Lecture 2: Constitutive Relations: E. J. Hinch
Lecture 2: Constitutive Relations: E. J. Hinch
Lecture 2: Constitutive Relations: E. J. Hinch
E. J. Hinch
1 Introduction
This lecture discusses equations of motion for non-Newtonian fluids. Any fluid must satisfy
conservation of momentum
Du
ρ = −∇p + ∇ · σ + ρg (1)
Dt
where ρ is the density of the fluid, u is the velocity field, p is the pressure and σ is the
deviatoric stress tensor (the trace-free component of the stress). 1 We can absorb the body
force ρg into a modified pressure, and in turn we can absorb the modified pressure into the
stress giving ρ Du
Dt = ∇ · σ. Much of the modeling in non-Newtonian fluids concentrates on
finding a constitutive relation between σ and the flow velocity distribution.
The fluids we use are incompressible unless stated otherwise, so we have assumed
∇ · u = 0.
2 Phenomenological
2.1 Simple materials
In a simple material the stress σ depends on the deformation and the rate of deformation.
To understand this relationship we begin by considering how the fluid deforms. Using a
Lagrangian fluid description we follow a fluid particle in the flow. The flow u maps a
material element to a new position x that depends on its initial position X
X → x(X, t).
If we follow a material line element, δX, it is stretched and rotated in the flow according to
∂xi
δX → δx = A · δX, AiJ = .
∂X J
1
In these notes σ is used for either the stress or just the deviatoric stress. It is usually obvious from the
context.
20
We assume that the system is local and causal. That is, the stress at a material point
depends only on the history of that material point, and the stress cannot depend on future
time. This gives a functional for stress
We now adopt the assumption of material frame indifference which states that our
constitutive equation should not depend on the translation, rotation or acceleration of the
frame of reference. Except in extreme cases this should be a good approximation. Thus,
we should get the same result if we calculate our stress before or after rotating the frame
of reference.
Consider a change of frame of reference given by
with Q(t) a rotation matrix and a a translation vector. The stress in the new frame of
reference is given by
21
2.2 Time derivative problem
In a new reference frame the stress is given by
σ 0 = QσQT
and so
σ̇ 0 = Qσ̇QT + Q̇σQT + Qσ Q̇T .
Thus the transformation from σ̇ to σ̇ 0 does not follow the same relation as the transformation
from σ to σ 0 . We will try to find some other derivative that does.
The new flow velocity is
u0 = Qu + Q̇x + ȧ
and the velocity gradient is
∂u0 ∂u
0
= Q QT + Q̇QT . (7)
∂x ∂x
The velocity gradient can be separated into symmetric (strain rate) and antisymmetric
(vorticity) parts The transformed strain rate is E0 = QEQT and the transformed vorticity
is Ω0 = QΩQT + Q̇QT .
Putting these elements together we can show that the co-rotational (Jaumann [1, 2])
time derivative
◦ Dσ
σ≡ −Ω·σ+σ·Ω (8)
Dt
has transformation 0◦ ◦
σ 0 = Q σ QT ,
◦0
(where σ 0 denotes the co-roational derivative of the stress in the new frame of reference,)
as does the co-deformational (Oldroyd [3] or upper convected) derivative
5 Dσ
σ≡ − ∇uT · σ − σ · ∇u. (9)
Dt
◦ 5
Note I = 0 but I 6= 0.
The co-rotational time derivative is the rate of change as observed while rotating and
translating with the fluid. The co-deformational derivative is the rate of change as observed
while deforming and translating with the fluid.
3 Exact approximations
3.1 Linear viscoelasticity
Linear viscoelasticity is valid in the limit where AT A ≈ I. The most general form of the
history-dependent linear constitutive law is
Z ∞
·
σ(t) = R(t) G(s) AT A (t − s)ds RT (t). (10)
0
22
This is a co-rotational time integral. G(s) represents the elastic memory. It is the Fourier
transform of the frequency dependent elastic modulus G∗ (w) defined in section 3.3 of Lecture
1. For a Newtonian fluid G(s) = δ(s) and for an elastic solid, G(s) = 1.
For simple shear with shear rate γ̇
Z ∞
σ(t) = G(s)γ̇(t − s)ds (11)
0
R∞
and since for steady shear γ̇ is constant, the steady shear viscosity is given by 0 G(s)ds.
4 Semi-empirical models
Many fluids are too non-linear to be described by the linear viscoelastic or slightly non-
linear second order models discussed above. For these fluids there are no exact solutions or
exact approximations and other models must be considered.
23
• Power Law [4]
µ(γ̇) = k γ̇ n−1 , (15)
where k and n are fit parameters.
where µ0 and µ∞ are the viscosities at the limits of zero and infinite shear rate,
respectively; a, n, and λ are fit parameters.
