LL ————————————— Rheometry K. WALTERS. Professor of Applied Mathematics University College of Wales Aberssewyth @ CHAPMAN AND HALL - LONDONie First published 1975 by Chapman and Hall Led 1] New Fetter Lane, London EC4P 4EE Typeset by Preface Lid Salisbury, Wiltshire Printed in Great Britain by Whitstable Litho Whitstable, Kent ISBN 0 412 120909 © 1975 K. Walters All rights reserved. No part of this book may be reprinted, or reproduced or utilized in any form or by any electronic, mechanical or ‘other means, now known or hereafter invented, ingluding photocopying and recording. oF in any information storage and retrieval system, without permission in writing from the Publisher Distributed in U.S.A. by Halsted Press, «a Division of John Wiley & Sons, Ine., New York Library of Congress Catalog Card Number 75-2105 Preface Many industries, government research establishments and university research groups are concerned with a wide range of fluids which ean te broadly classified as ‘non-Newtonian’; polymer melts and polymer solutions come readily to mind in this connection, but the complete list is seemingly endless. Those who are practically concerned with the flow behaviour of these fluids invariably require measurements of their mechanical properties, so that “Kheometry’ plays a major role in the expetimentalists’ approach to non-Newtonian fluid mechanics. The theoretician is also concerned, since apart from his interest in supplying the background theory to a variety of existing rheometers, he is also involved in suggesting and describing new thcometer flow Situations and these are becoming progressively more sophisticated. ‘At the 6th International Congress on Rheology held at Lyon, France, in September 1972, two new commercially available rheometers were displayed — the Mechanical Spectrometer (Rheomettics Inc.) and the Rotary Rheometer (Instron Ltd.), together with established instruments such as the Weissenberg theogoniometer (Sangamo Weston Controls Ltd.) and the Balance Rheometer (Contraves, A.G.), These are sophisticated instruments requiring a working knowledge of both theoretical and experimental ‘heology on the part of those who use them. The present book is. intended to be a text book for such instruments as well as providing a background to ‘Rheometry” in general “There are two basic objectives in Rheometry. The first involves straightforward attempt to characterize the behaviour ofoj vi Preface non-Newtonian liquids in a number of simple (theometrical) flow situations, with a view to correlating material behaviour with either ‘molecular structure or observed behaviour in practical situations, The second objective concerns the construction of rheological equations Of state for the liquids which can be later used in the solution of flow problems of practical importance. To accomplish these objectives, especially the second, a working knowledge of theoretical theology {a difficult subject by common consent) is essential. This book is written by a Mathematician, but a genuine attempt has been made to avoid unnecessary mathematical rigour. It is hoped that the overall impression is one of sympathy for the needs of those with only a limited mathematical background, ‘The author was introduced to practical aspects of the subject of Rheometry during a three months visit to the Dallas Field Research Laboratory of the Mobil Oil Corporation in 1964 and the help and encouragement of Mr. J. G. Savins in particular were largely instrumental in motivating the formation of an experimentat programme within the author's research group. A further visit to the U.S.A. in 1973, (this time to the Mathematics Research Center of the University of Wisconsin) was also beneficial in formulating the general structure of this book. Stimulating discussions with Professors R. B. Bird and A. S. Lodge were of significant help in this connection. Lowe a great debt of gratitude to Professor J. G. Oldroyd who first awakened my interest in Rheology as a graduate student, and to Professors T. V. Davies and J. Heading who have in turn provided 1 very congenial atmosphere for research within the Applied Mathematics Department of the University College of Wales, Aberystwyth. Since 1967, experimental work within our research group hs been in the capable hands of Dr. J. M. Broadbent. His quiet and helpful influence is apparent in various sections of this book. ‘Some of the work contained in the book is a summary of research carried out at U.C.W., Aberystwyth with a succession of research students. | am happy to acknowledge their significant contributions to the text. Finally, Mr. J. F, Hutton of Shell Research Ltd. and my colleague, Dr. W. M. Jones, read an early draft of the text and made a number of useful suggestions for improvement, K. WALTERS, October 1974 © Contents Preface 1. Introduction and basic equations LL Viseoetasticity 1.2 From viscometry to eheometry 1.3 Basic equations 2. Rheological equations of state 2.1 Basie principles 2.2. Application of the theory ~ the simple fluid of Coleman and Noll 2.3. Other general rheological equations of state 24 Some relatively simple rheological equations of state 2.5 Restricted flow situations 2.6 Boundary conditions 2.7 Conclusion je sheomettical concepts 3.1 Steady simple shear flow 3.2. Small-amplitude oscillatory-shear flow 3.3. Combined steady and oscillatory shear flow 3.4 Extensional flow 3.5. The measurement of material functions in various geometries 3.6 The measurement of extensional viscosity in various geometries 3 7 18 19 20 21 2 2 28 30 31 32 404. The measurement of the material functions using 9 rheogoniometer 4.1 Introduction 4.2. Basic theory for cone-and-plate flow 4.3 Basie theory for torsional flow 4.4 Basic theory for Couette flow 4.5. Basic equations for other sheogoniometer flows, 4.6 Possible sources of error 46.1 The shear rate in the cone-andt-plate rheometer 4.6.2. Inertial effects 4.6.3 Edge and end effects 4.6.4 Hole-pressure errors 4.6.5 Viscous heating 4.6,6 Instrument imperfections 4.6.7 Errors inkerent In the interpretation of experimental results 4.6.8 Miscellaneous sources of error 4.7 Experimental ‘The measurement of the material functions using capillaries, slits and similar devices S.1 General introduction 5.2 Flow in capillaries and slits S21 Introduction 5.22. Basic theory for normal stress and shear stress ‘measurements in capillary tubes 5.2.3 Basie theory for normal stress and shear stress ‘measurements in slit dies 5.24 Possible sources of error and experimental results {for low Reynolds-number flows 5.2.5. Possible sources of error and experimental results for high Reynolds-number flows 5.3. Determination of the second normal stress difference using axial annular flow 6. ‘The measurement of the complex dynamic viscosity and related functions 6.1 Introduction 6.2. The determination of the complex viscosity from conventional rheometers Contents 44 45 48 33 56 59 60 6 66 15 80 82 84 85 87 93 ys 94 94 94 102 105 ut 1g. 120 120 123 Contents 6.2.1 Basic equations 6.22 Theory for oscillatory shear between parallel plates and other geometries 6.2.3 Possible sources of error Inertial effects Non-linear effects ‘Natural frequency problems Stiffness problems Miscellaneous sources of error 6.3. Experiments involving free oscillations 6.4. The determination of the complex viscosity from the new rheometers 64.1 lntroduction and basic theorem 64.2. Basic theory for the orthogonal rheometer 64.3 Basic theory for the balance rheometer 6.44 Basic theory for the eccentric-cylinder rheometer and similar devices 6.4.5. Possible sources of error Inertial effects Non-linear effects Edge and end effects Friction and lag ‘Miscellaneous sources of error 6.4.6 Conclusion and recent developments 6.5 Combined steady and oscillatory shear 7. ‘The measurement of extensional viscosity 7.1 Introduction 7.2. The importance of extensional viscosity 7.3. Theoretical considerations 7.4 Methods of measurement 7.4.1 Controllable experiments 74.2 Non-controllable experiments 7.5. Experimental results 8, Novel rheometrical flow experiments 8.1 Introduction 8.2 Approximate methods of determining v3 8.3 Restricted flow conditions 123 128 133 133 140 Ist 155 160 160 164 164 168 2 176, 181 181 183 186 190 193 + 197 201 210 210 213 216 219 220 222 234 234 235 238 ES8.4 Simple unsteady flows 8.5 Complex flow situations ‘Appendix I Equations of motion and continuity in cylindrical-polar and spherical-polar co-ordinat Appendix 2 Measurement of phase lag in oscillatory testing ‘Appendix 3 Determination of the complex viscosity from oscillatory testing with parallel-plate geometry ‘Appendix 4 Determination of decay parameter Basic notation References Index Contents 246 247 251 253 255 357 259 262 m7 CHAPTER! Introduction and Basic Equations 1.1. Viscoetast y In the development of classical mechanics, the distinction between solids and liquids was assumed to be quite sharp and separate physical laws were formulated to account for the behaviour of each the solids obeying Hooke’s kaw and the fluids Newton’s law of constant viscosity. It is now common knowledge that many materials cannot be described by either of these classical extremes. Classical elasticity can of course be extended to meet the finite strains which are often involved and the shear behaviour of many seemingly inelastic liquids may require a shear-dependent viscosity to describe it. But these generalizations (which give rise to finite-strain elasticity theory and inelastic novr-Newtonian fluid mechanics) can still be regarded as being al the opposite ends of a spectrum of possible material behaviour. They are completely inadequate to describe the behaviour of a wide class of materials that may be broadly described as ‘viscoelastic’. Such materials are called viscoelastic solids if they do not continually change their shape when subjected to stresses: and elastico-viscous (or simply ‘elastic’ liquids if they do change their shape continually when subjected to stresses, irrespective of how small these stresses may be. In this book, we shall be concerned very largely with the mechanical behaviour of elastic liquids. this being an important part of a more general study of the deformation and flow of matter, which is now embraced by the term ‘Rheology’. We shall further limit attention to liquids that are both isotropic and€ 2 Rheometry incompressible and we shall not refer explicitly to (visco-plastic) liquids having a yield stress. Pethaps the most familiar examples of elastic liquids are the polymer melts and polymer solutions associated with the synthetic fibre and plastics industries, and certainly, on any commercial basis, it is this area which provides the main motivation and funding for any research effort in rheometrical studies. However, it would be ‘wrong to conclude that other examples of elastico-viscous behaviour are not plentiful. For example, the polymeric V.L. improved oils, $0 ‘well known to the motorist, can certainly be classified as elastic liquids and there is strong evidence to show that even the base mineral oils exhibit elastico-viscous behaviour under operating conditions{ 1,2]. Lubricating greases, liquid detergents, printing inks, asphalt, bitumens and paints all manifest rheologically complex. material behaviour[3]. Some of them exhibit time-dependent (thixotropic’ and ‘rheopectic’) behaviour in addition to viscoelasticity (o further complicate a complex situation.* In the field of biology, many body Muids are now known to be elastico-viscous in behaviour, including blood, synovial fuid and various forms of mucus. It is not surprising, therefore, that Rheology “is becoming increasingly more important in medical research and there is now an International Society of *Biorheology’ to look after these and related developments. 1.2. From viscometry to rheometry A-constant viscosity coelticient is sufficient to determine the behaviour of incompressible Newonian liquids under any conditions of motion and stress, The measurement of this viscosity coefficient involves the use of a VISCOMETER defined simply as “an instrument for the measurement of viscosity’, The viscosity of non-Newtonian clastic liquids may be shear-rate dependent and this is certsinly not the only complication of viscoelasticity, as we shall see, The viscometer is therefore inadequate to characterize the behaviour of these materials and has to be replaced by a RHEOMETER defined as ‘an instrument for measuring rheological properties’. The definition ‘tn this book, we do not explicitly consider this type of timedependent behaviour C Introduction and Basic Equations 3 is purposely vague for reasons that will become all too apparent later. The advent of the rhoometer has given rise to the science of “Rheometry’ There appear to be two basic (not necessarily unrelated) objectives in Rivometry, The First (which we shall refer to as Objective 1) involves a straightforward attempt to determine the behaviour of non-Newtonian liquids in a number of simple (theometrical) flow situations using suitably defined material functions. The simple desire here is to seek a correlation between molecular structure and ‘material behaviour or alternatively between material properties and ‘observed behaviour in practical situations, Such w study sometimes. involves a search for suitably defined non-dimensional parameters to augment the ‘Reynolds number" Found so useful in classieal fluid imcluanies. However, this is only one small part of the correlation process. ‘The second objective (which we shall refer to as Objective 2) is more sophisticated and decidedly more difficult. It involves the prediction of behaviour in nion-simple flow situations from the results of simple rheometrical experiments. The rheometer data here is used to construct rheological equations of state (constitutive equations) which may then be used together with the stiess equations of motion and wuutinuity to predict behaviour in more pructical situations. The more data that is available, the more accurate will be the rheological model, so that a very detailed research programme is required to meet this objective. The work is Trauight with problems inherent in the complex nature of the materials being studied and there is no guarantee of general success. 