(s) 1 1 = = s E(s) B(s) 2s 2s 2s 6 6s K(s) 6s D(s) 1 1 J(s) 1 1 E(s) = = s E(s) HJ (s) , s s 2 s 2s G(s) 2s D)s) M(s) 2G(s) . ] 2s [M(s) G(s) (3.11)

. From Equations (2.10) we nd because 1/s E(s) = s D(s) (s) = (3.12)

and from Equation (2.11) we obtain (s) = (3.13)

The relations in the time-domain may then be obtained from these transform relations through the inverse Laplace transformation (cf. Section 3.2.2). When dealing with a sinusoidally steady-state excitation instead of the imposition of a step function of time, the correspondence principle simply requires that the elastic constants be replaced by the corresponding complex quantities according to: i i (), and K K (), M M (), G G (), (), E E (), (3.14) i i (),

where is the radian frequency. The indicated substitutions follow directly from the denition of the (generalized) Fourier transform, f () as the purely imaginary form of the Laplace transform (cf. Tschoegl, 1989, p. 573). We have

16

N.W. TSCHOEGL ET AL.

f () = lim

0 0

f (t) exp[( + j )t ] dt = f(s)|s =j = F [(f (t)], (3.15)

where F [ ] denotes the Fourier transform operator, and f(s) is a relaxance or a retardance. Applying this to the Poissons ratio, we have () = [s (s) ]|s =j = j F [(t)], (3.16) since s (s) = (s) is the Poisson retardance. Equation (3.16) establishes the (complex) frequency-dependent Poissons ratio as the j -multiplied Fourier transform of the time-dependent ratio. From Equations (2.8) to (2.9), () becomes 3J () 2B () 3K () 2G () = (3.17) () = 6K () + 2G () 6J () + 2B () in terms of the frequency-dependent fundamental moduli and compliances, E () 1 B () 1 1 1 = = E ()B () (3.18) () = 2 6K () 2 6D () 2 6 and J () 1 E () 1 = 1 = E ()J () 1 (3.19) () = 2G () 2D () 2 in terms of the frequency-dependent stretch moduli and compliances in addition to those in shear. We further have M () 2G () (3.20) () = 2[M () G ()] in terms of the frequency-dependent longitudinal bulk and shear moduli. The complex quantities must then be separated into their real and imaginary parts (see Section 3.3.2). 3.1.2. The General Stress-Strain Relations In the transform plane the general relations between the three-dimensional stress ij (s), take the form tensor, ij (s), and strain tensor, 2 (s)ij + 2s G(s) s G(s) ij (s) (3.21) ij (s) = s K(s) 3 in terms of the transforms of the fundamental moduli (cf. Tschoegl, 1989, p. 514). Here, (s) is the transform of the time-dependent cubical dilatation and ij is the unit tensor (Kroneckers delta). We now use the expressions s E(s) s E(s) and s G(s) = (3.22) s K(s) = 3[1 2s (s) ] 2[1 + s (s) ] [cf. Equations (2.15)], to recast Equation (3.21). This yields the general stressstrain relation as

POISSONS RATIO IN LINEAR VISCOELASTICITY

17 (3.23)

s (s) 1 + s (s) (s)ij + ij (s) = ij (s) 1 2 s (s) s E(s)

in terms of the Carson transforms of Poissons ratio and of the stretch modulus. Clearly, at the limits of the range of admissible values for this relation becomes indeterminate so that the restriction 1 < < 0.5 must apply. 3.1.3. The Transform Relation for Poissons Ratio The relation between the longitudinal and the transverse deformations in uniaxial extension as a step function of time is given in the transform plane by (s) 1 (s) = s (s) 1 (s) 2 (s) = (3.24)

where 1 (s) is the transform of the longitudinal deformation in the 1-direction, while 2 (s) is the transform of the deformation in one of the transverse directions in the uniaxial deformation of an isotropic material. Since the latter deformation is a contraction, the term carries the negative sign. The time-dependent Poissons ratio, (t), is dened here as resulting from the response of the material to the imposition of a step function of time, h(t). Given that the deformation is a strain, 0 , applied in the 1-direction, it becomes 0 (t) = 1 h(t). Laplace transformation of 1 (t) yields 1 (s) = L[0 h(t)] = 0 /s. Introducing the latter into Equation (3.24) gives 2 (s) = (s) 0 (3.26) (3.25)

as the transform relation for Poissons ratio. Equation (3.26) refers to a relaxation experiment. We have written (s) for the more pertinent rate (s) to conform to the usual notation. When considering other time regimes, such as constant rate of strain or creep, Equation (3.24) must be modied to yield rate (t), and creep (t), respectively (see Sections 3.6 and 3.7). 3.1.4. The Delay Time Distributions in the Transform Plane We have pointed out in Section 1.2 that the viscoelastic Poissons ratio is characterized by distributions of delay times, that are analogous to but not necessarily identical with the well-known distributions of retardation times. Accordingly, we now consider the delay time distributions for (s) in both their discrete and continuous forms. In accordance with the retarded nature of the viscoelastic Poissons ratio, the behavior of all its functions formally resembles that of the corresponding compliances functions. The time-dependence of Poissons ratio is therefore expressed in the transform plane by a distribution just as is that of the transforms of

18

N.W. TSCHOEGL ET AL.

(s), and the bulk compliance, B(s) . The transform of the the shear compliance, J discrete delay time distributions of Poissons ratio thus becomes (s) = e s
i =N

i
i =1

i 1 + i s

(3.27)

or, equivalently, (s) = g + s


i =N

i
i =1

1 , s(1 + i s)

(3.28)

where e and g are the equilibrium and glassy (or instantaneous) Poissons ratios, respectively, and the i are the delay times, while the i are the associated lateral contraction ratios, i.e., the strengths (heights) of the spectral lines that compose the discrete distribution of delay times. Because the latter are neither retardation, nor relaxation times, we distinguish them here with the symbol . The two expressions for (s) are equivalent because
i =N

e g =
i =1

i .

(3.29)

The set {i , i ; i = 1, 2, . . . , N } represents the discrete distribution (or discrete spectrum) of delay times. Denoting the continuous distribution (or continuous spectrum) of the delay times of (s) by () , we have

(s) = g +
0

()

1 d, 1 + s

(3.30)

and it then follows from s (s) = (s) that g + (s) = s or, equivalently; e (s) = s

M ()

1 d ln s(1 + s)

(3.31)

M ()

d ln , 1 + s

(3.32)

where M () = (). The two expressions for (s) are again equivalent because now

e g =

M () d ln .

(3.33)

POISSONS RATIO IN LINEAR VISCOELASTICITY

19

The integral and summation formulations for e g do not imply that the sum and the integral are equal. The discrete and continuous representations cannot be converted into one another in any simple way (but see Baumgaertel and Winter, 1992). If the material contains more than one distribution (e.g., a secondary transition in addition to the main transition), Equations (3.27), (3.28), (3.31) and (3.32) must take this into account (see Section 3.8). 3.2. P OISSON S R ATIO IN THE T IME D OMAIN From the transform relations introduced in Section 3.1.3 we obtain the corresponding time-dependent relations through transform inversion. Let L1 [f(s)] denote the inverse Laplace transformation. Retransforming Equation (3.26) onto the real 2 (s) = (s) time axis of the transform plane according to L1 [ 0 ] gives 2 (t) = (t)0 from which (t) = 2 (t) 0 (3.35) (3.34)

follows as the denition of the time-dependent Poissons ratio. Equation (3.35) clearly indicates that the ratio must be determined in the longitudinal direction in response to a step function of time, i.e., in a stress relaxation, and not in a creep experiment. Measurements can be made in creep but must be converted (see Section 3.7). 3.2.1. The Delay Time Distributions in the Time Domain Equations (3.27) and (3.28) yield the expressions for the delay time distributions of (t) as
i =N

(t) = e
i =1

i exp(t/i )

(3.36)

or
i =N

(t) = g +
i =1

i [1 exp(t/i )]

(3.37)

in the discrete representation, while Equations (3.31) and (3.32) furnish them as

(t) = e

M () exp(t/) d ln

(3.38)

20 or

N.W. TSCHOEGL ET AL.