• Yield fluids: the fluid flows only above some critical stress σy .
– Bingham [8]
∞ if |σ| < σy
µ= σy
µ0 + if |σ| > σy
γ̇
– Herschel Bulkley
∞ if |σ| < σy
µ= σy
k σ̇ n−1 + if |σ| > σy
γ̇
5 5
σ + λ1 σ = 2µ(E + λ2 E), (17)
where λ1 is the relaxation time and λ2 is the retardation time. For a given pressure p, the
Oldroyd-B model often appears in an equivalent form for the total stress:
G
σ = −p∗ I + 2µ∗ E + A (18)
τ
5 1
A = − (A − I) (19)
τ
where p∗ = p + 2(1 − λ2 /λ1 )µ/λ1 , G = 2(1 − λ2 /λ1 )µ, τ = λ1 and µ∗ = λ2 µ/λ1 . The
Oldroyd-B model reduces to Upper Convective Maxwell (UCM) when λ2 = 0 and viscous
Newtonian when λ2 = λ1 .
In simple shear an Oldroyd-B fluid has constant viscosity µ = G/2 + µ∗ and the normal
stresses are
N1 = 2µ(λ1 − λ2 )γ̇ 2 , N2 ≡ 0. (20)
The uniaxial extensional viscosity is
24
10
−2
−4
−6
−8
−10
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Elongation rate: ˙
This gives negative viscosities at some elongation rates (figure 1) which is unphysical. This
happens because the Oldroyd-B model is derived using Hooks Law springs which are in-
finitely extensible.
The Oldroyd-B model can be reformulated to eliminate the negative viscosity. Assuming
that the microstructure is not infinitely extensible, we get
5 f
A+ (A − I) = 0 (22)
τ
for some function f (A). The stress σ is then
Gf
τ = −pI + 2µs E + (A − I). (23)
τ
Occasionally f appears only in Equation (22) and not in (23).
The FENE (finitely extensible nonlinearly elastic) modification keeps A from growing
too fast by setting
L2
f= 2 , (24)
L − trace A
where L represents a nondimensional length scale for the stretching of the microstructure.
The more A stretches, the stiffer it becomes.
25
tion of shear rate. The deviatoric stress is given by
µ(γ̇) 5
σ+ σ = 2µ(γ̇)E. (25)
G
The shear thinning viscosity µ(γ̇) often follows a power-law.
• The Giesekus model [10] adds quadratic nonlinearity and divides the deviatoric stress
into a solvent contribution (σs ) and a polymer contribution (σp ).
σ = σ s + σp
σs = µ s E
5 αλ1 2
σp + λ 1 σ p + σ = 2µp E
µp p
• The PTT (Phan-Thien-Tanner) model [11] is similar to Giesekus but has a different
nonlinear term
σ = σ s + σp
σs = µ s E
5 λ1
σp + λ1 σ p + exp trace σp − 1 σp = 2µp E
µp
The Wagner model [13] is a special case of the K-BKZ model with
∂w ∂w √
= 0 (N2 = 0) and = e−k α−3+θ(β−α) .
∂β ∂α
• One can choose σp to be the sum of several components each of which has its own
relaxation time. This allows us to introduce multiple relaxation times into most of
the above models.
26
References
[1] G. Jaumann, Grundlagen der Bewegungslehre (Springer, Leipzig, 1905).
[2] S. Zaremba, “Correlation of dynamic and steady flow viscosities,” Bull. Acad. Sci.
Cracovie 594 (1903).
[3] J. G. Oldroyd, “On the formulation of rheological equations of state,” Proc. Roy. Soc.
London 200, 523 (1950).
[4] de Waele, “Viscometry and plastometry,” Oil Color Chem. Assoc. J. 6, 33 (1923).
[5] P. Carreau, “Rheological equations from molecular network theories,” Trans. Soc.
Rheol 16, 99 (1972).
[7] M. Cross, “Rheology of non-newtonian fluids: a new flow equation for pseudo-plastic
systems,” J. Colloid 20, 417 (1958).
[9] J. L. White and A. Metzner, “Rheological equations from molecular network theories,”
J. Appl. Polym. Sci. 7, 1867 (1963).
[10] H. Giesekus, “A simple constitutive equation for polymer fluids based on the concept
of deformation dependent tensorial mobility,” J. Non-Newtonian Fluid Mech. 11, 69
(1982).
[11] N. Phan-Thien and R. I. Tanner, “A new constitutive equation derived from network
theory,” J. Non-Newtonian Fluid Mech. 2, 353 (1977).
[13] M. H. Wagner, “Network theory of polymer melts,” Rheol. Acta 18, 33 (1979).
27