1,3 Basic equations ‘This book is written by an Applied Mathematician and there will inevitably be a tendency to stress certain theoretical aspects of the subject in a way which may appear unnecessary to the reader who is basically concerned with more practical aspects of the problem. Such readers (especially those who ate concerned only with Objective 1) may find it unnecessary to study in depth the basic theoretical treatment given in this Chapter and in Chapters 2 and 3 and may. wish to concentrate on those parts of the book that are mainly concerned with rheological instruments and their functioning. At the same time, it eannot be overemphasized that the attitude whieh4 Rheometry considers mathematical expertise of any degree unnecessary in the study of Rheometry is not only old-fashioned, it is false! 1t would appear that this unfortunate attitude is more prevalent in the U.K. than elsewhere[4]. Following this line of thought still further, we remark that the fact that a steady simple shear flow gives rise to normal stresses as well as a shear stress in theologically complex materials (cf. Chapter 3) is sulficient motivation for the requirement of a working knowledge oF tensor analysis. In this book, we use standard tensor notation covariant suffixes are written below and contravariant suffixes above and the usual summation convention for repeated suffixes is assumed. Brackets placed round suffixes are used to denote ‘physical components’ of tensors, In a rectangular Cartesian reference frame, it is of course unnecessary to distinguish between covariant, contravariant and physical components To describe material behaviour, we require certain dynamic and kinematic variables. In the present chapter, we introduce the stress tensor as the basic dynamic variable and show how it can be used to develop the stress equations of motion, these being a direct Application of Newton's second law of motion. We shall also derive the equation of continuity, which is a mathematical statement of the “conservation-of mass’ principle. (a) The stress tensor Consider a small plane surface of area As drawn in a deforming medium (Fig. 1.1). Let, represent the components of the unit el Fig 1 Introduction and Basic Equations 5 Fig. 1.2 normal vector to the surface. It points in the direction of the *+ve” side of the surface. We say that the material on the positive side of the surface exerts @ foree /77As on the material on the negative side, it being implicitly assumed that the area As is small enough for the ‘stress’ AY") to be regarded as constant over the surface. Referred to suitably defined rectangular Cartesian co-ordinates (2), the stress will have components Py. Pay » Pas (Say). the first index: referring to the orientation of the plane surface and the second to the direction of the stress. Our sign convention is such that a positive Pex represents a tension (Fig. 1.2). The stress py (and similarly Py y and pez) are termed “normal stresses’ and pyy.Pxs etc. are called “shear stresses’ It can be shown that on transforming from one Cartesian co-ordinate system to another the set of nine quantities pj, transform as the components of a second-order tensor. In the absence of body couples, the stress tensor is symmetric, Le. pre = Pky and there are only six independent components of the tensor. (b) The stress equations of motion Consider an arbitrary volume V fixed in space, bounded by a surface ‘Sand let n, represent the components of the outward drawn unit hormal vector (Fig. 1.3). The rate of change of linear momentum, within V and the rate of outflow of momentum across S are ‘governed by the body forces throughout / and the surface forces over 5. The relevant balance can be expressed in the form Jf ewase fffeniar= [Jf 0 Rt ax ay6 Rheometry | f q 7 Fig 13 where pis the density of the fluid F, denotes the components of the body-force vector per unit mass and! D/DF represents the material derivative given in Cartesian co-ordinates by 24,2 1.2) Dior ax" oe vg representing the components of the velocity vector. An appliation of the Divergence Theorem to the First term on the left hha side enables (1,1) bo be written in the sltersative form « [2p bu; (00, FRE er oot ae aay Since the volume V is arbitrary, we require the integrand to be zero at every point of V, Therefore 2pm, Ox +R =e aay Dr We shall refer to the three equations embodied in (1.4) as the stress equations of motion. In component form, we have Aes Wx , Bes 5 op wp Dee aa Wan, Pies yp mp, 1s ax oy az 7 Pte Dt cD Bay 4 Pwr 5 Pm 4 pp, = p be 6) ax ay” a Dr Apes | Pye Wet, gp 2 DH Bias , BBs 4 Be Bes 1 ox ay oe ter Dr CED Introduction and Basic Equations Coq “The corresponding eauations in eylindrieal polar and spherical polar co-ordinates are given in Appendix 1 (e) The equation of continuity Ir there are no net sources or sinks within V, the rate of inflow of ‘mass across S must be balanced by the rate of increase of mass within Voie. Jf eames fff ar as Applying the Divergence Theorem to the left hand side of this equation, we obtain ni oe Blaceo a9) Since the volume Vis arbitrary, the integrand must be zero at every point of V, and we have (ou) Mop oe wl ant Or a.toy be, oH plier aan If the Nuid is incompressible, Do/D1 =O, and we require uy ax =o, a2) oe ay or 13) “The corresponding equations for cylindrical polar and spherical polar ‘co-ordinates are given in Appendix 1 The four equations embodied in (1.4) and (1.12) are not in themselves sufficient to determine the response of a material to an arbitrary state of stress, We require, in addition, tensor equations relating the stress and suitably defined kinematic variables. These are8 Rheometry known as theological equations of state or constitutive equations. We consider such equations in detail in the next chapter. , In view of the importance of the Newtonian viscous liquid in the study of mobile liquid systems, we shall anticipate the full discussion on theological equations given in the next chapter by documenting. the relevant equations for this classical extreme. ‘The equations of state for an incompressible Newtonian liquid can be written in the form Pa = Pb + 2nel?, (sy where p is an arbitrary isotropic pressure, 5. the kronecker delta, 1nis the constant viscosity coefficient and ef is the rate-oF strain tensor, given in terms of the velocity vector v, by ue, ae ei Las: we 4 (15) Substituting (1.14) into the stress equations of motion (1.5)-(1.1), nthe ‘Navier-Stokes’ equations oy [PH ary «Ge ag [2 5 2 ee aan ax? ay? a (1.18) Many common fluids such as water and air as well as most of the organic solvents used in the preparation of elastico-viscous polymer solutions are adequately described by the Navier-Stokes equations It is not surprising, therefore, that these equations occupy a prominent position in the study of continuum mechanics as well as providing a useful starting point for the study of rheologically complex fluids. CHAPTER 2 Rheological Equations of State Since one of the objectives of Rheometry is to assist in the construction of rheological equations of state, itis clearly in order to sive detailed consideration to this important topic. As the study of theoretical continuum mechanics has developed over the last three decades, two distinct approaches have emerged, The first, mainly sociated with the names of Oldroyd15,61 and Lodget7.8}, employs a (body) convected co-ordinate system to meet the desired aim of constructing theological equations of state of general validity. The second, associated with the names of Rivlin, Ericksen, Green, Coleman, Noll and others[9—18], employs Cartesian tensors to accomplish essentially the same end. Amongst interested non-theoreticians there is sometimes confusion conceming the respective formulation procedures and resulting equations,* ‘There is often the erroneous impression that since the respective notation and outlooks are different, the two approaches are mutually contradictory. This is, of course, not the case, and it is certainly possible to relate the alternative approaches. In the present book, we draw heavily on the notation and ideas of Oldroyd, although we shall attempt to give proper emphasis to the important contributions of Rivlin, Ericksen, Green, Coleman, Noll, ef al “This is borne out by an observation of Professor A. S. Lodge (pevate communication) concerning the response ¢o his paper (written jointly with 3 H Stark{19]), which shows that the simple Nuid: of Coleman and Noll fs included in Oldcoya's general classifications, The paper has apparently generated considerable interest, and the requests for reprints have been unusually high, A Felated paper by the present author! 20] generated a similar responseC 10 Rheometry 2.1. Basie principles Equations of state for elastic liquids have to satisfy certain formulation principles, These may be stated as follows{5] {@) The equations of state must be consistent with the requirement that the behaviour of a material element depends only Cn its previous theological history and not in any way on the state of neighbouring elements, In order to satisfy this principle, it would clearly be desirable to work with a Lagrangean reference. However, in any flow problem, equations of state have to be solved in conjunction with the familiar equations of motion and continuity (Chapter 1), together with certain prescribed boundary conditions, and an ‘Eulerian’ system is best suited for this purpose.* The basic problem is therefore one of translating @ situation which, from a physical standpoint, is best suited to a Lagrangean approach to ‘an Eulerian reference which has significant advantages from @ mathematical standpoint. (Gi) The equations of state must be consistent with the requirement that the behaviour of a material element does not depend on the translatory or rotatory motion of the material as a whole in space. In one of its simplest forms. the principle implies that a material undergoing a rigid-body motion can experience no change in internal stress on account of the motion over and above any isotropic pressure generated by stich a motion. A ‘charge in shape’ (or deformation) must be present if there is to be any change in stress. ‘This principle has been stressed by a number of theoreticians, many of whom prefer to regard it us being concerned with indifference to an observer (identified with a moving rectangular Cartesian frame) rather than indifference to the absolute motion of the material in space (as in the formulation of Oldroyd) but the two approaches are essentially equivalent when ‘indifference to an observer is given a proper physical interpretation(8,19,20]. ‘The usual “frame — indifference’ requirement of any physical theory can be thought of as being included in this general principle ‘Oldroyd[5} was able to satisfy the two principles given above by introducing a convected co-ordinate system ¥ drawn in the material ‘Papers by Lodge(21] and Kramer22] contain noteworthy efforts to solve ‘How problems without specific reference to fixed coordinates, The sparsity of Such paper's, however, an indication of the problems involved in this procedure Cr and deforming continuously with it. This co-ordinate system has the property that a material element which is at &/ at time ¢ will be at the Sime position (referred to the convected co-ordinate system) at all ther times. There is an unlimited number of co-ordinate systems of the type envisaged here, but the equations & = E'() * relating two sets of convected co-ordinates # and &/ do not involve the time explicitly. This means that convected components of tensors associated with the same point of the material can be manipulated in the same way as the fixed components of tensors at a fixed point in space. Since principle (i) involves the consideration of a given ‘material element over a period of time, the convected co-ordinate system is a suitable frame of reference with which to work “The conveeted co-ordinate system has the further advantage oF being unaffected by any superposed rigid body motion (the co-ordinate surfaces being embedded in the material), This means that, provided one works in a convected co-ordinate system and uses variables and operations which do not introduce any dependence lupon absolute motion in space, the resulting equations of state ‘automatically satisfy the second principle. ‘A suitable kinematical variable follows naturally from a description of the material properties in terms of a convected co-ordinate system. Since all quantities which refer to absolute notion in space are irrelevant, one