(t) = g +

M ()[1 exp(t/)] d ln

(3.39)

in the continuous representation. These expressions all show that the time-dependent Poissons ratio, (t), is a monotone non-decreasing function of t , i.e. d/dt 0. None of them contain the ow term, f , familiar from the corresponding equations for the creep compliance because the lateral contraction cannot increase indenitely, not even in an uncrosslinked material. We also note that e is always nite. 3.2.2. (t) as Function of Two Other Response Functions Inversion of Equations (3.10) to (3.13) leads to expressions containing either of four equivalent convolution integrals that result from t df1 (u) f1 (0)f2 (t) + f2 (t u) du, du 0+ t df1 (u) f2 (u) du, f1 (0)f2 (t) du + 0 (3.40) L1 [sf1 (s)f2 (s)] = t d f (u) 2 du, f (0)f1 (t) + f1 (t u) 2 du 0+ t df2 (t u) du, f2 (0)f1 (t) f1 (u) du
0+

(see Tschoegl, 1989, p. 563). To save space and avoid unnecessary duplication, we will mostly display only the rst (topmost) of the four Equations (3.40) in the inversions that follow. These convolution integrals cannot be made explicit in (t) and they must normally be solved numerically. This invariably entails numerical errors. Certain of the four convolution integrals may be subject to less errors than others (Emri et al., 2002). To obtain the expressions for (t) in terms of the bulk, shear, and stretch moduli and compliances, Equations (3.10) must rst be cleared of fractions. We thus obtain ] = 1.5K(s) s (s) [3K(s) + G(s) + G(s), ] = 1.5J(s) + B(s). s (s) [3J(s) + B(s) (3.41)

POISSONS RATIO IN LINEAR VISCOELASTICITY

21

Applying now the inverse Laplace transformation, L1 [f(s)], we nd


t

g [3K(t) + G(t)] +
0+

d(u) [3K(t u) + G(t u)] du du (3.42)

= 1.5K(t) + G(t), and


t

g [3J (t) + B(t)] +


0+

d(u) [J (t u) + B(t u)] du du (3.43)

= 1.5J (t) + B(t). Equations (3.11) yield (t) = L1 1 1 s E(s) B(s) 2s 6


t

1 1 1 = Eg B(t) + 2 6 6

0+

dE(u) B(t u) du du

(3.44)

and from Equations (3.12) we obtain (t) = L1 = 1 1 s E(s)J(s) 2 s


t

(3.45)

1 Eg J (t) 2 + 2

0+

dE(u) J (t u) du . du

Finally, we rewrite Equation (3.13) as 1 ](s) G(s) [s M(s) s G(s) = M(s) 2 and obtain
t

(3.46)

(Mg Gg )(t)
0+

dM(t u) dG(t u) (u) du du du (3.47)

1 M(t) G(t), 2

where we have used the second of the convolution integrals, Equations (3.40).

22 3.3. P OISSON S R ATIO IN THE F REQUENCY D OMAIN

N.W. TSCHOEGL ET AL.

In the frequency domain we are concerned with the response of the material to a sinusoidal steady-state excitation of unit amplitude. If, therefore, instead of a step function of strain with height 0 , we apply a sinusoidal steady-state strain with amplitude 0 , we replace Equation (3.26), in accordance with the correspondence principle, by
2 () = ()0 ,

(3.48)

() is the complex sinusoidal steady-state lateral contraction, and where 2

() =

2 () 0

(3.49)

denes the complex frequency-dependent Poissons ratio. Because of the compliance nature of the viscoelastic Poissons ratio, in Cartesian coordinates the real and imaginary parts of () are () = () j (). (3.50)

In the literature one sometimes nds formulae for the real and imaginary parts following from considering () to be given by () = () + j () (see, e.g., Rigby, 1965, 1967; Waterman, 1977; Yee and Takemori, 1979). This, however, implies that Poissons ratio possesses the nature of a modulus instead of that of a compliance, i.e., that it is a relaxing rather than a retarded quantity, and this view is clearly erroneous. In polar cordinates we have exp[j ()] = () [cos () j sin ()], () = () where () = [ ()]2 + [ ()]2 (3.52) (3.51)

is the absolute (complex) Poissons ratio, and () = tan1 () () (3.53)

is the phase (or loss) angle. The real and imaginary parts are then obtained as cos () () = () and sin (), () = () respectively. (3.55) (3.54)

POISSONS RATIO IN LINEAR VISCOELASTICITY

23

3.3.1. The Delay Time Distributions in the Frequency Domain The delay time distributions governing the behavior of the frequency-dependent Poissons ratio may be represented either in discrete or in continuous form. Thus, Equations (3.27) and (3.28) yield () = e
i =1 i =N

j i 1 + j i

(3.56)

or () = g +
i =N

i
i =1

1 1 + j i

(3.57)

for the discrete representation of (). Its real (or storage) parts become
i =N

() = e
i =1

2 i2 1 + 2 i2

(3.58)

or
i =N

() = g +
i =1

1 , 1 + 2 i2

(3.59)

and the imaginary (or loss) part results as


i =N

() =
i =1

i . 1 + 2 i2

(3.60)

For the continuous representation Equations (3.31) and (3.32) yield

() = e

M ()

j d ln 1 + j

(3.61)

or

() = g +

M ()

1 d ln 1 + j

(3.62)

with the real or storage parts as

() = e

M ()

2 2 d ln 1 + 2 2

(3.63)

24 or

N.W. TSCHOEGL ET AL.

() = g +

M ()

1 d ln 1 + 2 2

(3.64)

and the imaginary or loss part as

() =

M ()

d ln . 1 + 2 2

(3.65)

The real parts are seen to be monotone non-increasing functions of , i.e., d/d 0. The imaginary part, of course, displays a maximum in the transition region. as Functions of Two Other Response Functions 3.3.2. () and () We list the real and imaginary parts of (), that is, () and (), as functions of two other response functions, expressing the latter both in Cartesian and in polar form. One or the other of these relations is sometimes quoted incorrectly in the literature. In Cartesian coordinates () becomes
() = (3.66)

[(3K () 2G ()][(3K () + G ()] + [(3K () 2G ()][(3K () + G ()] = 2{[3K () + G ()]2 + [3K () + G ()]2 } [(3J () 2B ()][(3J () + B ()] + [(3J () 2B ()][(3J () + B ()] , 2{[3J () + B ()]2 + [3J () + B ()]2 }

in terms of the bulk and shear moduli [cf. Equations (3.17)]. From Equations (3.18) and Equations (3.19), () follows as () = = and as () = = E ()G () + E ()G () 1 2{[G ()]2 + [G ()]2 } D ()J () + D ()J () 1. 2{[D ()]2 + [D ()]2 } (3.68) 1 E ()K () + E ()K () 2 6{[K ()]2 + [K ()]2 } 1 D ()B () + D ()B () , 2 6{[D ()]2 + [D ()]2 } (3.67)

POISSONS RATIO IN LINEAR VISCOELASTICITY

25

Finally, Equation (3.20) gives () = (3.69) [M () 2G ()][M () G ()] + [M () 2G ()][M () G ()] . 2{[M () G ()]2 + [M () G ()]2 } Turning now to the imaginary part, (), this follows as () = = and () = = or () = = Further, () = M ()G () M ()G () . 2{[M () G ()]2 + [M () G ()]2 } 2 () 3K() 2 () 9K G() cos[K () G ()] 2G 2 () + 6K() 2 ()} 2{9K G() cos[K () G ()] + G 2 () 9J2 () 3J()B() cos[J () B ()] 2B 2 ()} 2{9J2 () + 6J()B() cos[J () B ()] + B E() 1 cos[E () K ()] 2 6K() B() 1 cos[B () D ()] 2 6D() (3.76) (3.73) E ()G () G ()K () 2{[G ()]2 + [G ()]2 } D ()J () D ()J () . 6{[D ()]2 + [D ()]2 } (3.72) E ()K () E ()K () 6{[K ()]2 + [K ()]2 } D ()B () D ()B () , 6{[D ()]2 + [D ()]2 } (3.71) 9[K ()G () K ()G ()] 2{[3K () + G ()]2 + [3K () + G ()]2 } 9[J ()B () J ()B ()] , 2{[3J () + B ()]2 + [3J () + B ()]2 } (3.70)

In polar coordinates the real part of Poissons ratio, m (w), becomes () = and () = (3.75) (3.74)

in terms of the bulk and shear responses, () = =

26 or 1 1 E()B() cos[E () B ()] 2 6 in terms of the bulk and stretch responses, and () = () = = or 1 () cos[E () J ()] 1 E()J 2 in terms of the stretch and shear responses. In terms of the longitudinal bulk and shear moduli we have () = () = E() cos[E () G ()] 1 2G() J() cos[J () D ()] 1 2D()

N.W. TSCHOEGL ET AL.

(3.77)

(3.78)

(3.79)

2 () 2M() 2 () M G() cos[M () G ()] + 2G . 2 () 2M() 2 () 2{M G() cos[M () G ()] + G 4.5K() G() sin[G () K ()] 2 2 () 9K () + 6K()G() cos[K () G ()] + G 4.5J()B() sin[B () J ()] 2 () 9J2 () + 6J()B() cos[J () B ()] + B

(3.80)

The imaginary part, (), takes the forms () = and () = (3.82) (3.81)

from the bulk and shear responses, () = or 1 E()B() sin[B () E ()] 6 from the bulk and stretch responses, and () = () = or () = 1 () sin[J () E ()] E()J 2 (3.86) (3.84) E() B() sin[K () E ()] = sin[D () B ()] (3.83) 6K() 6D()

() E() J sin[G () E ()] = sin[D () J ()] (3.85) 2G() 2D()

POISSONS RATIO IN LINEAR VISCOELASTICITY

27

Figure 2. Comparison of the time- and the frequency-dependent Poissons ratios for the Standard Linear Solid.

from the stretch and shear responses. Finally, we have () = M() G() sin[M () G ()] 2 () 2M() 2 () 2{M G() cos[M () G ()] + G (3.87)

from the longitudinal bulk and shear moduli. All of these equations contain differences between measured quantities. To obtain meaningful results from them, the quantities in question must clearly be determined by adhering strictly to the Standard Protocol. 3.4. M ODEL B EHAVIOR Figure 2 illustrates schematically the behavior of (t), () and () as functions of log t and log , respectively. The functions were calculated from the relations of the Standard Linear Solid (see Tschoegl, 1989, p. 531). The Laplace transform of the viscoelastic Poissons ratio for this model becomes (s) = (e g ) e s 1 + s (3.88)

using e = 0.5, g = 0.333, and = 1 for the constants. The plot thus shows (t) = e (e g ) exp(t/), (3.89)

28 () = e (e g ) and () = (e g ) , 1 + 2 2 2 2 , 1 + 2 2

N.W. TSCHOEGL ET AL.