is interested in the relative distange between and the relative motion of the parts of the arbitrary element at £), The length ds between elements at g/ and &/ + de! is given by Rheological Equations of State sce 91? = yl del dz, << ey where ris taken to represent the present time and /” an earlier time Ue is clearly independent of fo that 7,(&,0') is direct measure Of dstr?) and isan obvious choice as the kinematical variable. (tis not the only one, since 7", r) can be taken as @ measure of area-segment changes [261). ‘We note here that for some purposes a more convenient deformation variable may be-yi(E, 1.) ~ (Es 1s since small values of this variable correspond to small deformations. “an operator which is likely to occur frequently in equations of state is that which corresponds to a time derivative holding convected ‘The argument & is meant to imply &, 2, 212 Rheometry corordinates constant, denoted by D/Dt. This operator (and the corresponding integral operator) are of significant importance in the formulation of equations of state for materials with memory. For example, such quantities as Dy(&,')/Dt! ... D™yp( VIDE ate important kinematical quantities, being measures of the relative ‘motion of the parts of the arbitrary element at f! In convected co-ordinates, the covariant stress tensor 3 (E, 1) is taken as the dynamic variable, This is related to the fixed rectangular Cartesian tensor pix defined in Chapter 1 by the usual covariant tensor transformation law, i. _ ox! axt TH oe oF Itis possible to develop equations of state using the variables and ‘operators introduced above. To be useful in the solution of flow problems, however, the formulated equations should preferably be , Feferred to co-ordinates x fixed in space, so that its necessary to apply certain transformation rules, These were given by Oldroydl 5} The stress tensors x and px ate related by (2.2) with x! now representing any suitable fixed ca-ordinate system (not necessarily Cartesian) and similar tensor relationships can be derived relating other variables. For example, ig is the metric tensor of the fixed co-ordinate system x! related to 7y(E, 0) by ax! act ; ww? 5g ee aa Pix 2. the deformation tensor y(8,) ~ YC, 1 is weated to a fixed tensor Gx by ax! ayt = yl = 57 Ser Gia 2.4 alt Wh 57 aS 24) where Gn es) and xis the position at time of the element that is instantaneously at the point x’ at time ¢. The fixed component equivalent of the operator D/Dr is represented by 9/91, which is given by* ‘The corresponding equations when 0/9¢ operates on a contravariant or mixed tensor are alzo avilable[ 5] Rheological Equations of State 13 ay" int + 5k Eins 2.6) + 2.6) where Ex isa symmetric covariant tensor. ‘The fixed-component counterparts of quantities such as (1/2) ynlG,# DE" evaluated at f= 1, are “rate-oF strain” tensors. AP S209", en ef) being the familiar (first) rate-of-strain tensor of classical fluid dynamics (cf, equation 1.15). We note for future reference that the nih Rivlin-Ericksen[9] tensor is given by AD? 2 24h, 28) 2.2 Application of the theory — the simple fluid of Coleman and Nol In the notation of the last section, the rheological equations of state for the general class of fluids* known as incompressible ‘simple fluids’ can be essentially written in the form{13~16] aul = py B0* wh, a9) y= Engl) — Yul. 91 2.10) in convected co-ordinates, and Pa =P + Pies aay Dials O° F Gals, 629) 2.12) in fixed co-ordinates, where F is a tensor-valued functional. We are considering only incompressible fluids, so that p is an arbitrary isotropic pressure and we can write FO) =0 2.13) without loss of generality. It can easily be deduced that the current time f cannot appear explicitly in the equations of state for the materials under *oldroyd{6) argues that the simple-Nuid model is not completely generale “4 Rheometry ‘consideration and these then essentially reduce to pin = E 1G), @.14) where s == 1" Elastic liquids necessarily have a fading memory in the sense that the deformation history of a material element in the distant past rust be expected to have a weaker influence on the current stress response than the deformation history of the recent past, Coleman and Noll{15] make provision for this by defining a “norm’ which involves an influence function satisfying (a) H6(s) is defined for 05 Se and has positive real values: A((s)> 0. (b) Als) decays to zero according to Eg, sthG) <0 monotonically for large s. ‘A convenient form for the norm is* NG) = [7 1G) PAO" Ias, 2.15) where | G(s) | is the magnitude of the tensor G(s). The norm is clearly designed to give more prominence to the recent past i. small values of s More general norms than (2.15) have been considered[13,18,231, but the results are essentially the same. The norm given by (2.15) ensures that the space of symmetric tensor-valued functions G(s) is a” Hilbert space. In their development of the theory. Coleman and! Nollf 15] make use of the Fréchet derivative and certuin formal continuity requirements to obtain a series of approximations to the functional appearing in (2.14), based on the ‘size’ of the norm (2.15). The first, second and third approximations for an incompressible liquid are sgiven below by equations (2.16), (2.17) and (2.18) respectively. In these equations use has been made of certain simplifying relations given by Pipkin 17] At this stage itis slightly more convenient to work with s matrix notation ‘These ate given fora rectangular Cartesian frame to avoid any confusion at this point concerning covariant and contravariant suffixes, ¢ ple = My (sGia(s0ds, 2.16) Rheological Equations of State 15 [faosrcatons fans snrtenGleari en vin = [Maen tos+ [°F Mats, 8216 (60% (85 de + fiat sa sanGulsadGu(ss als) $+ MalS1 4531 53)Gi51 IGylSs Grass dS, ds dss, (2.18) where, from the symmetry of the stress tensor, we require Mal5,452) = Ma (52,5) Ma(S1,S2y $9 2.19) Ma(53. 52,51). ‘The rheological equations of state (2.11) and (2.16) are called the equations of finite linear viscoelasticity; (2.11) and (2.17) are called the equations of second-order viscoelasticity. and so on. We now consider the conditions under which the norm (2.15) may bbe small, ie, the conditions under which (2.16)-(2.18) are useful approximations to (2.14), 1. Clearly, the norm is small if the deformation has been small throughout the history of the motion, as for example is the ease when a fluid is subjected to an oscillatory motion of small amplitude (cf. Chapter 6), However, itis sufficient only that the deformation should be small in the recent past on account of the influence function. ‘2. With a sufficiently strong influence function, there need be very little restriction on the deformation, except, of course, that the norm must exist, This means that the norm can be small for fairly large deformations in the case of slightly elastic liquids. In their work, Coleman and Nollf 15} lay more emphasis on condition 1, although condition 2 is discussed by Truesdell and Noll{24] (see also Wang{ 18). 3. The norm is small in the case of retarded motions. Coleman and Noll{ 14] consider a given history G(s) and then a further set ofee ——————— 16 Rheometry histories which may be regarded as Yessentially the same’ as G(s), but having been carried out ata slower rate. They introduce a variable af")s where 0 0 forall, arf) isa monotone inressing function of 4, W@) 39, 66) @ leo We know of no corresponding festitions on the second normal eda secures pes moor er He Lh, dncrly accepted by mont ther theoreticians ether te Proportion borne out by experiment, since inspite of some eal Evidence init favourl40.41 |, the overwhchming opinion now that ee rl The vicometrie functions occupy astrtele postin in many situation of technoogia importante the wseodty being specially Basie Rheometrical Concepts 25 important. For example, itis the viscosity which governs the rate at ‘which liquids can be pumped through pipes; it is also the dominant influence in ‘lubrication’ situations. For most elastico-viscous systems, the viscosity is a monotonic decreasing function of the shear rate, falling from a zeroshear ‘first Newtonian’ value to a lower value at higher shear rates (sometimes referred to as the second-Newtonian viscosity). For polymer systems, the second Newtonian value can often be several orders of magnitude lower than the first. Such behaviour is sometimes called ‘temporary viscosity loss’ to distinguish it from ‘permanent viscosity loss’ caused by mechanical degradation. However, it is more usually referred to as ‘shear-thinning’ behaviour. ‘Some non-Newtonian systems, including starch suspensions, exhibit shearthickening behaviour, the viscosity increasing with shear rate. Such systems are, however, relatively rare, The first normal stress difference », is often much larger than the shear stress 7 and can give rise to startling phenomena not found in conventional Newtonian liquids. The Weissenberg rod-climbing effect (Fig. 3.1) and the die-swell effect (Fig. 3.2) are perhaps the best known examples of such phenomena, The first i eevant in mixing problems and the latter to extrusion processes. We have already indicated that the second normal stress difference Fig 3.1 ‘The Weissenbers effect,26 Rheomerry Basic Rheometrical Concepts 27 Fig 3.3. Progressive development of streamlines fora solution of 1% polyacrylamide in 50/50 water glycerol in a square pipe of side 084 cin. Flow rate = 2 ce see". Dye inserted very near entre of tp face of square. The departure (rom rcilineat flow & Fig 3.2 The die-swell effect. an indication of a non-zero » : 7 ™28 Rheometry vy is usually much smaller than the first and in general terms is not regarded as being as important. However, in some situations » has a prominent influence, For example, it decides whether rectilinear flow is possible in flow through pipes of non-circular cross section{ 6,42] (Fig. 3.3). It also plays an important role in hydrodynamic-stability considerations(43]. 3.2 Smal-amplitude oscillatory-shear flow In this section, we consider a small-amplitude oscillatory-shear flow with the velocity distribution given by =O, an where the amplitude e is small enough for sécond and higher powers to be neglected. The corresponding stress distribution for viscolastic, ‘materials can be written in the form* vl =xtew cost, uF =v! . ¢ oe Ga) Pit —Paa = P22 ~ Pas = P13 ™ P23 =O, where n' and G' are even functions of w. ‘nis given the name proportional to the rate-of-strain, so that G and 1! is simply the the stress is proportional to the strain so that in this case n' = O and G’ is the (constant) rigidity modulus. xtewelt, vt =u? =0, G9) where i=/—1 and the ‘real part’ is implied. The corresponding stress distribution is now given by pia =ntewe!, other Pix G.10) As with the corresponding stress dstsibution for steady simple-shear flow, no {Specific theological equation of state is required for this exercise ee ————————————— Basic Rheometrical Concepts 29 where nt @.1) is known as the ‘complex dynamic viscosity’. Equivalent definitions for other functions associated with. smalkamplitude oscillatory-shear flow are to be found in the literature. For example G" = onl 3.12) js known as the “loss modulus’, and Gt=G' +i" @.13) as the ‘complex modulus’. The complex viscosity is sometimes written in the form ate in” @.14) with an obvious relationship existing between n” and G', Since we shall be mainly concerned with liquid systems showing some elastic effects, we shall find it convenient to work with nt, 9 and G” ‘The behaviour of a simple fluid undergoing small amplitude oscillatory-shear flow can be represented by the single integral ‘equation of state (2.16). When the deformation tensor appropriate to the velocity field (3.9) is substituted into (2.16), we obtain a relationship between n* and the kernel function M of the form wn 2 fl mpi eMtnae, 3.15) or, what is equivalent, 1 s , My (8) sin et dB, cn) : 3.16) G'=— [7 My @U ~ cov bye : For the Oldroyd model (2.25), n’ and G’ take the special forms nolL tude] Gy Tow?@r =m) Toa u+en) | G17)30 Rheometry For a simple fluid, the viscometric functions are given by (3.3) for sufficiently slow flow, and it is not difficult to deduce from (2.24), (3.3) and (3.16) that CQ) lq--0 = 1'() Iso» G.18) w@) Gi) a9 ag? [79 at de! G.19) ‘which relate the viscometric functions to those associated with small-amplitude oscillatory-shear flow. ‘When the restriction on the ‘size’ of the amplitude of the ‘oscillatory-shear flow is removed and non-linear effects are considered, the normal stress differences are no longer negligible and higher harmonics have to be considered when calculating the shear stress response fo such a motion. This is considered in detail in Chapter 6. 