(3.90)

(3.91)

The curve is shifted upward by 0.333 to yield a more convenient display. Signicantly, neither (t) nor () support the formation of any maxima or minima. They are simply monotone non-decreasing and monotone non-increasing functions of their arguments. 3.5. S UPERPOSITION OF I SOTHERMAL AND / OR I SOBARIC S EGMENTS OF P OISSON S R ATIO Procedures for shifting isothermal or isobaric segments of Poissons ratio into superposition along the logarithmic time or frequency axis have not been established unambiguously. There is no information on shifting isobaric segments of the ratio. In amorphous polymers isothermal segments do seem to shift according to the WLF equation (Ferry, 1980, p. 274; Lu et al., 1977). However, the shift parameters have not been compared reliably with those that apply to other viscoelastic response functions in particular the compliance functions preferably determined according to the Standard Protocol. The question of whether K(t) and B(t) shift with the same parameters as G(t) and J (t) is still not resolved satisfactorily either, although recent experiments lend support to this contention (cf. Deng and Knauss, 1997; Sane and Knauss, 2001). 3.6. T HE L ATERAL C ONTRACTION R ATIO IN C ONSTANT R ATE OF S TRAIN M EASUREMENTS One of the most widely used rheological experiments is the tensile experiment performed at a constant rate of strain, 0 . The extension is, of course, accompanied by a lateral contraction that may be measured. However, these measurements do not directly yield Poissons ratio as dened by Equation (3.35). We now have 0 t, 1 (t) = and its transform becomes 0 /s 2 . L[1 (t)] = Substituting this into the basic relation, Equation (3.24), yields rate (s) 0 /s. 2 (s) = (3.94) (3.93) (3.92)

POISSONS RATIO IN LINEAR VISCOELASTICITY

29

Elimination of 2 (s) between Equations (3.26) and (3.94) leads to (s) = 0 rate (s) 0 s (3.95)

0 t = 0 , thus 0 /0 = 1/t , as the transform relation between (s) and rate (s). But and division by s in the transform plane implies integration in the time domain. Accordingly, we obtain
t

(t) = (1/t)
0

rate (u) du,

(3.96)

and see that (t) can be recovered from measurements at constant rates of extension by applying Equation (3.96). Experimentally, the chief drawback of measurements of the lateral contraction ratio in this way is that one is likely to soon exceed the linear range. This imposes the need for a correction for the strain dependence of the measured ratio, perhaps along the lines of Equation (4.1). 3.7. T HE L ATERAL C ONTRACTION R ATIO IN C REEP The lateral contraction can evidently be measured also in uniaxial creep. The application of a constant stress, 0 , produces creep in both the longitudinal and the transverse direction. In such a deformation, therefore, two retardation processes take place simultaneously. It would obviously be quite erroneous to consider the ratio creep (t) = 2 (t) 1 (t) (3.97)

as yielding the time-dependent Poissons ratio dened by Equation (3.35). The problem can again be worked through the representation in the transform plane. Partial retransformation of Equation (3.24) unto the real time axis gives the timedependent lateral contraction as 1 (s)]. 2 (t) = L1 [s (s) (3.98)

Thus, simultaneously measuring both the uniaxial extension, 1 (t), and the lateral contraction, 2 (t) as functions of time in a creep experiment, one can, in principle, obtain (t) by solving either of the integral equations
t t

(u) 1 (t u) du = du
0

(t u) 1 (u) du = 2 (t) + g 1 (t) du

(3.99)

30 or
t t

N.W. TSCHOEGL ET AL.

1 (t u) du = (u) du
0

(t u)

1 (u) du = 2 (t) + 0 Dg (t) (3.100) du

(see Tschoegl, 1989, p. 563), having previously determined the glassy ratio, g , or, alternatively, the tensile glass compliance, Dg , in a separate experiment. Suppose now that creep (t) is given but neither 2 (t) nor 1 (t) are known. How can one then convert creep (t) to (t)? By Equation (3.24) we have 2 (s) = 1 (s) for the transform relation in creep. On the left of this relation we s creep (s) , and on the right of it we introduce 1 (s) = now substitute 2 (s) = 0 (s) where D(s) is the tensile creep compliance, and 0 is the constant stress 0 D(s) that produces it. This yields (t) = (0 /0 )creep (t)D(t) = (1/Dg )creep (t)D(t), (3.101) as the relation between (t) and creep (t). Van der Varst and Kortsmit (1992) have derived an equivalent form of Equation (3.24) as a Stieltjes convolution involving the viscoelastic analogues of the Lam constants, G, and = K 2G/3. Their gure 2 exhibits plots of arbitrary 3-parameter models for creep (t) and (t) (see their curves for c (t) and s (t), respectively). The authors did not, however, show how to solve their relation in the general case. 3.8. M ATERIALS WITH M ULTIMODAL D ISTRIBUTIONS The relations introduced so far apply to unimodal materials, that is, materials that display a single distribution of respondance (relaxation, retardation, and thus also delay time) distributions. Amorphous homopolymers and random copolymers can often be treated as unimodal materials. Even though they might contain secondary transitions in the glassy region, these are often too small to be picked up. Many other materials such as typically block and graft copolymers and blends, however, may display two or more so-called main transitions. Such materials may be called multimodal materials. In such materials two or more distinct molecular mechanisms contribute to the material response, generally in different portions of the time scale. The relations in Sections 3.2 and 3.3 must then be modied to reect this behavior. Thus, for a bimodal distribution (Tschoegl, 1989, pp. 489ff.) Equations (3.27) and (3.28), for example, become (s) = and g + (s) = s
i =M i =1

e s

i =M

i
i =1

i i i 1 + i s i =M +1 1 + i s
j =N

i =N

(3.102)

1 + i s(1 + i s)

i
j =1

1 . s(1 + + j s)

(3.103)

POISSONS RATIO IN LINEAR VISCOELASTICITY

31

The corresponding equations in the time or frequency domains are easily derived from those in the transform plane using Section 3 as a guide. must all three still be monotone Importantly, the ratios (t), (), and () non-decreasing functions of their arguments, displaying steps (plateaus) for each distribution, and no maxima nor minima. By contrast, (), (), and M () will display distinct peaks if the distributions are well separated, and shoulders and peaks if they are not. Tschoegl (1898, pp. 489ff.) has modeled some aspects of the behavior of bimodal distributions. For an example of a multimodal Poissons ratio see, e.g., Waterman (1963b), reviewed in Section 4.3.1. 3.9. L ARGE D EFORMATIONS Poissons ratio is strictly dened in innitesimally small deformation. In fact, the restriction to innitesimally small deformation is inherent in the denition of all linear viscoelastic response functions. An innitesimally small deformation is a mathematical abstraction. In practice one must invoke the maxim that near enough is good enough. This allows one to get away with a deformation that is not innitesimally small but small enough to yield satisfactory results. Experimental measurements of Poissons ratio reveal it to depend signicantly on strain (cf. Equation (4.1)). The relative change in volume resulting from a uniaxial extension is V 2 = 1 2 (3.104) 2 1 = (1 + 1 )(1 + 2 ) 1 (3 = 2 ) V0 where the i are the stretch ratios and the i = i 1 are the strains in the three principal directions. Introducing Poissons ratio, Equation (2.7), we nd V 2 3 = (1 2)1 (2 )1 + 1 . V0 (3.105)

Thus, V = 0 yields the traditional incompressibility relation for Poissons ratio, = 1/2, only upon neglect of all higher strain terms. This implies that the stress-strain relation must be linear, and lets us appreciate the requirement that the uniaxial extension be small. No critical examination of the behavior of the lateral contraction ratio in moderately large (Chang et al., 1997) or in large deformations has been undertaken although corrections have been applied by some authors when the lateral contraction was measured under conditions of strain that exceeded the linear strain limit (see, for instance, Kugler et al., 1990). It would seem that more work on this aspect is needed because relations such as Equations (3.21) and (3.23) must not be used where a good enough approximation to an innitesimally small deformation cannot be assumed to yield satisfactory results. Currently, there does not seem to be any information on the error that may occur when the traditional denition of Poissons ratio is employed in deformations that are not small enough to justify its use.

32 4. Determination of the Viscoelastic Contraction Ratio

N.W. TSCHOEGL ET AL.

This section reviews published work on the results obtained in experimental determinations of viscoelastic lateral contraction ratios. Work on the elastic Poissons ratio will not be considered but some attention will be paid to publications that report essentially only the glassy value, g , or the equilibrium value, e , of the viscoelastic Poissons ratio (cf. Section 4.4). Our main concern is the determination of the ratio as a function of time or frequency in uniaxial extension, although we include work on other viscoelastic lateral contraction ratios wherever such work appears relevant for one reason or another. Determinations of the time- or frequency-dependent Poissons ratio may be direct or indirect. In the former, Poissons ratio is obtained from actual measurements of the lateral contraction. The latter involve calculation from two other timeor frequency-dependent material functions, short circuiting any measurements of the lateral contraction as such. Indirect determinations generally necessitate the solving of integral equations. This may be accomplished, for instance, by means of the method of Hopkins and Hamming (1957), or the Windowing Algorithm (see the Appendix). With respect to numerical errors arising in the solutions of the convolution integrals, see Emri et al. (2002). Work on the determination of the viscoelastic Poissons ratio reported before 1963 has been ably reviewed by Kstner and Pohl (1963) and will not be repeated here. No direct measurements of the ratio had apparently been made prior to that date. We review later attempts in this direction in Section 4.1.1. Although some indirect measurements made isochronally as a function of temperature (cf. Section 4.3.2) were successful, most of those made as function either of time or of frequency did not do well in determining the viscoelastic Poissons ratio. The reported results were generally suspect, showing maxima or minima where none should have appeared, and were often in contradiction with each other and with the linear theory of viscoelasticity. We believe that these determinations were inadequate, at least in part, because they did not conform to the Standard Protocol described in Section 1.3. Measurements of Poissons ratio that are carried out as a function of temperature at xed (isochronal) values of the time or frequency reveal the viscoelastic nature of the ratio and may shed some qualitative light on its time-dependent behavior. Indeed, if several isochrones have been determined, the time or frequency dependence (i.e., the isotherm) can be obtained simply by crossplotting (Onogi and Ui, 1956; Ferry, 1980, p. 313). If only a single isochrone is available but the timetemperature shift factor, aT , has been determined through measurements of any of the standard response functions, the isochrone can be transformed into an effective isotherm by plotting against t/aT or aT (see the references above) provided that those shift parameters have been shown to be applicable.

POISSONS RATIO IN LINEAR VISCOELASTICITY

33

We divide this section into four main subsections according to whether the determinations of the lateral contraction ratio were carried out as a function of time (4.1), as a function of frequency (4.2), as a function of temperature (4.3), or as a function of strain (4.4). 4.1. T IME -D EPENDENT M EASUREMENTS Either direct or indirect determinations resulting from the imposition of a tensile strain as a step function of time are the only ones that can yield a true time-dependent Poissons ratio. Any other deformation may produce a lateral contraction ratio (see Sections 3.6 and 3.7), but this must not be taken without further ado for the genuine article. 4.1.1. Direct Determination of the Time-Dependent Poissons Ratio In these determinations a constant uniaxial deformation (a strain as a step function of time) is imposed on the specimen, and the lateral contraction is measured as a function of time also. The ratio is then calculated using the appropriate relation reviewed in Section 3.2. Kugler et al. (1990) measured both the uniaxial extension and the lateral contraction in stress relaxation tests on a glass bead lled, compounded Hypalon rubber using electro-optical extensometers. They made their measurements at ve constant strain levels over two decades of time from 10 to 1000 seconds. Their lowest strain level was 0.01. All curves showed the expected increase in the lateral contraction ratio with time but did not reach either the glassy or the equilibrium region. They accounted for the strain-dependence of the ratio through the equation of Smith (1959) L (t) = d log(V (t)/V0 ) 1 1 , 2 d log (t) (4.1)

where V is the volume and = 1 + is the stretch ratio. The function L (t) is independent of strain for constant volume and is still 0.5 for an incompressible material. The authors reported a mastercurve of L (t) vs. log t/aT for one of their elastomers resulting from the superposition of the measured isotherms. Knauss (1992) employed a speckle interferometric technique to obtain the timedependent Poissons ratio on a sample of commercial poly(methyl methacrylate) (PMMA) at 12 temperatures ranging from about 25 to 125 C, i.e., mainly in the glassy region. The data showed considerable scatter. E(t) was measured concurrently with (t). An attempt was made to obtain K(t) from E(t) and a smooth curve of (t) that lay between the error bounds of the measurements. The E(t) data required both horizontal and vertical shifts for a reasonable superposition. The data clearly showed the expected lack of superposition due to the presence of the -relaxation mechanism in PMMA arising from motion of the side chains.

34

N.W. TSCHOEGL ET AL.

Figure 3. The time-dependent Poissons ratio, (t), of PMMA.

Delin et al. (1995) determined Poissons ratio directly on low density (LDPE) and on linear low density (LLDPE) polyethylenes at 23.6 C. The specimens were stressed at a constant rate of strain of 3.03%/min until the desired strains of about 0.2 to 0.33% were reached. The changes in the lateral dimensions were measured using a non-disturbing inductive extensometer and were conrmed using a liquid dilatometer. The data followed smooth courses and showed neither maxima nor minima. Each sample was stressed at two levels of the nal strain, never exceeding, however, the limit of linear viscoelastic behavior. Higher values of Poissons ratio were measured at the higher strains, clearly indicating the dependence of the ratio on strain. Lu et al. (1997) measured the time-dependent Poissons ratio of poly(methyl methacrylate) directly, using an image moir method. Their measurements, reproduced in Figure 3, again showed considerable scatter, which they attributed to the lack of accuracy afforded by their experimental method. They concluded that the ratio must be determined with at least four gure accuracy to deduce bulk response. The central curve between the error bounds in Figure 3 represents a best t through the data. It shows the theoretically expected monotone non-decreasing behavior and is the only more or less acceptable determination of (t) to date. Regrettably, none of the cited efforts produced an unimpeachable result. Even though the data reported were aficted by various shortcomings, they are significant nevertheless in that none of them showed any of the maxima and minima, inconsistent with linear viscoelastic theory, that are generally reported when Pois-

POISSONS RATIO IN LINEAR VISCOELASTICITY

35

sons ratio is determined indirectly from independent measurements of two other viscoelastic response functions. 4.1.2. Indirect Determination of the Time-Dependent Poissons Ratio No successful indirect determinations of the time-dependent Poissons ratio, (t), seem to have been reported so far. Theocaris (1964) described lateral contraction ratios of a cold setting epoxy polymer on which he had measured the tensile relaxation modulus, E(t), and the shear creep compliance, J (t), in two separate experiments. These determinations did not conform to the Standard Protocol. He calculated (t) using the relation t dE(t u) 1 Eg J (t) 2 J (u) du (4.2) (t) = 2 du
0

[cf. Equation (3.45)], and evaluated the convolution integral by an approximation procedure utilizing the fact that the bulk responses are weak functions of time. No maxima or minima appeared in the plot of (t) vs. log t which showed qualitatively the expected behavior. The author also calculated (t) from t dJ (u) 1 Jg E(t) 2 + E(t u) du , (4.3) (t) = 2 du
0

again evaluating the convolution integral by the same approximation procedure that he used to evaluate Equation (4.2). He believed that this equation furnishes what he called the creep lateral contraction ratio. In fact, his two equations are just two equivalent forms of each other as is easily seen by transforming both into the transform plane [cf. Equations (3.45) and (3.40)]. By rights, therefore, the two equations should have yielded the same results and the two curves representing the calculated values should have been identical. They were not. Since the same data were used in both calculations, the discrepancy would appear to lie in differences in the effects of the approximations used in the evaluations of the two convolution integrals. Unfortunately this affects the credibility of both curves. Tsou et al. (1995) reported values of Poissons ratio for polycarbonate and for cellulose acetate at 21 C over a time span of four logarithmic decades. The values were calculated from measurements of the tensile relaxation modulus and the timedependent bending modulus. The curve for polycarbonate was essentially at while that for cellulose acetate was roughly parabolic with a denite maximum around 1000 seconds. The result on polycarbonate may be acceptable as g , but that on cellulose acetate is puzzling and probably occurred because the measurements were not executed according to the Standard Protocol.

36

N.W. TSCHOEGL ET AL.

Figure 4. The real part of the complex Poissons ratio, (f ), of PMMA.

4.2. F REQUENCY-D EPENDENT M EASUREMENTS Frequency-dependent measurements are most often performed in response to a strain excitation since this is usually easier to impose than a stress excitation. Sometimes only the real (or storage) part of the response is reported because the imaginary (or loss) part is generally more difcult to determine. It is often quite small, particularly in the glassy and in the rubbery region. Many of the frequencydependent measurements were reported on polymers in the glassy region that showed little, if any, time dependence. Some frequency-dependent measurements were reported not in terms of the radian, but of the cycles/second frequency, f = /2 . We report the latter uniformly in Hz. 4.2.1. Direct Measurements of the Frequency-Dependent Poissons Ratio Kstner and Pohl (1963) directly determined the real part of the complex Poissons ratio, (f ) on a sample of poly(methyl methacrylate) (PMMA) at six temperatures between 90.5 and 106 C, and at frequencies between 5 104 and 101 Hz. They measured the changes in dimensions of the rectangular cross-section of a cantilevered beam. The imaginary part was found to be too small to measure. Their mastercurve of (f ) as function of frequency is displayed in Figure 4. The plot shows the expected behavior. They also plotted (f ) as a function of temperature (see Section 4.3) and this plot also displayed qualitatively the expected behavior but is not shown here. The authors further determined the real and imaginary parts of G (f ) and E (f ) on the same material but did not undertake