3.3. Combined steady and oscillatory shear flow Since the mid-1960’s, considerable attention has been paid to the Situation of combined steady and oscillatory shear flow. Such a situation is of interest in ils ows sight as a relatively simple ‘rheometer flow which can be interpreted in terms of certain well-defined material functions. It also provides a much more critical test of any proposed rheological equations of state than steady and/or oscillatory shear separately. ‘The relevant velocity distribution for combined steady and. oscillatory shear flow can be written in the form at tetwc!, 3.20) vw? = Bxtwelt, where € and B are small enough for second and higher powers to be neglected. It is usual for either € or B to be zero. When B= 0, we have ‘parallel superposition’, and when ¢ = 0 we have ‘orthogonal ‘superposition’. In either case, itis possible to interpret the results in terms of a complex viscosity and to investigate the effect of a non-zero q on this function; ie. we have (when B= 0) Piz sense", ean Basle Rheometrical Concepts 31 or (when € = 0) P32 = Bunge", (3.22) where in both cases, n* is to be regarded as a function of q as well For the simple fluid, it ean be shown that (when = 0)(44~46] a Mylo. @ loo 3 lane a (3.23) ClO, D loro and (when €= 0) To.) lo~0 = (4), 3.24) C00, lua 0. | : Itis also of interest to investigate the effect of the oscillatory shear on the mean shear stress, but to do this it is necessary to relax tite restriction on the magnitude of and B, since a restriction to linear powers cannot result in any change in the mean conditions. ‘This is considered in detail in Chapter 6. 3.4 Extensional flow Most rheometer flows can be represented by the general velocity distribution (3.20) (with a suitable modification to accommodate the rheometer geometry). However, there is a growing interest in the study of ‘extensional’ or ‘elongational’ flow. This is usually related to a velocity distribution of the form (3.25) where k is a constant extensional rate of strain. The corresponding stress field can be written as Py ~ P22 Pay P33 =k), Py =O fori #j, (3.26) where ne is the elongational or extensional viscosity. When k is small, we have the familiar “Trouton’ result mele = 3N@) qo G2C 32 Rheometry Attempts have naturally been made to relate ne to the material functions n, »; and v3 (associated with steady simple shear flow) for general values of k and [47]. However, these have proved largely unproductive and there is no a priori reason why such a relationship should exist (ef. Chapter 7), 3.5. The measurement of material functions in various geometries One of the objectives of Rheometry is to measure the material functions defined in this Chapter over as wide a range of conditions as possible. The development of these functions in terms of a simple Cartesian representation of the sort given in Sections 3.13.4 is conceptually appealing and allows the basic results to be expressed in ‘a simple way without unnecessary trimmings. The Cartesian representation does not however lend itself to easy experimentation, Itis therefore convenient to consider simple rheometer flows in other (more suitable) geometries and to attempt to relate the relevant stress distribution to the material functions already defined. ‘This is not a trivial exercise and may be carried out in more than one way. One may, for example, use a ‘base-vector’ treatment[ 8,16] or alternatively, a convected-co-ordinate ‘theological-history" development|6). We shall use the latter, and the reader is referred to other texts for the equivalent base-vector treatment. (i) Rheological history for a steady simple shear flow ‘The displacement functions x", x'?, x"? corresponding to the steady simple shear flow (3.1) are given by* xfhex! -gxta 0), 3.28) We now choose a convected co-ordinate system ¥ which is coincident with the Cartesian system at time, i. xt sgt abe 1), x7 2B, G29) * ‘14 willbe recalled that xis the position at time ¢' ofthe clement that is {nstantaneoualy atthe point x at timer. © Basie Rheometrical Concepts 33 The components of the metric tensor of the &/ co-ordinate system are given by nm ax alt) = Ser oer 3.30) so that, from (3.29) i a1), 0 = |-aa—0), 1e@a—er, of, 31) 0 0, 1 which we define as the ‘deformation history’ forthe steady simple shear flow (3.1), The corresponding (extra) stress history in the &/ co-ordinate system (Eis given by whe Sar gg PO. (3.32) From (3.2), the components of the stress tensor pig(x', ¢') are given oy @INi@)+ N21, ang, 0 Pals’ t= alg), PNA, O}, 3.33) 0, 0 were, for convenience, we ive wed the normaizaion* Py = PUN) + 2N2(Q)1 (3.34) From (3.29),(3.32) and (33), we have FINDEN, Ghq,t—1), 0 ae =| abate), @¥@t-0), 0] 3.35) 0, °. 0 where P= HQ ~ GIN *N OIE, . G36) We Na(q) = 2n(qnt~ F+4? IN CQ) + NCQ = 1? "This is permissible, since we are dealing with incompressible liquids and there is an indeterminacy inthe stress tensor (othe extent of an added wotropiec pressure34 Rheometry We are now in a position to write down other legitimate theological histories which are expressible in terms of the functions a. and Np by simply making an arbitrary change of co-ordinates from & to 8, for if yu(E, ) and m,(E, 1") describe a rheological history in the system, then %(&.') and #(€, 1" also represent a permissible theological history, with FWlE, OY = ahah ya lk, t'), (3.37) HAE Y= ahah nig, 3.38) where the a=ag'/@ &) can be any nine finite numbers with a non-vanishing determinant, each constant so far as the time ¢’ and particle under consideration are concerned] 6] Our main concern at this point is to be able to take a simple velocity distribution in a fixed orthogonal co-ordinate system and deduce the corresponding stress components in that system. The problem is therefore one of determining the deformation history for such a flow and associating this with (3.31) through (3.37) by means of a suitable convected co-ordinate transformation. ie. through suitable af. The stress history is immediately given by (3.35) and (3.38) and the required stress distribution in the fived orthogonal coordinate system follows from the usual tensoF-transformation law. In this chapter, we give no consideration to the important problem of determining whether the proposed flows and resulting stress distribution are compatible with the stress equations of motion. This is left o later chapters where specific rheometer flows are considered in some detail. (ii) General rectilinear flow under a constant pressure ‘gradient | 6,42,48) We apply the above method first to general rectilinear flow under a constant pressure gradient, for which the relevant velocity field, referred to Cartesian co-ordinates x, is given by =O, (3.39) vw oF ‘The corresponding displacement functions are given by xt ex! 00 PME), xt, (3.40) Basic Rheometrical Concepts 38 If we choose a convected co-ordinate system €/ which is coincident with the Cartesian system at time, we have Pe), 7 G41) ‘and the components of the metric tensor of the system are easily shown to be f gt, net) WWE = | gate fa Vega PF, gaqatt~ 1713.42) Hatt qagste FF, Vege 14 where a _ av 25a tapi aa 3.43) aa apt Oa OE It is easily verified that the components of the metric tensor el. #') are related to those of the metric tensor given by (3.31) through (3.37), with 1 0 0, asia. aala |. (3.44) 0, ala, asl Vig tail (3.45) Having determined dit isa straightforward matter to calculate the ‘extra-stress history #)(£. (") From (3.35), (3.38) and (3.44), The components of the extra stress tensor in the x" system at time f” follow by using #)(&,"). (341) and the usual tensor transformation Jaw. It is found that(6) BIND*NEDI, — QomtQ). gala) Pic’ = asm, a3N2@) 4243N2(q)], ang), aaqaNal@. g3N2(a) 3.46) ‘afin the th row and jth column of the array,a { 36 Rheometry ive, the stress components are independent of (3.46) indicates that the extra-stress components at time / are given in terms of the ‘material functions n, N, and Nz, which being functions of q,q3 and qy are to be regarded as functions of x? and x9. We have therefore determined the extra-stress components corresponding to the velocity distribution (3.39) from primitive steady-shear concepts without recourse to any particular theological equation of state, (ili) Torsional flow Consider 2 ‘torsional’ flow with velocity components, referred to a suitably chosen cylindrical polar co-ordinate system (r, 8,2) of the form Hy =O, Hey = PLE), M4) = 0, 47) which automatically satisfies the equation of continuity. The corresponding displacement functions, which we write as 7,8", "are given by = w=), (3.48) We now choose a convected co-ordinate system £/ which is coincident with the cylindrical polar system at time ¢ and we waite re, OF WEE 1), 3.49) vee . ‘The components of the metric tensor in the E! system are easily shown to be Tage), Bq), 0 a= | Bae-2, «@), 0], 50) 0, 0 1 where est) ( ( Basic Rheometrical Concepts ” It is now possible to show that the components of the metric tensor 4n(G, 1 are related to those of the metric tensor ¥4(6, ) through (3.37), with Oh 0) a=11, 0, 0 3.52) 0, 0, =I Following a similar procedure to that of the last section, it can be shown that the components of the extrastress tensor piy(”, ¢’) are given by{6] 0, 0, o Pies’ f= 10, aI (Qt NQ, rang) 3.53) 0, rant. @N3@). ‘The physical components of the tensor are obtained by omitting the rand F factors. We see that torsional flow is equivalent to steady simple shearing flow, the relevant local shear rate being rduo/dz (iv) Flow between rotating cones with a common vertex ‘An important rheometrical flow is that generated between rotating cones, The relevant velocity components referred to a suitably chosen spherical polar co-ordinate system (r, 8, x), are given by 49 20, 048) gy = rsin 0216). (3.4) ‘The corresponding displacement functions are Per 638, x=x-M0OKE-¢), 3.55) We now choose a convected co-or inate system &! to be coincident with the (7,8, x) system at time fund write roB 3.56) xs FE- Fe 1°), {tis easily verified that the components of the metric tensor yl, are given by38 Rheometry @Y sin? B, alk. | EP Best Bu, 0, sey ae Br-1'), 0 ervey sn BSB) o-, of. os 0, 1 “The relevant form for the components of afin this case ate eos q=| 0 Bo (358) 0 0 1 with the appropriate shear rateq given by qrsinE fe G59) The physical components of the extra stress tensor in the fixed spherical polar co-ordinates are given by 0 0 ° Poy =| 0, a? N20Qs ania) 0, gn @PUNs(@) + N(@)) with 4a = sing 3.66 an sind (3.60) We see from (3.60) that flow between rotating cones is equivalent to a steady simple shear flow with local shear rate sin @ dS2/d9. In fact, Oldroydl 6} has shown that the flow caused by the steady rotation of solids of revolution of general shape is equivalent to steady simple shear flow. The situations considered in this and the preceding section are special cases of particular importance in a theometrical context. Basic Rheometrical Concepts C5 (v) Helical flow Consider a ‘helical’ flow with a velocity distribution referred to cylindrical polar co-ordinates of the form dry 20, ey = POL, He) FM, G61) which automatically satisfies the equation of continuity. When = 0, we have a velocity distribution suitable for use in the problem of flow between rotating concentric cylinders, and when co = 0 we have a velocity field appropriate to steady flow through a pipe of circular cross section, ‘The displacement functions corresponding to (3.61) are (2B, =P WEE), 2B WE, (3.62) where, as before, we have taken a F/ system to be coincident with the cylindrical polar system at time ¢. The components of the metric tensor ¥(E, 1") are given by 1 0, -alt-¢) inlE = 0, er, -Pase-0) |. mantt= 0), -Pase— 0), 1a ey (3.63) - where q =V1g3 +a31 and ga = 4v/dE?, gy = E°de/dE?. The approptiate form for the components af is ala. Basia, 0 d=} 0 1 (3.64) gla. Bazia, 0 and the physical components ofthe extra stress tensor are given by @ Nata), aaa)» ana GIN) +1, 42g), aaaotNilad + Na(@)]s Pin aana) 429stNiq) +2) |. G65) GIN) +N SE TNE TS40 Rheometry We see therefore that helical flow is characterized by the three material functions m, Ny and Ny Numerous theoretical analyses of helical flow of varying generality are to be found in the literature, together with a small number of | ‘experimental papers|6,49—53]. In this latter connection, the papers of Schowalter and co-workers[51,52] deserve special mention, since their experiments confirm the prediction that helical flow is equivalent to a steady simple shear flow. In the light of this, one should not expect significant research effort on helical flow in the future, since the viscometric functions can be determined with greater ease from other steady simple shear flows, including two special cases of helical flow, namely, flow between cylinders in relative rotation (Couette flow) and rectilinear flow through a stationary pipe (Poiseuille flow). For the former, the relevant shear rate is rdoofdr and for the latter du/dr. In this book, we shall deal ‘exclusively with these particular cases of helical low. 2.6 The measurement of extensional viscosity in various geometries ‘The method used in the last section to verify that certain rheometer flows are equivalent to a steady simple shear flow can also be applied to the ease of steady extensional flow [54]. The problem is to construct simple flow fields which are equivalent to a steady extensional flow and which are attractive from an experimental point™ of view. ‘The displacement functions corresponding to the velocity jstribution (3.25) can be shown to be xi axle, Pe xteMe“nn, (3.66) xf) = pte-OR, We now choose a convected co-ordinate system &/ to be coincident with the Cartesian system at times f, ie. Peter, . xt epeteon, 3.67) xia peter, Basic Rheometrical Concepts 4 ‘The components of the metric tensor of the &/ co-ordinates are, from (3.30) and (3.67). MeO, 9, ° Piet) =e On ere Oe 3.68) 0, 0, ee, which is the deformation history for the steady extensional flow. From (3.26), we write the stress tensor pi (s",’)in the form Aneto), 0. 