POISSONS RATIO IN LINEAR VISCOELASTICITY

37

to obtain Poissons ratio from these. Neither did they try to nd K (f ) from (f ) and E (f ) by calculation. Giovagnoni (1994) attempted to measure the absolute Poissons ratio, (f ), and the phase angle, tan (f ), on four materials: poly(vinyl acetate), poly(methyl methacrylate), a polyamide, and a polyacetal resin. The measurements were made in bending in the frequency range from 80 to 720 Hz. The axial and the lateral deformations were measured with strain gauges, at the same location, on the surface of the tensile part of the beam. From these data Giovagnoni then calculated (f ) and tan (f ). The measurements were performed at room temperature (25.5 26.5 C). The absolute Poissons ratios and the phase angles varied at most only slightly as a function of the frequency since the measurements were made in the glassy region. Yee and Takemori (1982) reported measurement of the absolute Poissons ratio, () , at 0, 20, and 40 C on poly(methyl methacrylate) (PMMA). Their experiments are more fully described in Section 4.3.1. 4.2.2. Indirect Measurements of the Frequency-Dependent Poissons Ratio In the frequency domain indirect determinations generally necessitate the taking of differences [cf. Equations (3.66) to (3.71)]. If the imaginary parts are found to be negligibly small, one may obtain the real parts with reasonably good approximation from E () 1 E () 2G () 1 = 2 6K () , (4.4) lim () = ()0 1 E() E() 1= . 2 6K() 2G() This shortcut has been used by several authors. Philippoff and Brodnyan (1955) attempted to evaluate the real part of Poissons ratio, (f ), from separately measured values of E (f ) and G (f ) of poly(vinyl chloride) and polyethylene specimens. The measurements were carried out in the frequency range from 0.3 MHz to 5 Hz, and at temperatures ranging from 25 to 95 C. They also measured the compressibility, and from this obtained the bulk modulus as K (f ) = 1/ (f ). From K (f ) and E (f ) they then calculated (f ) again. They used Equations (4.4) for both calculations, and found not surprisingly that their two real parts of Poissons ratio did not agree. ) as well as Koppelmann (1959) measured the absolute Youngs modulus, E(f the absolute shear modulus, G(f ), of poly(methyl methacrylate), at various temperatures between 20 to 100 C in the frequency range from 105 to 101 Hz. He calculated the absolute Poissons ratio, (f ), from the third of Equations (4.4) and found the average value of (f ) to be 0.31 independently of either frequency or temperature. The measurements were clearly made in the glassy region so that the quoted gure represents the glassy value of Poissons ratio, g .

38

N.W. TSCHOEGL ET AL.

Thomson (1966) calculated values of the real and imaginary parts of Poissons ratio, (f ) and (f ), for a polyurethane rubber compound from literature values of the complex moduli, G (f ) and E (f ) at frequencies between 100 and 700 Hz, estimating values near 0 Hz. He showed a strong maximum in (f ) and a minimum in (f ). His values are meaningless, however, because he used incorrect formulae to obtain them. In their study of physical aging in poly(vinyl acetate) Kubt et al. (1998) obtained data on (f ) as a function of aging time at three frequencies. The data were apparently calculated from experimental values of K (f ) and E (f ) using the equation (f ) = E (f ) 1 2 6K (f ) (4.5)

[cf. the second of Equations (4.4)]. This may apply when (f ) is small enough to be neglected. The authors are, however, silent on this point. The data lie between 0.2 and 0.3. They are the smaller the lower the frequency and they decay with aging time. 4.3. T EMPERATURE -D EPENDENT M EASUREMENTS Temperature-dependent measurements of Poissons ratio, (T ), must be made isochronally, i.e., at a given xed time or frequency. Since (T ) thus does not vary with either, if we wish to express it in terms of two other material functions taken at a xed time, we simply use the relations furnished by the theory of elasticity but show the temperature dependence of the functions explicitly. We thus have [cf. Equations (2.8), (2.9), and (2.10)], (T ) = E(T ) 1 E(T ) 3K(T ) 2G(T ) = 1= 6K(T ) + 2G(T ) 2G(T ) 2 6K(T ) J (T ) 1 B(T ) 3J (T ) 2B(T ) = 1= 6J (T ) + 2B(T ) 2D(T ) 2 6D(T ) (4.6)

in terms of the bulk, shear, and stretch moduli, and (T ) = (4.7)

in terms of the corresponding compliances when the functions are determined at a xed time. When the functions are measured at a xed frequency, we substitute the complex functions for the elastic constants, again marking them as functions of the temperature. Thus, (T ) = E (T ) 1 E (T ) 3K (T ) 2 G(T ) = 1 = , 6K (T ) + 2 G(T ) 2G (T ) 2 6K (T ) (4.8)

where the real and imaginary parts must be separated. The resulting expressions for (T ) and (T ) can be derived from Equations (3.66) to (3.68), and (3.70) to (3.72) by formally replacing by T .

POISSONS RATIO IN LINEAR VISCOELASTICITY

39

In the limit as the imaginary parts of the moduli vanish, (T ) takes on the same form as the complex expressions changed to the real ones. Thus, e.g.,
K ,G 0

lim

(T ) =

3K (T ) 2G (T ) , 6K (T ) + 2G (T )

(4.9)

and analogously in all similar cases. The temperature-dependent Poissons ratio can, of course, also be determined either directly or indirectly. We emphasize again that the measurements must be based on stress relaxation, and not on strain retardation in the longitudinal direction (recall the comment at the end of Section 3.2). They should therefore be obtained from isochronal values generated using Equation (3.35), not Equation (3.97). 4.3.1. Direct Measurements of the Temperature-Dependent Poissons Ratio Kstner and Pohl (1963) obtained (T ) on poly(methyl methacrylate) at 0.05 Hz in the temperature range from 25 to 120 C by the same method that they had used to determine (f ) (cf. Section 4.2.1). The ratio increased monotonically, as expected. Heydemann (1963) reported direct isochronal measurements of the real part of Poissons ratio, (T ), at 4 Hz as a function of temperature over the range where he had found (see Section 4.3.2) a strong minimum in a sample of plasticized poly(vinyl chloride) when he calculated the ratio from indirect measurements. No minimum appeared in the direct measurements. Waterman (1963a) obtained values of (T ) and (T ) in isotactic polypropylene and high density polyethylene by measuring the velocity and attenuation of ultrasonic longitudinal and transverse pulses. The latter showed two transitions that exhibited terraces in (T ) and peaks in (T ) as required by linear viscoelastic theory. Using the same method, Waterman (1963b) then investigated several polyethylenes. Almost all of his samples showed more than one transition and these showed up clearly in the Poissons ratio measurements. Figure 5 shows the real and imaginary ratios obtained on one of his polyethylenes. Yee and Takemori (1982) directly determined Poissons ratio and Youngs modulus on poly(methyl methacrylate) in the glassy region by measuring the axial deformation and lateral contraction, reporting the absolute Poissons ratio, (t) , and the phase angle, tan (f ). The authors were primarily interested in the secondary, i.e., the -transition. The experiments were carried out on a servo-hydraulic Instron machine at different temperatures ranging from 160 to 200 C. A longitudinal and a transverse extensometer were attached to the specimen. Most of the data were reported as isochronal measurements as functions of temperature. Some data were also reported as isothermal measurement at 0, 20, and 40 C (see Section 4.21). Migwi et al. (1994) determined (T ) in the temperature range from about 0.37 to about 175 C on a silicone elastomer in a Mettler Thermal Mechanical Analyzer

40

N.W. TSCHOEGL ET AL.

Figure 5. (T ) and (T ) of PE showing a secondary transition.

(TMA) by measuring the linear thermal expansion coefcients in an unconstrained and a constrained conguration. The disk-shaped specimen was constrained by conning it within a copper annulus. The Poissons ratio was obtained from c (T ) u (T ) , (4.10) (T ) = c (T ) + u (T ) 2(T ) where c (T ) and u (T ) are the thermal expansion coefcients of the material in the constrained and the unconstrained conguration, and (T ) is that of the conning material. Their (T ) values varied between almost zero and about 0.35. The authors assumed purely elastic behavior throughout. Their approach requires reexamination. Simultaneously with the measurements of the expansion coefcients the authors also obtained the temperature-dependent stretch modulus E(T ) and then calculated the temperature-dependent shear and bulk moduli from E(T ) E(T ) and K(T ) = (4.11) G(T ) = 2[1 (T )] 3[1 2(T ) [cf. Equations (2.15)]. These, of course, are meaningful only under the assumption of purely elastic behavior.

POISSONS RATIO IN LINEAR VISCOELASTICITY

41

Figure 6. (T ) of PS and PMMA.