0. Pix’, 0 0 0 (3.69) 0 where we have used the normalization meth) 3.70) for convenience. From (3.32), (3.67) and (3.69), we obtain Piy ameter“, 0,0 idk’) 0 0, 0 0 0, 0 67) Which isthe stress history corresponding to the steady extensional flow (3.28). The problem is now to decide whether other meaningful histories 7,’ can be related to yu(§. given by (3.68) through (3.37) and suitable f components. itis easy t0 show that biaxial extensional flow with velocity components 2px 79 (where 8 is the vonstant biaxial rate of strain) is equivalent to steady extensional flow with a corresponding stress field 0, 0, 0 Pik =] 0. 0. 0 . (3.73) 0, 0, *~26ne(-26) a result obtained by Dealy[55] for the simple fluid (2.14),42 Rheometry Consider next the motion of a circular tube undergoing deflation and simultaneous stretching described by a velocity field in cylindrical polar co-ordinates 0, He) 3.74) Hy =z HM, He where 6 isa constant. The appropriate metric tensor can be written ene, °. o aE= |] 0, RMN, 0 75) 6, °. out “This is again easily shown to be simply related to (3.68) and the appropriate stress field is| S41 0, 0, 0 0, 0, 0 (3.76) 0, 0, Bnet) We now consider the inflation of a spherical shell tiking a velocity field (referred to spherical polar co-ordinates) of the form Pla mit) 4) ea) = Hey =O. a7 In the general cise, it does not appear possible to express the corresponding stress field solely in terms of the extensional viscosity ng. However, when attention is confined to the expansion of ruin shells, the flow ean be approximated to a steady extensional flow by ‘an appropriate choice of m(1)156] Finally, we note that the present method can be used to deduce that the following low fields (referred to cylindrical polar co-ordinates) are equivalent to steady extension: 3 Uy FM, 9) FMWD, Yay = ZF (3.78) yy Fm, Ug) =A Yay = IME G.79) where m and @ are constant, Unfortunately. itis not immediately clear how such flows ean be realized in the laboratory. _—$—$5 $< rrr Basic Rheometrical Concepts as {The search for easly realizable flows which are equivalent to simple extension is a subject of current interest and much effort is being extended on the possible development of extensional viscometers for mobile liquid systems, At the time of writing, progress in this direction is slow and difficult (cf. Chapter 7).CHAPTER 4 The Measurement of the Material Functions using a Rheogoniometer 4.1 Introduction ‘The new British Standard Glossary of rheological terms[$7) defines a rheogoniometer as ‘a rheometer designed or the measurement of normal as well as shear components of the stress tensor’. This rather general definition includes within its scope capillary and similar theometers as well as thos¢ of the ‘rotary-type’. However, in practice, the rheogoniometer has become to be associated exclusively with rotary flow situations and very often the rheogoniometer is regarded as being synonymous with the Weissenberg cheogoniometer, a commercial instrument (manufactured by Sangamo Controls, Ltd., Bognor Regis). There have been commercial developments elsewhere, which would discourage this latter practice, but there is no doubt that the rheogoniometer is popularly regarded asa rotary rheometer which normally employs the cone-and-plate geometry, but can be adapted to simulate, for example, torsional flow between parallel plates or Couette flow between co-axial cylinders. In the rheogoniometer, the test fluid is contained between two instrument members, one of which is driven at some prescribed angular velocity, while the other is stationary. Facilities are invariably available to determine the torque on the stationary ‘member. In favourable circumstances, facilities are also available for ‘measuring the ‘total-normal-force’ on, for example, the plate in the ccone-and-plate geometry and for determining the distribution of pressure over one of the instrument surfaces. The commercial Weissenberg rheogoniometer, for example, has torque and total-force 4a Measurement of Material Functions using a Rheogoniometer 45 facilities but the ‘distribution-of-pressure’ measurement is more complex and fraught with problems, as we shall see. In the present chapter, we consider first the basic theory for sheat- and normalstress measurement in rotary situations, giving particular attention to the assumptions involved. The numerous. problems, both theoretical and experimental, real and apparent, which make interpretation of the results difficult are then critically discussed and methods for minimizing their effect are outlined. 4.2 Basic theory for *cone-and-plate’ flow We consider the flow situation described in Fig. 4.1 and a suitably defined spherical polar co-ordinate system (r, 9, x). The test material is contained between the plate (= #/2) and the cone (@ = /2 +89) both of which are assumed at this juncture to be infinite in extent. We assume that a solution to the various equations exists in the form 1%) =, %9) =0, an gq) sin BLA) with the boundary conditions implying Qx12=0 — onthe stationary plate, and : ay From (3.2) and (3.60), the corresponding extra stress distribution Flot ‘Statonary Pate Rototing F ‘cone Fig. 1 Basic coneand-plate geometry,¢ 46 Rheometry js given in terms of the material functions n, v; and v2 by Phony = Pies) = Pirw Plow) =¥2(Ms Poxxy = ¥1(Q) + P2(q) 4.2) Peoxy = 4nl@), where the shear rate q is given by aa sino 4.3) a=sind 3) We have no guarantee at this point that the stress distribution (4.2) is compatible with the stress equations of motion (AI.S~AI.7). ‘These reduce in the present context to ap 2 Pie) PU pr sin?@ 9? FS = ion ~ eM Pon, (4.4) 1 ap ~prsin 8 cosa. 92 => pr sin 8 aa ie cote + ab po (Sit Orie) — —— Pox as) to = (sin? 0 pio. (4.6) = rp Bg sin? Plows 4.6) where we have incorporated any body forces into the isotropic pressure p. ‘The solution of (4.6.) on using (4.2.) is immediately given by gn(q)= A cosec?8, 47) where 4 is a constant fo be determined from the boundary conditions. Eliminating the pressure between (4.4.) andl (4.5.) we obtain 2or sinto 9 2 = 4 ao do Equations (4.7) and (4.8) are in general incompatible. To enable us to retain the velocity distribution (4.1) we make the following assumptions: (Gi) ‘Inertia’ effects are negligible, so that we can set p = 0 in (4.8) or assume that & is small enough for second-order terms to be ignored {rq + (@)). (48) ‘Measurement of Material Functions using a Rheogoniometer 47 (ii) We have a cone-and-plate situation with a very small gap angle 60, small enough, in fact, to enable us to write cosec? @ = 1 in (4.7). ‘We now have, from (4.18) and (4.7), ae 49) i.e. a constant shear rate, implying a constant state of stress from 4.2). In practice, the cone and the plate are of finite extent. We therefore require the further assumptions: [7,16,29,58] (iii) The cone and plate are each of radius a (iv) ‘The free surface of the liquid is part of a sphere of radius a with centre at the cone vertex. (v) The steady simple shear flow represented by (4.1) continues up to the free surface. (vi) Surface-tension forces are negligible. Under these conditions, the boundary condition at the free surface becomes{58} Pot P30) =P» (4.10) where po is the atmospheric pressure, The couple C required to keep the plate stationary is given by c=f" mo ower aay and from (4,2) and (4.9), we have c= an) (4.12) This equation indicates thatthe measurement ofthe torque on the stationary pate asa function of the angular velocity of the cone GGe-of the shear rate q through (4:9) can be immestately used to determine the shearrate dependence of the apparent viscosity Ir pis the presure ata point onthe plate in excess of that due to atmospheric pressure, we have B=—De00) lexan ~ Por (4.13) and from (4.2) and (4.4), it is easily shown that aa (@)+2.@l, (4.14) eS8 Rheometry ox whats equivalent, _ dan) ‘This equation predicts a logarithmic dependence of the pressure pon rand the slope o the @ In?) cure ean be immelatly ised 60 determine the normab-stress function »,(q) + 2 ¥3(Q). Integrating (4.15) and applying the boundary condition = by @ +21. (41s) Pla)= po — Ya (4.16) we have BON =1y, +20, )In* — vy, (47) ‘The total normal force F on the plate is given by Fe ff amp an, (aus) and performing the integration, we obtain a (4.19) ‘This equation implies that measurements of the total normal force as 4 function of rotational speed &, can be used to determine the variation of the first normal stress difference », with shear rate g In principle, (4.15) and (4.19) can be used to determine vj and v2 * separately. A further check on v2 is provided by (4.16), although the assumptions (iv) to (vi) are more crucial here than in, for example, the employment of (4.15), where the slope of the (p, In r) curve can be taken at points far enough away from the edge, for edge effects to bbe unimportant. The status of (4.19) as regards edge effects is somewhat between that of (4.15) and (4.16) and will be considered in detail later. 4.3 Basic theory for torsional flow For torsional flow, we consider the arrangement shown schematically . in Fig. 4.2. Referred to suitably chosen cylindrical polar co-ordinates (7,8, 2), the plate at z = 0 is stationary while that at z = h rotates O) Measurement of Material Functions using @ Rheogoniometer 49 fs Spotnay aa Pate “Sere Pate Fig. 42 Basic torsionalflow geometry. about the z axis with angular velocity 2 . We assume a velocity distribution of the form B20, Ye) = POLE). My =O, (420) where, on account of the boundary conditions, we require 10) =0, w(h) = 25 Gey (3.2) and (3.53) give the extra-stress distribution corresponding to (4.20) in terms of the material functions m, »; and v;. We have 0, 0, 6. Play =] 0, ms(a)+¥3(@). ania]. 4.22) 0 ang at@) Where the shear rate q is given by (4.23) Uf we incorporate any body fores into the isotropic pressure p, the stress equations of motion corresponding fo the velocity distribution (4.20) are from (A1.1)—(A1.3) pu Lae, (00 — Blo «38 1 2 ig) HED, a2 (4.25)o 6 Rheometry (4.26) ‘Substituting (4.20) and (4.22) into (4.24)- (4.26), we have a wen) prt = 22 ith), 427) arr a 0-2 nan, (428) a 0-2 +? ox@ (4.29) From (4.21), (4.23) and (4.28), we require ou (430) ot gee so that the shear rate q depends on r but not on = Eliminating the pressure between (4.27) and (4.29) and using (4.30), we obtain a further equation for «2, namely do 2pre ere a 0. 431) This is incompatible with (4.30) unless we neglect inertia terms (i. unless we set p = 0 or asstume that «1 small enough for product terms to be neglected). Here, we do not require any geometrical condition for compatability as in the case of cone-and-plate flow. If pis the pressure on the stationary plate over und above that arising from the atmospheric pressure po, i B= Pas) lea ~ Pos (432) we have, from (4.22) and (4.27), @” &; Sted Ex} gee ate (4.33) This equation does not in itself yield 4 normal stress function directly unless ¥, is arbitrarily set equal to zero[59]. This is in contrast to the situation in cone-and:-plate flow, where the corresponding equation (4.15) can be used to give the normal stress function », + 209. For convenience, we now assign a sulfix 1 to the cone-and-plate ( Measurement of Material Functions using a Rheogoniometer $1 situation discussed in section 4.2 and a suffix 2 to the torsional-flow situation, From (4.15) and (4.33), we havel $8) Ws 4.34 a (4.34) or va (435) pf oar where the term in brackets is to be regarded as a function of q’. From (418) and (4.38) we ae tht pres distribution hnemurmtents ake i the eoneant-pats and plateand pate Inorder to determine the (otal normal force ating on one of the plates, we mae the folowing asuimptons| 36] {The Ire urfce spurt of the eyinder r= a, where asthe Ci) The state of low given by (4.20) exists up to the fee suriace, ese can nelet edge effet ‘The relevant boundary condition atthe fre surface is now given by for O<:2&h, (436) Integrating (4.33) and applying the boundary condition (4.36) we obtain Pry = Po on 2 [oy tye) po arn, 437) where », and v; are to be regarded as functions of 7. If Fis the total normal force on the stationary plate over and ubove that arising from atmospheric pressure, we have =f, ter pte) ew ar. (4.38) Substituting (4.37) into (4.38) and reversing the order of integration, in the double integral, we obtain nf) 7 fe PON rardt — 2n Ff matrr, (4.39)( 52 Rheometry and writing q, for the maximum shear rate, ie Qa as, (4.40) equation (4.39) can be reduced to the more convenient form whe ae aa 4 (q) ~ v2(q) dg. ay Jy, 1) rsa de. (441) Substituting for ©, in terms of gg and differentiating, we have finally 59,60] 2F [1 dink aealire wa@ | *2ding. ‘This equation taken together with (4.19) provides a method for determining », and v2, separately, from total force measurements in two geometries. The torque C required to keep the plate stationary is given by cman [" anayae aan) 1 — 2) lew (4.42) Using (4.40), this can be written in the alternative form fo" ania) da. (aay (449) or, what is equivalent, 44) = t one] (4.46) pace i+ 2a? dy 3 ding. ‘This equation can be used to calculate the apparent viscosity function from measurements of the torque Cas a function of the maximum shear rate gq. In both cone-and-plate flow and torsional flow, the velocity field is independent of the material properties (see equations 4.9 and 4.30). Pipkin{61} defines such flows as ‘controllable’. C Measurement of Material Functions using a Rheogoniometer 53. 4.4 Basic theory for Couette flow 16,62,63] Couette flow supplies another possible method of determining the apparent viscosity 7 and the first normal stress difference », . We consider the situation shown schematically in Fig. 4.3. The test fluid is contained between coaxial cylinders (assumed infinite in extent) ‘which are in relative rotation. The inner-cylinder radius is denoted by and the outer-ylinder radius by ra. Referred to suitably defined cylindrical polar co-ordinates, the appropriate velocity distribution is taken as , ty = re(H), ty =O, (47) which implies the absence of ‘end effects’. From (3.2) and (3.65), the the corresponding extra-stress distribution is given in terms of the material functions m,», and v3 by vq antqs 0 pay = | ang (a+r), Of. (4.48) 0, 0, ° where the relevant shear rate q is given by (449) or On account of (4.47) and (4.48), the stress equations of motion (AL1)-(AL3) reduce t0 on, 708 =v ()— pret, 4.50) Me (4.50) [TT as I TTT, rir Fig 4.3 Basic Couette-flow ecometry54 Rheometry pee Pa (7 P¢0)) = 0. (431) If we neglect inertial effects, (4.50) can be integrated to give a Porro) Penritd= far, (4.52) and if we now write B for the pressure on the cylinder walls end 1 for the shear stress rye), we have from (4.51) and (4.52) Lore ode Hoopes [nine 433) with 5, and ro having the obvious meaning. Writing Ap for Blra) ~ Bly) und = In| (4.53) becomes Loa apes, tore, asa where, from (4.51), 8 = 2 In (ro/n). (4.55) Inverting 4.54), we nave 740) — 94 (0, ~ 6) = AP (4.56) 20; This equation can be alternatively expressed in the form woz [202 with the hope that the series converges rapidly enough for only a few terms in the series to be needed in the computation. (4.57) implies that the first normal stress difference », can be calculated if (84/80; lq, is determined empirically for n = 0, 1, 2,3, Experimentally, a large annular gap (ie. (ry ~ rity large) is usuelly required to produce a measurable pressure difference Ap in this flow situation. The problem of determining the viscosity function n (q) from torque measurements in Couette flow is 2 non-trivial exercise except in the ease when the annular gap is very small. From (4.51), the shear ~ ‘Measurement of Material Functions using a Rheogoniometer 55 stress Dg is given by Ay Poo) = : (4.58) where A, is a constant of integration. If Cis the couple (per unit, length of cylinder) on the cylinder of radius r, we have C= 2ar*p.0), (4.59) so that C is independent of r on account of (4.58) and we write for eC Posy () = 100 (4.60) We now assume that the shear stress/shear rate relationship has an inverse given by a= Xe. (4.61) (4.49) can then be integrated to give A= 22-2, =f (4.62) where &, and 2; are the rotational speeds of the inner and outer cylinders, respectively. A change of the integration variable yields anal [2 a aa oe cle LES ae 4.64 : nr? ar ead ‘The problem is now one of inverting (4.63) to find Ms) and hence, by implication, the shear stress/shear rate relationship. We follow the method given by Coleman, Markovitz and Nolll 161 * and Gifferentiate (4.63) to obtain (4.65) "Other more sophisticated methods are given by Kreiger and co-workers 68]. aSC 56 Rheometry Writing B= 2/73 and Yisy)= 2COAN/AC, we have V6) = AG.) ABH). (4.66) Since 6 <1, this equation can be written in the alternative form Me= 5, vigts aon so that Xs, ) can be calculated from measurements of the couple on the inner eylinder and 9 ~ 9. The above method is only rally applicable when @ is relatively small (ie. when the gap is wide). Under these conditions, end effects may be nor-negltible, and the use of this method can only be justified when a wide-sap Couette rheometer only is available. There are clearly more satisfactory means of measuring the viscosity function using other geometties oF considering 3 narrow-gap Couette rheometer. tn the latter ease, the shear rate q can be effectively taken as a vonstant throughout the liquid, given by nan enh (4.68) ‘The shear stress + is directly related to the torque C through (4.60) (with r replaced by 7, oF rg) and the apparent viscosity 1 (q) follows from r= gn(q). ‘The approximation involved in assuming & constant shear stress throughout the liquid is easily assessed from (4.60). For example, when n/t is 0.98, there is approximately a 4%% variation in shear stress and when 1/79 is 0.99 this drops to 2%. Whether such an approximation can be tolerated is bound up to a large extent with such matters as the precision of the measurements and the particular application in mind. Tn conclusion, we note that, in contrast to cone-and-plate flow and torsional flow, the velocity field in Couette flow is dependent on the ‘material properties of the test fluid, the apparent viscosity function being required in its determination. Pipkin defines such a flow as “partially controllable'| 61}. q 4.5 Basic equations for other rheogoniometer flows. Rheogoniometers designed to investigate cone-and-plate or parallel plate flow can be (and have been)(63,65~671 easily adapted to ic Measurement of Material Functions using a Rheogoniometer $7 Fev Flot Statenary Fig 44 Uxtended coneand:plate geometry study the flow situation described in Fig. 4.4. The angle 9, and the ratio Ji/a are assumed to be small. Clearly torsional flow is a special ‘case given by 0, = Oand vonc-and-plate flow is given by h = 0. In analysing this ‘extended cone-and-plate flow’ the following assumptions are made: (i) inertial effects may be neglected. Gi) the boundary condition at the interface ean be satisfied without the introduction of secondary flows, and an appropriate bounclary condition is introduced which includes (4.10) and (4.36) as special cases. With reference to eylindrical polar co-ordinates (r, 8,2), itis assumed that conical surfaces z = k{h +r 69], 0 0. This would Indicate that total force measurements in cone-anc-plate flow and torsional flow might be expected to lead to a better estimate of rp than extended cone-and-plate flow, although the need for one ‘geometry in the generalized flow as against two in the alternative Measurement of Material Functions using a Rheogoniometer $9 situation may be a compensatory factor in favour of extended. cone-and-plate flow. The above observations led Cowsley167} to suggest the use of a reentrant cone (i 8y <0) in extended cone and plate flow. In this case. the fuctor h/t +40, } can be made very large and total force measurements can be made very sensitive to 1 We conclude this section by remarking that numerous authors hhave suggested alternative methods of determining the normal stress functions by suitable modifications of the cone-and-plate and parallel-plate situations|63, 68, 69]. Such methods are novel and interesting and may be important in isolated circumstanes However, itis very unlikely that any of them will superse methods we have described in Sections 4.2, 4.3 and 4.5, 4.6 Possible sources of error ‘The basis equations derived in Sections 4.24.5 for the determination of the material functions n,», and v3 rely on the existence of a simple type of flow throughout the liquid and ‘numerous assumptions are made to ensure the existence of such a flow. An important consideration is therefore the extent to which these assumptions are valid in practical situations and the limitations, if any, they impose on the operating conditions. In the following sections, we critically review the more important assumptions made in the development of the theory and also other possible sources of error which are likely to affect the interpretation of the experimental results For convenience, we consider the possible sources of error individually and isolate the main cause of concern in each case. For example, in examining the ‘constant shear rate” assumption in cone-and:plate flow, we concentrate on the effect of the gap angle on the shear rate and leave any consideration of other influences such as or inertial effects to the appropriate sections dealing with these general problems. This isolation of individual factors is necessary, since the general problem incorporating and coupling all possible sources of error would be extremely difficult (if not intractable at the present time) and any solution, if it were forthcoming, would have little predictive value in the present context. ‘The eeentrant cone geometry might be expected to lead to more severe ‘edge effvet” problems than the other geometries we have considered in this chapter,¢ 60 Rheometry 46.1 The shear rate in the cone-and:plate rheometer ‘The usual interpretation of experimental results in the conventional cone-and-plate situation assumes the existence of a constant shear rate throughout the liquid and by implication a constant state of stress. The validity of this assumption has been considered by Adams and Lodge{$8] and by Walters and Waters{70], amongst others. It ig clearly dependent on the gap angle and the point at issue is whether those used in practical situations are small enough for the constant shear rate assumption to be a reasonable one. In most cases, a gp angle of less than 4° is used (with angles in the region of 2° being the ‘most popular), although there have been cases of cone angles as high as 10° being preferred for special reasons(71] TABLE 4.1 Gap angle Variation of shear Error in wring degrees} rateacross gop formula ‘) 7 0a 2 021 3 0.28 4 os 5 077 > Vs wo an ‘The error involved in assuming a constant shear rate can be cestimated in various ways. For example, it is possible to solve the low problem exactly fora Newtonian liquid in the case of negligible uid inertia and the computed shear rate can then be compared with the approximate expression q =, /@. Alternatively, an approximate error analysis can be based on (4.7). Any reliable method should lead to the same order of magnitude estimates of the error and in the present context this i all that is required. Table 4.1 (after Adams and Lodge|$8]) gives the relevant information for gap angles up to 10°. It will be seen that the errors involved for gap angles employed in normal operation of the rheogoniometer can be tolerated since they are well within experimental error. The assumption of a constant shear rate is therefore a reasonable one, () Measurement of Material Functions using a Rheogoniometer 61 6.2 Inertial effects {In Sections (4.2) and (4.3), we saw that the steady simple shearing lows proposed for cone-and-plate flow and torsional flow are not compatible with the governing equations when inertial effects are non-negligible, The relevant compatibility conditions (4.8) and (4.31) point to the need to invoke ‘secondary flows’ to satisfy all the equations involved. Even when inertial effects are negligible, secondary flows are to be expected in the cone-and-plate situation unless »; + 2v, is zero or the gap angle is very small. No comparable restriction on the material functions o the geometry is necessary in torsional flow, a fact which influenced Ginn and Metzner{59] to prefer the plate-and-plate geometry to the corresponding cone-and-plate situation. In this section we consider secondary flows of inertial origin and we attempt to estimate their effects on the interpretation of experin ‘The secondary flows predicted for the (infinite) cone-and-plate geometty are of the type illustrated in Figs. 4.5 and 4.6 (the gap angle is exaggerated for illustration purposes). In the viscous case, the direction of the streamlines is outwards near the rotating cone and inwards near the stationary plate. In the elastico-viscous case, the streamlines are divided into two regions. In the outer region, the streamlines behave similarly to those in the viscous ease, but the inner region contains closed streamlines, the direction of which is in the opposite sense to those in the outer region (ie. inwards near the ‘cone and outwards near the plate). General experimental confirmation of the theoretical predictions have been supplied by ‘numerous workers{ 70.72.73]. However, the experiments of Savins sand Metzner{74] indicate the presence of a series of vortices, the origin of which may be linked with edge effects in the experimental configuration (see Section 4.6.3) tal results Viscous quid Fig 4.5 Sccondaty flows for a Newtonian liquid.o Rheometry Eloste tque Fig. 4.6 Secondary flows for an elastic Hiquid Secondary flows give rise to ‘secondary pressures’ and ‘secondary stresses’ which may be significant when fluid inertia is important as in the case when liquids of relatively low viscosity are in high-speed rotation. Existing mathematical analyses of inertial effects are essentially perturbation problems in a non-dimensional parameter M sven by 2 pa? M . (4.74) a where a isa typical length and 1 is an appropriate viscosity — usually the limiting viscosity at small rates of shear. For mathematical convenience, the analyses are usually restricted to small values of M. At this point it is important to discuss the status of theoretical analyses based on the infinite cone-and-plate situation, These can and have been useful in assessing when fluid inertia effects are important and in estimating their likely orders of magnitude, However, it would bbe misleading to seek quantitative significance in the predictions, except in isolated circumstances. For example, the appearance of secondary flows leads to the theoretical prediction of an increase in the torque on one of the instrument members and a decrease on the other, whereas one would expect (and indeed observe) an increase in torque on both members