4.3.2. Indirect Measurements of the Temperature-Dependent Poissons Ratio Kono (1960) obtained the real part of the temperature-dependent Poissons ratio, (T ), of polystyrene (PS) and poly(methyl methacrylate) (PMMA) at 1 MHz over the temperature range from 20 to 190 C, using Equation (4.9). He measured the longitudinal bulk and shear moduli, M (f ) and G (f ) from the velocity and attenuation of longitudinal and transverse waves, and calculated the bulk modulus from the equation 4 K (T ) = M (T ) G (T ) 3 (4.12)

[cf. Equation (2.1)]. He then obtained (T ) from K (T ) and G (T ) using Equation (4.9). The plots for his two materials are reproduced in Figure 6. The author gave no reasons for using Equation (4.9) for Poissons ratio instead of
(T ) = (4.13) [(3K (T ) 2G (T )][(3K (T ) + G (T )] + [(3K (T ) 2G (T )][(3K (T ) + G (T )] 2{[3K (T ) + G (T )]2 + [3K (T ) + G (T )]2 }

[cf. the rst of Equations (3.66)], even though the imaginary parts of the moduli had also been determined. Nevertheless, the values of (T ) for both polymers appear reasonable, probably because the values of K (T ) and G (T ) were indeed small enough to be neglected. The curve for PMMA shows a small maximum at about 120 C that could possibly be due to the well established -mechanism in

42

N.W. TSCHOEGL ET AL.

that polymer. Figure 6 displays an example of an indirect determination that owed its success to the fact the source functions were determined following the Standard Protocol. Heydemann (1963) attempted to calculate the real part of the temperaturedependent Poissons ratio, (T ), at 1000 Hz from his measurements of the complex shear and tensile moduli on a sample of poly(vinyl chloride), plasticized with 30% of dioctyl phthalate over the temperature range from 50 to 80 C. He found a strong minimum at about 45 C. This minimum was absent when the ratio was calculated from measurements of the complex tensile and bulk moduli, also at 1000 Hz. It was also absent in direct measurements at 4 Hz. He ascribed the appearance of the minimum to the rate at which the temperature was changed. It vanished at lower rates of the temperature change. Crowson and Arridge (1979a) measured the torsional shear compliance, J (t), of an epoxy resin, diglycidyl ether of Bisphenol A, using a long thin-walled circular cylinder as specimen. They also simultaneously obtained the bulk compliance, B(t), through ination of the cylinder. The experiments were performed in the temperature range from 20 to 200 C, and covered a time span of 3.25 logarithmic decades. Plots of J (t) and of B(t) were shown but the authors interest was primarily in J (T ) and B(T ). They obtained these (Crowson and Arridge, 1979b) by crossplotting the values of the time-dependent data attained 20 seconds after loading. They then calculated from them the tensile compliance as function of temperature, D(T ). From the values of J (T ) and D(T ) they calculated the lateral contraction ratio, (T ). They wrote that they found it from (T ) = 2 J (T ) 1, D(T ) (4.14)

where they had multiplied by 2 instead of dividing by it [cf. Equation (2.10)]. According to their paper they had used the formula D(T ) = 1 1 3B(T ) + J (T ) = + 9B(T )J (T ) 3J (T ) 9B(T ) (4.15)

for D(T ) instead of the correct equation, D(T ) = J (T ) B(T ) + . 3 9 (4.16)

They reported values of (T ) increasing monotonically from about 0.35 to 0.46. These values are reasonable and make one wonder if they had indeed used the wrong equations they had reported. An interesting aspect of the work of Crowson and Arridge (1979b) is their demonstration (see Figure 7) that a difference of even only 1 C in the experimental temperatures can have a large effect on the Poissons ratio calculated from two independently measured response functions.

POISSONS RATIO IN LINEAR VISCOELASTICITY

43

Figure 7. The effect of differences in the experimental temperatures in the determination of (T ) from two functions.

Curve A represents the Poissons ratio with no difference in the temperatures. Curves B and C show the ratios calculated with +1 and 1 C difference in temperature, while Curves D and E were obtained with +2 and 2 C differences, respectively. The authors thus showed that differences in the temperature at which the two source functions were obtained can result in spurious maxima and minima in the calculated Poissons ratio. These results vividly demonstrate that the indirect determination of Poissons ratio from two independently measured source functions must be carried out following the Standard Protocol, i.e., they must be made simultaneously on the same specimen, under identical test conditions. Some indirect determinations of limited range were performed by Gilmour et al. (1979). They obtained the glassy Poissons ratio, g , of polystyrene, of poly(methyl methacrylate), and of polycarbonate using a bidirectional strain gage. Husler et al. (1987) determined Poissons ratio on a modied epoxy as a function of the temperature in the range from 0 to 50 C using an ultrasonic pulse transmission technique at a nominal frequency of 12 MHz. Poissons ratio increased linearly with temperature and decreased with an increase in the degree of crosslinking.

44 4.4. S TRAIN -D EPENDENT M EASUREMENTS

N.W. TSCHOEGL ET AL.

The experiments reviewed here were concerned only with the glassy, g , or the equilibrium Poissons ratio, e . The dependence of Poissons ratio on the level of strain is documented in some cases. Lobdell et al. (1964) performed measurements on a standard tensile testing machine at room temperature on several metallic and polymeric materials. They measured the lateral contraction with a contact (clip-gauge) extensometer positioned directly on the specimen. Our own experience showed that using contact measurements on polymeric materials near or below their glass transition temperature becomes questionable due to creep caused by the contact force. The authors showed no curves but tabulated average values for the g values of six different plastics. Laufer et al. (1978) measured the specimens volume change during uniaxial straining. They constructed a special dilatometer attached to a standard tensile testing machine. From the volume change and the simultaneously measured uniaxial deformation at constant crosshead speed they calculated the lateral contraction ratio of a nearly incompressible polybutadiene rubber. They reported values of e that decreased from 0.49633 to 0.40752 as the strain varied from 0.0085 to 0.3869. Fedors and Hong (1982) used triaxial contact extensometers. They performed their measurements at room temperature, and measured the deformation in all three perpendicular directions. The main problem that they encountered was relative movement of the specimen with respect to the stationary extensometers. They tried to solve this problem by polishing the specimen surfaces. Measurements were obtained as a function of the rate of strain at a crosshead rate of 0.002 in/min on an unlled polycarbonate (Lexan), an ABS polymer, and a lled poly(phenylene oxide) (PPO). The authors calculated Poissons ratio as lat /ax and observed no time-dependence. This is not surprising since the measurements were taken at about 23 C, that is, in the glassy region. The values of the glassy Poissons ratio, g , was about 0.35 for the Lexan and the ABS polymer, and about 0.4 for the lled PPO. Bauer and Farris (1989) also used dilatometry. They determined Poissons ratio on polyamide (Kapton) lms, and obtained the glassy ratio, g , by measuring the stress, , as a function of the pressure, P , using the equation = 1 2. P (4.17)

The measurements were made at room temperature. They found g to be 0.34 at approximately 1% strain, and 0.48 at 5%. Summerscales and Fry (1994) attempted to avoid the contact problems by measuring the extension and the contraction with strain gauges. Their data were obtained on a glass ber reinforced epoxy resin at a constant crosshead speed in an Instron tensile tester. They plotted the lateral contraction ratios as a function of the load.

POISSONS RATIO IN LINEAR VISCOELASTICITY

45

Michaeli and Dassow (1994, 1995) used a biaxial optical strain extensometer. They performed experiments on a polyamide in uniaxial tension at three different rates of deformation and at ve different temperatures between 23 and 100 C. The data showed considerable scatter. Their overall observation was that the lateral contraction ratio decreased with elongation. Weber et al. (1997) also found a strong strain dependence of Poissons ratio in their work on polyethylene foams.

5. Conclusions and Recommendations Careful consideration of the published work leads to certain inescapable conclusions. We summarize these below and also offer recommendations for work in areas which have so far been neglected. We emphasize: The current literature only too often shows a distressing ignorance of the true nature of the viscoelastic Poissons ratio. Consequently, it contains a number of factual misstatements and erroneous formulae. To date, the best determinations of the time-, frequency-, or temperaturedependent Poissons ratio were made on poly(methyl methacrylate) PMMA. Others were made on materials of complex composition. This often hinders an insightful interpretation of the data. To establish the methodology for an unambiguous determination of the viscoelastic Poissons ratio, measurements on many more well characterized unlled amorphous polymers or random copolymers are needed. To date, there apparently exists only a single successful direct determination of (f ) (see Figure 4), none of (f ), and only one more or less acceptable determination of (t) (see Figure 3). Measurements as a function of temperature are experimentally often easier to perform, and thus there exist reports on reasonably successful determinations of (T ). In any determination of Poissons ratio the measurements must be based on stress relaxation, and not on strain retardation, in the longitudinal direction. Measurements can be made in creep but must be converted (see Section 3.7). A comparison of the true Poissons ratio with that determined in creep (see Section 3.7) and also in constant rate of strain measurements (see Section 3.6) would be useful. Although direct determinations of the time- (or the frequency-) dependent Poissons ratio are desirable, successful indirect determinations would offer the possibility of a complete characterization of a linear viscoelastic material in a single experiment (cf. Section 1.2). Indirect determinations of Poissons ratio have hitherto been singularly unsuccessful. Usually they led to ratios in which there appeared maxima or minima.

46

N.W. TSCHOEGL ET AL.