in the real situation. The discrepancy between the theoretical predictions and the experimental observations in this respect is resolved when the role of the free surface in the experiments and the absence of such a surface in the infinite cone-and-plate theory is taken into account, Predictions of @ quantitative nature might be expected in a study of the inertial pressures generated by the steady simple shear ‘primary flow’ itself, these being essentially independent of the Measurement of Material Functions using a Rheogoniometer 63 presenee and form of any secondary flow. To discuss this, we replace {G18 forthe dstrbuton ot pressure on the pain the coneandplate situation by ip [ were inthe ease when fd inertia and secondary Rows are o-y, (4.15) @=0, (4.76) Y= rg) + 22()1 We regard 4 as an inertial correction which is independent of uid rheology, to a first approximation at least. It arises as a result of the term on the left hand side of (4.8) which depends on the density p and the primary flow angular velocity $2(@), but does not depend directly on the material functions A satisfactory mathematical analysis gives & as{70] (4.77) (4.78) which was used by Adams and Lodge[58] and results from an ‘adaptation of an approximate theory originally proposed by Groensmith and Rivlin{75] for torsional flow (see Fig. 4.7). Higher-order corrections to (4,77) can also be computed and these ‘might be required when fluid inertia is a dominant influence and the parameter M is very large (see, for example, Waters and King{761).iIn such extensions to the theory, the velocity (0) is now found to depend on the mechanical properties of the liquid and in consequence so does the inertial correction & on account of (4.4), The correction term & gives rise to what is commonly referred to as the ‘negative normal-stress effect’ We have next to consider the way that W differs from the form given in (4.76) on account of secondary flows. Here, existing theoretical analyses can only be a guide due to the complexity of the full problem, but the indications are that W does not differ significantly from (4.76)(70]. This means that so far as inertial64. Rheometry Fig. 4.7. Pressures attributed to centrifugal forces, btained by Adame and Ledge [SA Dimethylphthalate. Continuous line based on 4.78, Broken line based on 4.77. (After Walters and Waters (ro). effects are concemed, we can replace (4.14) by _ 30 "a 10 = leq) +2 @) (4.79) with a reasonable degree of confidence (see also Olabisi and Williams[77}), The corresponding equation for the total force place of (4.19)) is 3mpQat | x pt (4.80) ‘The negative-normal-stress terms given in (4.79) and (4.80) can also be determined experimentally by measuring p and F for a Newtonian liquid of the same density us the test Muid under the same experimental conditions. ‘The arguments given above for cone-and-plate flow may be applied ‘even more strongly in torsional flow to justify the following, ‘Measurement of Material Functions using a Rheogoniometer 65 ‘modifications of (4.33) and (4.41). =o, ta, sn aq aH a0 + oy Sp lana da (4.82) We have already indicated that, in assessing secondary flow effects, theoretical torque predictions are hindered by “free surface’ considerations in addition to other intrinsic difficulties in the problem. However, theory can be used with success to indicate when secondary flow effects are likely to be important (in the case of Newtonian liquids at least) and also their likely order of magnitude Fig. 4.8 contains experimental torque results obtained by Chengl 78} Tor supposedly Newtonian liquids. Cy is the observed torque and Co the torque expected on the basis of Newtonian behaviour and no 1ry-flow effects. The apparent ‘shear-thickening’ behaviour is attributable to sevondary-flow efleets, a point of view substantiated by the theoretical predictions which are seen to be in qualitative agreement with the experimental results|76}. Fig. 4.8 points to the need for care when using the cone-and-plate geometry (or plate-and- plate geometry for that matter) for viscosity measurements in the cease of very mobile liquid systems at high shear-rates. Vem fff Sauron" Le THEORETICAL CURVE, os a oem aa ee oe 7 eel)? Fig 4.8 Comparison of experimental data of Chengl 78] with the theory of Walters and Waters 70]. B isa dimensionless function ofthe gap angle which has Becn tabulated by Cheng.¢ 66 Rheometry In Couette flow, no inertial correction is required in torque measurements provided the laminar flow regime implied by (4.47) is not affected by instabilities of the Taylor-vortex type or by turbulence, The necessary inertial correction in pressure ‘measurements is given by (cf. 4.53) Bere) — BU) rot dr (4.83) ‘The inertial correction depends on eo(r) which in tuen depends on the apparent viscosity function (see Section 4.4), IL is therefore necessary to determine this function before the inertial correction ccan be made. 4.6.3 Edge and end effects For convenience we consider edge and end effects under three main headings, (a) Effects arising from the fact that the instrument members have “finite’ dimensions, (b) Effects arising from the shape of the free surface and related surface-tension problem. (c) ‘Fracturing’ effects. Under (2), if we confine attention in the ‘ist instance to the cone-and-plate configuration, we are interested in the distortion to the flow field near the edge of the instrument members (which necessarily have finite dimensions) from that predicted for the infinite cone-and-plate situation. To isolute this type of edge effect from that mainly associated with the shape o surface tension considerations, we may conveniently limit attention to the problem of a cone of finite dimensions in a sea of liquid as shown schematically in Fig. 4.9, the gap ungle is again magnified for illustration purposes and the geometry in the ‘sea of liquid!” region is chosen for convenience in making theoretical predictions, There is no doubt that some distortion of the flow field near the cone rim will ‘occur, but the problem is to decide how violent this distortion is and its sphere of influence in the range 0 << a, a being the radius of the cone. Tanner[79] comments that, intuitively, one would expect edge effects to penetrate a distance /, where fis the distance between the cone and the plate at the rim. Linked with this is the expectation that any edge effect correction will depend on the gap angle in the cone-and-plate device or the ratio ha in the plate-and-plate the free surfage and ‘Measurement of Material Functions using a Rheogoniometer 67 be LL Fig. 49. Geometry of cone In SeaoFliquid configuration geometry. For reasons we have made clear in earlier seetions, the gap angle in cone-and-plate flow is purposely kept small (< 4°), one of the few exceptions being the 8° gap favoured by Meissner 71], whose choive was governed by a desire to isolate material and instrument effects in transient experiments on high-viscosity polymer systems, Our main concern in this section is with the small gap-angle Rheologists have carried out theoretical and experimental studies of edge effects in the “sea-of-liquid’” configuration, Theoretical analyses of varying complexity have been carried out by Slattery{80], Tanner 79], Greensmith and Rivlin| 75] and Griffiths and Walters{72]. Tanner has even proposed a method for determining the second normal-stress difference on the basis of an analysis of edge effects in the cone-und-plate situation. Here, we concentrate on the theoretical predictions of Griffiths and Walters{72.81] for Newtonian and second-order fluils, Edge effects may be expected to make their presence felt more in torque ‘measurements than in pressure distribution of total force experiments, a fuct borne out by the theoretical results. Fig. 4.10 contains percentage errors in torque measurements due to edge effects in the parallel-plate system for a Newtonian liquid. Errors for the cone-and-plate situation can be estimated by identifying h/a with the gap angle. (In Fig. 4.10, C, is the torque with edge effects accounted for and Cy is the torque calculated on the basis of no. aberration to the flow field near the edge). Griffiths and Walters conclude that the’ error for a gap angle of 4 is less than 2% and for a cone angle of 2° it is less than 1%. These estimates might be expected to apply approximately for non-Newtonian liquids also. They might68. Rheometry -6,Veq100% o To tae Fig. 4.10 The percentage dference between Gqland Cy ofa funstion af hla. (ter Gri lthe sha Walterst721) also be expected to provide the upper bounds for the errors in pressure distribution and total force measurements. In fact, edge effects should not be a major problem in pressure-distribution measurements, since the slope of the (P, In r) curve can be taken at + distances far enough away from the rim for edge effects to be irrelevant. The situation concerning total-force measurements is more complicated and needs careful consideration, Of particular importance, isthe effect of the flow conditions in the sea of liquid beyond the cone rim. A change in flow conditions here can produce ‘a change in the pressure p on the plate. The inerease or decrease in B is predicted to be the same at all points of the plate ie. itis independent of r{ 72]. This would not affect the pressure gradient on the plate, but it would affect the resulting total force. These theoretical findings are consistent with the experimental results of Kaye, Lodge and Valel 82). Fig. 4.11 contains (p, ln r) results for one shear rate in the cone-and-plate apparatus using different gp angles (with the cone rotating in a sea of liquid). Altering the gap angle changes the flow conditions in the sea of liquid with a corresponding ‘Measurement of Material Functions using a Rheogoniometer 69 “1s osm en) lz tem Fig 4.11 Pressure distribution at one shear rite (27-4 "+ n'a cone-and-plate apparatus wsing different gap angles with sone fotating i a sea of liquid. Cone radius = 4.4 cm, (After Kaye, Lodge and Valet 821.) influence on the absolute values of p. The gradient of the (, In 7) curve is unichanged, so that the pressure-distribution method of measuring normal stress differences is not affected, Measurements of the rim pressure in these circumstances are of little use and total force measurements can also be significantly affected as Olabisi and Williams{77] and Miller( 83] have observed. ‘When the sea of liquid situation is replaced by the alternative configuration with the sample held in the gap by surface tension, total-Force meusurements showing 0 change on altering the gap angle in the cone-and-plate configuration or the separation fifa in torsional flow (but keeping the shear rate fixed) can be taken as implying the absence of any significant edge effects due to the finite dimensions of the instrument members| 591 Finally, we note that the presence of the edge can significantly affect the secondary-flow patterns in the ease of elastic liquids and a70 Rheometry Fig 412 Theoretical streainline projestions when 30" fora) a Newionaa Uiguid, 1b) lightly elastic liquid 4 moderately sstic Inguid (After Grtfths and Watters 721) strong ‘reversal’ vortex is predicted to form near the edge in cone-and-plate and torsional flow (see Figs. 4.12 und 4,13). tn fact the presence of an edge can affect the secondary-flow field throughout the liquid [81 Turning now to (b) ie. effects arising from the shape of the free surface and related surface tension problems. there is a sense in which it is difficult to isolate such effects from those discussed above. However, here we attempt to isolate spurious pressure problems which are directly caused by changes in contact angle ‘during start up or shut down of shear. Pressure-distribution measurements are again not our immediate concem, since itis difficult to see how surface tension effects can effect the gradient of the (B, la 7) curve when this is taken far from the edge. [tis rather total normal force measurements which are potentially subject to error. It must also be borne in mind that it is | ; Measurement of Material Functions using a Rhieogoniometer <7 Fig. 4.13. Streamlines observed when the cone is rotated ina 13% aqucous solution of polyacrylamide Rotational speed i revmin, (After Griffiths and Walter 721) only the forces generated by cranges in contact angle during start up + or shut down which are relevant in the present context, since any forces due to surface tension which may be present in the rest situation ean be exsily accommodated in the interpretation of the results by simply subtracting it out In their detailed studies of possible aberrations to normal stress measurements, Ginn and Metzner|59} tested several Newtonian oils before concluding that surface tension effects were negligible in their experiments. Kaye, Lodge and Vale[821 on the other hand were more specifie and made total normal force measurements with different free surface shapes before concluding that the resulting changes were within experimental error (<6). Berry and Batchelor!84] reached very much the same conclusion and state that the thrust data are not dependent on the configuration of the boundary. This effect was checked directly in the paralle-plate system by varying the gap between the plates so that the profile of the boundary was concave, flat (ideal) and convex in tum, The differences in the measured thrusts were no larger than those =2 Rheometry observed in replicate experiments under nominally standard ideal conditions’. fi It would be wrong to draw general conclusions from the detailed and precise experiments discussed above, since any free surface shape/surface tension correction must depend to some degree on the surface-tension characteristics of the test fluid and the instrument members and to 2 large extent on. the relative total force levels arising from true normabstress effects. Most of the test materials considered by the investigators cited above were purposely chosen to be highly elastic and it is not surprising therefore that surface tension effects were insignificant when compared to the total forces generated by large normalstress effects. At the other end of the spectrum, the more recent work of Hutton! 83] isa salutary warning not to underestimate surfuce tension cffects when slightly elastic liquids are being investigated. He gives the excess pressure p, inside the liquid in torsional flow due to surfac non [282] a where T, is the surface tension and g, is the contact angle of the liquid (measured through the liquid), This pressure multiplied by the plate area is the surface tension contribution to the total normal thrust tending to separate the plates. Inu series of detailed experiments on a Weissenberg theogoniometer, Hutton investigated the effect of p, for Newtonian * liquids and dilute polymer solutions. He found changes in contact tangle for water at start up and shut down to be of the order of 15° — 20° and the associated change in total thrust was in the region of 250mN. Forces as small as 20mN are detectable in a Weissenberg rheogoniometer, so that surface tension effects are far from being insignificant and must clearly be accommodated if reliable normal stress information is required for slightly elastic liquids. Hutton also ‘warns that there are circumstances when contact angle and surface tension changes with shear may be indistinguishable from true normal forces, Fortunately, approximate considerations of the sort embodied in (4.84) cun at lewst be used to assess whether surface tension effects are likely to be important in a given situation. ‘The third type of edge effect to be considered in this section is concerned with ‘shear-fracture’ instabilities, In a vense, this could be ——— ( Measurement of Material Functions using a Rheogoniometer 73 included as a separate ‘possible source of error’, but it appears to be bound up with conditions at the periphery of the sample to a sufficient extent to consider it under the general heading of ‘edge effects’.* Most of the work in this connection has been carried out by Hutton[87]. He observed the occurrence of an instability in cone-and:-plate flow when a certain critical stress is reached, The instability manifests itself as a rapid fall in stress and the higher the rite of shear the shorter the time before the fall occurs. When the stress falls o a low value, it oscillates considerably. The effect is reversible provided the sample does not gerate and is not expelled from the shearing gap. Hutton concludes that the breakdown is localized to a conical fracture surface which starts at the periphery of the sample and grows inwards (Fig. 4.14). The torque falls because as the fracture grows so the effective sheating radius is decreased. Some recent experiments of Gleissle[88] indicate that such an instability may be the cause of the ‘stress-overshoot’ phenomenon which is occupying the attention of many rheologists at the present UUme, Clearly, we have liere Une possible explanation of the overshoyt phenomenon, but it would be premature to make the fracture instability the sole cause of all stress-overshoot effects Hutton{87] associates the instability under consideration with “fracture in shear’. Now it has always been acknowledged that solids Fig. 4.14 Representation of sheat facture in a parallel plate heometer. ‘Joseph 6| has revently completed a theoretical study of the free surfac problem in torsional flow in which he dixeuses the relevance ofthe theoretical predictions to shear fracturea | Rheometry can facture in both tension and shea, but until recently twas though hat guts could rature in iension (giving rie to “Cvtation’ effet) but notin shear, Hulton experiments sem to ,_—— phonomenon inked withthe shstcity oft tigld. Attempts fave alt been made to relat the shear facture phenomenon ‘mel rate he phenomenon of extrudate dsvorton i epiary fw lation proposed an energy criterion forthe ascurence of shear feature and obtained the following formula for the eel Tat normal stress difference », ay" (4.85) where nis an empirical factor. Such a formula appears to be generally consistent with available experimental results. The fracture phenomenon is present in torsional flow between parallel plates as well as in cone-and-plate flow and this is taken by Hutton as evidence that the phenomenon is not of secondary-flow origin. Some further justification for including shear fracture in the prevent section is provided by the observation that the critical conditions for the onset of the instability are affected by any excess liquid at the periphery of the shearing zone. When this is removed, the critical stress decreases and the delay period before the stress falls is also reduced, At the same time, the amount of exeess liquid has no direct influence on the stable shear-stress levels as one would have anticipated. ‘We end this sub section on a practical note. The experimenter is often required to determine the apparent viscosity of elastic liquids, over as wide a runge of shear rates as possible. From the foregoing. considerations, itis clear that a compromise has to be reached between a low value of a0, to avoid shear fracture (see equation 4.85) and a high value of ag to achieve accuracy of measurement{87] ‘Thus far in this section, we have concentrated on edge effects in the cone-and-plate and parallel-plate situations, We must now give some consideration to ‘end effects’ in Couette flow which arise because the cylinders are necessarily not infinite as required by the theory. Such effects may be expected to be especially severe in the : Measurement of Materlal Functions using a Rheogoniometer ‘75 case of the wide-gap instruments required to make normal-stress measurements (ef. Section 4.4). Indeed, there is evidence of the existence of secondary flows in the wide gap Couette apparatus which cannot be associated with ‘Taylor-instability’ vortices and are apparently generated by end effects([89]. Such secondary flows are confined to elastic liquids. Various theoretical attempts have been made to assess end effects in Couette flow and a number of experimenters have gone to great lengths to overcome them by means of guard rings and other devices. The best known of these methods is due to Mooney and Ewart{90], who designed the eylinder ends as cones, the angles of which were calculated to produce approximately the same shear rate bet: conical surfaces as between the cylindrical surfaces, In the case of elastic liquids, the presence of a free surface is especially troublesome in Couette flow on account of Weissenberg rod-climbing effects and this provides a timely warning not (0 carry over enc-effects analyses for Newtonian liquids to the elastico-viscous cease without due care In addition to the secondary flows already discussed, the combination of the presence of a cylinder edge and an elastico-viscous liquid can give rise to severe aberrations to the velocity and pressure fields near the edge which are not present to anything like the same degree in the case of Newtonian liquids. All these factors point to the need to reduce the sphere of influence of end effects by making the zap between the cylinders as narrow as possible.* In the narrow-gap situation, experimenters take consistency of results for a number of difference lengths of column of fluid as implying the absence of significant end effects. 4.6.4 Hole-pressure errors In Sections 4.2 ~ 4.4, we saw that pressure distribution measurements, provide a means for determining the two normalstress differences in a steady simple shear flow. In principle, this method requires a knowledge of the pressure at a series of points along one of the ‘We have already seen that there are other reasons for preferring «narrow gap in the Couette apparatus, when this Is used for viscosity measurements, Fortunately, Wwhen a wide gap is necessary in order to determine the fist normal siress ‘iffeence, the required point pressures on the cylinder surfaces canbe taken at ‘sufficient distance fom the ends ofthe cylinders to substantially reduce any effects, Broadhent and Lodgel62] conclude from their experiments that Secondary flows are of neuigible importance when this safeguard is observed = - -16 Rheometry instrument surfaces. In practice, itis of course necessary to consider surfaces of small but finite area surrounding the various points to ‘make the pressure measurements. Since the pressure is in general varying from point to point, itis important that the area of the pressure-sensing surface should be as small as possible to enable one to associate meaningfully a pressure with a given point somewhere in the surface. This position will not nevessarly be at the eentre of the pressure-sensing area, since the pressure varies logarithmically with radius in cone-and-plate flow, for example. However, we are here more concerned with emphesizing the need for small pressure-sensing areas than with the problem of associating the measured pressure with some point, which can be handled without too much difficulty Until recently, pressure distribution measurements were always carried out by drilling small holes in the surface over which the test liquid is flowing; the holes were connected either to a manometer oF toa pressure transducer located at the bottom of a suitable cavity containing the test liquid. Manometers have the disadvantage that excessive time is often required for equilibrium to be reached and the latter method ius been favoured by most experimenters. The argument used is that the required pressure jis the same as the hhydrostatie pressure p, measured at the bottom of the eavity by & suitable pressure transducer. The self consistency in the experiments of Markovite{911 on a number of instrument geometries seemed to confirm this belief. However, a series of very detailed and careful experiments carried out by Lodge and his co-workers|62.82,92] led them to the conclusion that there was a large systematic error arising from the use of holes’ to measure pressure. They required a ‘hole pressure’ py, given by Pa = Dem Bs (4.86) (vith py, negative and about one fifth of the First normal stress difference) to bring about consistency in their measurements in various geometries. The discovery of the hole-pressure error effect has led toa substantial theoretical and experimental programme of research and a reasonably consistent picture is emerging Theoretical work{93~95] has been concerned with deep two-dimensional slots normal to or aligned to the flow as wel asthe circular-hole situation. The sit situations are easier to handle in ‘mathematical analyses, but even here simplifying assumptions have Measurement of Material Functions using a Rheogoniometer 77 to be made to render the mathematics tractable. Existing analyses are either confined to the situation of slow flow of Newtonian and second-order fluids or rest crucially on the flow patterns being symmetric about the centreline of the hole, The presence of the slits or the hole must disturb the primary shear flow in their vicinity and a potential source of error is therefore located. Tanner and Pipkin{93] argue that this error in Newtonian flow must be an inertial effect and that the error can be made negligible by reducing the hole diameter until the Reynolds number based on itis sufficiently small (this isa further reason for making the diameter of the holes as small as possible). Experimental results are on the whole consistent with the arguments of Tanner and Pipkin, but Olabisi and Williams{77] have reported measurable pressure-hole errors for Newtonian as well as non-Newtonian liquids. Based on the assumption that the flow pattems are symmetric about the centreline of the hole or the slits, Higashitani and Pritchard{95] obtained simple expressions for the hole pressure error for a number of situations using the assumption that the normal stress differences and the shear stress are approximately related by* Pui Pas Pha 4.87) Par Pas © PP. (4.88) where n and m are constants, For the second-order uid, n= m For flow past a deep twondimensional slot placed normal to the flow direction, Higashitani and Pritchard{95] obtain the expression so that the hole pressure is negative if, as is usually observed, the first normal stress difference is positive. In (4.89), this difference has to be evaluated at the shear rate relevant to the undisturbed shear flow. If mis set equal to 2, the earlier result of Tanner and Pipkin{93} for a second-order Mud is retrieved. ‘When the slot is aligned to the flow, the formula for py is now ‘The first ofthese relationships i resonably well borne out by experiments on ‘number of non-Newtonian systems with n taking vale near 2. The validity of the second relation ir more dificult to assess on account af the lack of precise ‘measurements of the econd normal stress difference.