Since linear viscoelastic theory asserts that (t) is a monotone nondecreasing, while () is a monotone non-increasing function of their respective arguments (cf. Sections 3.2.1 and 3.3.1), the appearance of extrema in these functions clearly indicates that there is something amiss with either the measurements or the calculations. The demonstration by Crowson and Arridge (1979b) points conclusively to small differences in test temperatures as one likely source of such extrema. The viscoelastic Poissons ratio cannot be calculated indirectly by combining response functions such as G(t) and K(t), or E () and G (), obtained in separate measurements. If any bulk response function is to be calculated from two other response functions, it is absolutely mandatory that the source functions be determined in accordance with the Standard Protocol (cf. Section 1.3). This demands that highly accurate measurements be made on the same specimen, at the same time, and under the same conditions of the experimental environment. This may require the construction of novel apparatus. In particular, if Poissons ratio is to be used in conjunction with another response function it must be determined with at least four-gure accuracy (Lu et al., 1997). The experimental procedure which may offer the best hope for a successful indirect determinations of K(t) or B(t) might be the simultaneous measurement of E(t) and (t) (Samarin et al., 2002). Yee and Takemori (1982) claimed that the error analysis they had performed using typical values convinced them that K () calculated from E () and () incurs much smaller errors than K () calculated from E () and G (). An analogous claim would presumably apply to the calculation of K(t) or G(t) from E(t) and (t). The indirect determination of Poissons ratio often requires the numerical solution of convolution integrals. Emri et al. (2002) presented an analysis of the errors inherent in these solutions. If two simultaneously determined response functions are to be converted into two others, we recommend that this be done by using the results of the work of Emri and Tschoegl (1994) on the determination of line spectra from experimental responses. This work describes the windowing algorithm, an iterative computer algorithm for generating line spectra of experimentally determined response functions (see the Appendix). It would then be possible to compare line spectra in shear and in bulk relaxation, HiG (i ) and HiK (i ) with one another, as well as line spectra in shear B and in bulk retardation, LJ i (i ) and Li (i ), not only with one another, but also with Poissons ratio line spectra, i (i ).

POISSONS RATIO IN LINEAR VISCOELASTICITY

47

Appendix: The Windowing Algorithm This algorithm iteratively computes respondance time distributions from any experimental time- or frequency-dependent linear viscoelastic response function. The output is in the shape of line spectra equally spaced on the log -axis. The algorithm is described in a series of publications: Emri and Tschoegl (1993) deals with the relaxation modulus and the creep compliance, Tschoegl and Emri (1993) with the storage and loss functions, Emri and Tschoegl (1994) with its application to experimental data, and Tschoegl and Emri (1992) with the interconversion of relaxation, retardation, and delay time distributions into one another. The calculated spectra do not, of course, have anything in common with the underlying unknown true spectra but are capable of reproducing the source functions with entirely satisfactory accuracy. The algorithm obtains each spectrum line iteratively using only the appropriate part of the source function data, scanning the complete function through a window, as it were. This has led to its sobriquet as the windowing algorithm. Emri and Tschoegl (1998) discuss its success in view of the mathematical ill-posedness of the problem. We demonstrate the use of the algorithm by outlining its application to the conversion of E(t) and (T ) data into G(t) and K(t) data. From experimental values on the rst two functions the algorithm obtains the sets {HiE , i ; i = 1, 2, . . . , NE } and {i , i ; i = 1, 2, . . . , N } where HiE and i are the strengths of the discrete spectra of the stretch modulus, E(t), and of the time-dependent Poissons ratio, (t), while the i and i are the associated relaxation times and Poisson delay times. From these the algorithm then generates the sets of Laplace transforms of E(t) and (t), and using the expressions {Ee } + E(s) = s and e (s) = s
i =N i =NE

HiE
i =1

i 1 + i s

(A.1)

i
i =1

1 . s(1 + i s)

(A.2)

The Laplace transforms of G(t) and K(t) are then obtained using the discrete data sets i (s) = G and i (s) = K i (s) E 3[1 2s i (s)] (A.4) i (s) E 2[1 + s i (s)] (A.3)

[cf. Equations (3.22)].

48

N.W. TSCHOEGL ET AL.

i (s) and K i (s) data to obtain the sets of Finally, the algorithm is applied to the G G K discrete spectra, {Hi , i ; i = 1, 2, . . . , NG } and {Hi , i ; i = 1, 2, . . . , NK }. This essentially nishes the task since any desired response function is readily obtained from the last two sets of line spectra. As an example, the bulk relaxation modulus is obtained using
i =NK

K(t) = Ke +
i =1

HiK exp(t/i ).

(A.5)

While the method just outlined is straightforward in theory, the procedure nevertheless needs careful attention. Because of the minus sign in the denominator, Equation (A.4) in particular requires highly accurate data. Acknowledgement The authors gratefully acknowledge the cooperation of Matja Samarin of the Center of Experimental Mechanics in assembling and excerpting references. Note
1 Application of the correspondence principle is limited to cases where the time dependence of

the material properties is separable from that of the spatial boundary conditions.

References
Adams, R.D. and Peppiatt, N.A., Effect of Poissons ratio strains in adherends on stresses of an idealized lap joint, J. Strain Anal. 8, 1973, 134139. Bauer C.L. and Farris, R.J., Determination of Poissons ratio for polyimide lms, Polym. Engrg. Sci. 29, 1989, 11071110. Baumgaertel, M. and Winter, H.H., Interrelation between continuous and discrete relaxation time spectra, J. Non-Newtonian Fluid Mech. 44, 1992, 1536. Blatz, P.J. and Ko, W.L., Application of nite elastic theory to the deformation of rubbery materials, Trans. Soc. Rheol. 6, 1962, 223251. Chang, W.V., Bloch, R. and Tschoegl, N.W., Study of the viscoelastic behavior of uncrosslinked (gum) rubbers in moderately large deformations, J. Polymer Sci., Polymer Phys. Ed. 15, 1977, 923944. Clayton, D., Darlington, M.W. and Hall, M.M., Tensile creep modulus, creep lateral contraction ratio and torsional creep measurements on small nonrigid specimens, J. Phys. E: Sci. Instr. 6, 1973, 218226. Crowson, R.J. and Arridge, R.G.C., Linear viscoelastic properties of epoxy resin polymers in dilatation and shear in the glass transition region. 1. Time-temperature Superposition of creep data, Polymer 20, 1979a, 737746. Crowson, R.J. and Arridge, R.G.C., Linear viscoelastic properties of epoxy resin polymers in dilatation and shear in the glass transition region: 2. Measurement of the glass transition temperature, Polymer 20, 1979b, 747754. Darlington, M.W. and Saunders, D.W., The tensile creep behaviour of cold-drawn low-density polyethylene, J. Phys. D: Appl. Phys. 3, 1970, 535549.

POISSONS RATIO IN LINEAR VISCOELASTICITY

49

Delin, M., Rychwalski, R.W., Kubt, M.J. and Kubt, J., Volume changes during stress relaxation in polyethylene, Rheol. Acta 34, 1995, 182195. Deng, T.H. and Knauss, W.G., The temperature and frequency dependence of the bulk compliance of poly(vinyl acetate). A re-examination, Mech. Time-Dependent Mater. 1, 1997, 3349. Emri, I. and Tschoegl, N.W., Generating line spectra from experimental responses. Part I: Relaxation modulus and creep compliance, Rheol. Acta 32, 1993, 311321. Emri, I. and Tschoegl, N.W., Generating line spectra from experimental responses. Part IV: Application to experimental data, Rheol. Acta 33, 1994, 6070. Emri, I. and Tschoegl, N.W., An iterative computer algorithm for generating line spectra from linear viscoelastic response functions, Internat. J. Polym. Mater. 40, 1998, 5579. Emri, I., Nikonov, A. and Tschoegl, N.W., Numerical errors in the interconversion of creep and relaxation functions, Rheol. Acta, 2002, to be submitted. Eringen, A.C., Non-Linear Theory of Continuous Media, McGraw-Hill, New York, 1962, 177. Fedors, R.F. and Hong, S.D., A new technique for measuring Poissons ratio, J. Polym. Sci., Polym. Phys. Ed. 20, 1982, 777781. Fellahi, B., Fisa, B. and Favis, B.D., Welding strength in injection molded HDPE/PA6 blends. Inuence of interfacial modication, J. Appl. Polym. Sci. 57, 1995, 13191332. Ferry, J.D., Viscoelastic Properties of Polymers, 3rd edn, John Wiley and Sons, New York, 1980. Fisa, B., Favis, B.D. and Bourgeois, S., Injection molding of polypropylene/polycarbonate blends, Polym. Engrg. Sci. 30, 1990, 1051. Freudenthal, A.M. and Henry, L.A., On Poissons ratio in linear viscoelastic propellants, Progr. Aeronautics Rocketry 1, 1960, 3366. 3466. Gent, A.N. and Hwang, Y.-C., Elastic behavior of a double layer bonded between two rigid spheres, Rubber Chem. Tech. 61, 1988, 630638. Gilmour, I.W., Trainor, A. and Haward, R.N., Elastic moduli of glassy polymers at low strains, J. Appl. Polym. Sci. 23, 1979, 31293138. Giovagnoni, M., On the direct measurement of the dynamic Poissons ratio, Mech. Mater. 17, 1994, 3346. Harris, J.A. and Adams, R.D., Strength predictions of bonded single loop joints by non-linear FEM, Internat. J. Adhesion Adhesives 4, 1984, 6578. Husler K.G., Hauptmann, P., Meischner, C., Fedtke, M., Hartel, E. and Wartewig, S., Ultrasonic investigations of modied epoxies, Polym. Commun. 28, 1987, 154157. Heydemann, P., On the minimum of Poissons ratio in polymers, Kolloid-Z. u. Z. f. Polymere 193, 1963, 1215. Hopkins, I.L. and Hamming, R.W., On creep and relaxation, J. Appl. Phys. 28, 1957, 906. Karger-Kocsis, J. and Csikai, I., Skin-core morphology and failure of injection-molded specimens of impact-modied polypropylene blends, Polym. Engrg. Sci. 27, 1987, 241. Kstner, S. and Pohl, G., Ein Beitrag zur Frage der vollstndigen Erfassung des mechanischen Relaxationsverhaltens der Polymeren, Kolloid-Z. u. Z. f. Polymere 191, 1963, 114. Kimoto M., Nagata, I., Minowa, A., Moriwaki, K. and Watanabe, T., Evaluation of disbondings and measurement of Poissons ratio for plastic composites using holographic interferometry, J. Appl. Polym. Sci. 40, 1990, 10851093. Kinloch, A.J., Adhesion and Adhesives, Chapman and Hall, London, 1990. Knauss, W.G., The investigation of the mechanical properties of thermoviscoelastic materials, Soc. Exp. Mech. 1, 1992, 2226 (Proceedings of the VII International Congress on Experimental Mechanics, Las Vegas, NV). Knauss, W.G. and Emri, I., Non-linear viscoelasticity based on free volume consideration, Comput, & Structures 13, 1981, 123128; also in: Computational Methods in Nonlinear Structure and Solid Mechanics, A.K. Noor and H.G. McComb (eds), Pergamon Press, Oxford, 1981, 123128. Kono, R., The dynamic bulk viscosity of polystyrene and polymethyl methacrylate, J. Phys. Soc. Japan 15(4), 1960, 718725.

50

N.W. TSCHOEGL ET AL.

Kono, R., The dynamic bulk and shear viscosity of high polymers. I, J. Phys. Soc. Japan 16, 1961a, 15801586. Kono, R., The dynamic bulk and shear viscosities of polyvinylchloride, J. Phys. Soc. Japan 16, 1961b, 17931794. Kono, R. and Yoshizaki, H., Viscoelastic properties of polyvinyl-i-butyl ethers at high frequencies, Japan J. Appl. Phys. 12(3), 1973, 445457. Koppelmann, V.J., ber den dynamischen Elastizittsmodul von Polymethacrylsuremethylester bei sehr tiefen Frequenzen, Kolloid-Z. 164, 1959, 3134. Kubt, M.J., Vernel, J., Rychwalski, R.W. and Kubt, J., Bulk moduli from physical aging and stress relaxation data, Polym. Engrg. Sci. 38(8), 1998, 1261. Kugler, H.P., Stacer, R.G. and Steimle, C., Direct measurements of Poissons ratio in elastomers, Rubber Chem. Tech. 63, 1990, 473487. Ladizesky, N.H. and Ward, I.M., Determination of Poissons ratio and Youngs modulus of lowdensity polyethylene, J. Macromol. Sci.-Phys. B5(4), 1971, 661692. Laufer, Z., Diamant, Y., Gill, M. and Fortuna, G., A simple dilatometric method for determining Poissons ratio of nearly incompressible elastomers, Internat. J. Polym. Mater. 6, 1978, 159174. Lee, E.H., Viscoelastic stress analysis, in Structural Mechanics, J.N. Goodier and N.J. Hoff (eds), Pergamon Press, Oxford, 1960, 456482. Lobdell, A.J., Shinopulos, G.F. and Fillio, D.N., An instrument to measure transverse strain, Mater. Res. Stand. 4, 1964, 811. Lu, H., Zhang, X. and Knauss, W.G., Uniaxial, shear and Poisson relaxation and their conversion to bulk relaxation, Polymer Composites 18(2), 1997, 211222; J. Polym. Sci. Engrg. 37(6), 1997, 1053106. Malvern, L.E., Introduction to the Mechanics of a Continuous Medium, Prentice-Hall, Englewood Cliffs, NJ, 1969, 293. Michaeli, W. and Dassow, J., Describing Poissons ratio of thermoplastics as a non-linear viscoelastic parameter, in Proceedings 52nd Annual Technical Conference, ANTEC 94, San Francisco, CA, Society of Plastic Engineers, Brookeld, 1994, 18041806. Reprinted in British Plastics and Rubber, MCM Publishing, London, 1995, 2124. Migwi, C.M., Darby, M.I., Wolstenholm, G.H., Yates, B., Duffy, R. and Moss, M., A method for determining the shear modulus and Poissons ratio of polymer materials, J. Mater. Sci. 29, 1994, 34303432. Morita, E., Kono, R. and Yoshizaki, H., Bulk and shear relaxation processes in poly-i-butyl methacrylate, Japan J. Appl. Phys. 7(5), 1968, 451461. Nikonov, A., Tschoegl, N.W. and Emri, I., Numerical errors in the interconversion of creep and relaxation functions, Rheol. Acta, 2001, submitted. Onogi, S. and Ui, K., Frequency and temperature dispersions of high polymers, J. Colloid Sci. 11, 1956, 214. Philippoff, W. and Brodnyan, J., Preliminary results in measuring dynamic compressibilities, J. Appl. Phys. 26(7), 1955, 846849. Richardson, I.D. and Ward, I.M., Temperature dependence of the Poissons ratios in low-density polyethylene with parallel lamellas morphology, J. Polym. Sci. Polym. Phys. Ed. 16, 1978, 667 678. Rigby, Z., Phase shifts in stress-strain relationships of viscoelastic materials, Rheol. Acta 5, 1965, 2829. Rigbi, Z., The value of Poissons ratio of viscoelastic materials, Appl. Polym. Symp. 5, 1967, 18. Samarin, M., Tschoegl, N.W. and Emri, I., The simultaneous determination of Youngs modulus and the viscoelastic Poissons ratio, Mech. Time-Dependent Mater., 2002, to be submitted. Sane, S. and Knauss, W.G., The time-dependent bulk response of poly(methyl methacrylate), Mech. Time-Dependent Mater. 5, 2001, 293324.

POISSONS RATIO IN LINEAR VISCOELASTICITY

51

Smith, T.L., Volume changes and dewetting in glass bead Polyvinyl chloride elastomeric composites under large deformations, Trans. Soc. Rheol. 3, 1959, 113136. Summerscales, J. and Fry, S.A., Poissons ratio in bre-reinforced polymer composites with a high void content, J. Mater. Sci. Lett. 13, 1994, 912914. Theocaris, P.S., Creep and relaxation contraction ratio of linear viscoelastic materials, J. Mech. Phys. Solids 12, 1964, 125138. Thomson, K.C., On the complex Poissons ratio of a urethane rubber compound, J. Appl. Polym. Sci. 10, 1966, 11331136. Tschoegl, N.W., The Phenomenological Theory of Linear Viscoelastic Behavior, Springer-Verlag, Heidelberg, 1989. Tschoegl, N.W., Time dependence in material properties: An overview, Mech. Time-Dependent Mater. 1, 1997, 131. Tschoegl, N.W. and Emri, I., Generating line spectra from experimental responses. Part III: Interconversion between relaxation and retardation behavior, Internat. J. Polym. Mater. 18, 1992, 117127. Tschoegl, N.W. and Emri, I., Generating line spectra from experimental responses. Part II: Storage and loss functions, Rheol. Acta 32, 1993, 322327. Tschoegl, N.W., Knauss, W.G. and Emri, I., The effect of temperature and pressure on the mechanical properties of thermo- and/or piezorheologically simple polymeric materials in thermodynamic equilibrium A critical review, Mech. Time-Dependent Mater. 6, 2002, 5399. Tsou, A.H., Greener, J. and Smith, G.D., Stress relaxation of polymer lms in bending, Polymer 36(5), 1995, 949954. van der Varst, P.G.Th. and Kortsmit, W.G., Notes on the lateral contraction of linear isotropic viscoelastic materials, Appl. Mech. 62, 1992, 338346. Waterman, H.A., Determination of the complex moduli of viscoelastic materials with the ultrasonic pulse method (Part I), Kolloid-Z. u. Z. f. Polymere 192, 1963a, 18. Waterman, H.A., Determination of the complex moduli of viscoelastic materials with the ultrasonic pulse method (Part II), Kolloid-Z. u. Z. f. Polymere 192, 1963b, 916. Waterman, H.A., Relation between loss angles in isotropic linear viscoelastic materials, Rheol. Acta 16, 1977, 3142. Weber, H., Wolf, T. and Dnger, U., Determination of relaxation moduli and Poissons ratio in uniaxially loaded solid polyethylene foam specimens as part of full material characterization, Mech. Time-Dependent Mater. 1, 1997, 195208. Wilson, I., Cunningham, A. and Ward, I.M., The determination of Poissons ratio compliances for polyethylene terephthalate sheets using a Michelson interferometer, J. Mater Sci. 11, 1976, 21812188. Wilson, I., Ladizesky, N.H. and Ward, I.M., The determination of Poisson ratio and extensional modulus for polyethylene terephthalate sheets by an optical technique, J. Mater. Sci. 11, 1976, 21772180. Yee, A.F., Dynamic-mechanical studies of some polymers containing the bisphenol-A units, ACS Polymer Preprints 21(2), 1980, 285286. Yee, A.F. and Takemori, M.T., A new technique for measuring volume and shear relaxation in solid polymers: Dynamic Poissons ratio, ACS Polymer Preprints 20(1), 1979, 758761. Yee, A.F. and Takemori, M.T., Dynamic bulk and shear relaxation in glassy polymers. I. Experimental techniques and results on PMMA, J. Polym. Sci., Polym. Phys. Ed. 20, 1982, 205224. Zihlif, A.M., Duckett, R.A. and Ward, I.M., The determination of the lateral compliances and Poissons ratios for highly oriented polyethylene sheets, J. Mater. Sci. 17, 1982, 11251130.

Вам также может понравиться