Numerical Methods for Fluid-Structure Interaction

Problems

Thomas Richter
[email protected]

Heidelberg

July 23, 2010

(in process)
Contents

Bibliography 4

1 Introduction 7
1.1 Fluid-Structure Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Fundamentals of Continuum Mechanics 13

2.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.1 The strain tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.2 The rate-of-strain tensor . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.1 The Conservation Equations in Fluid Dynamics . . . . . . . . . . . . . 19
2.2.2 Material Laws in Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . 22
2.2.3 Incompressible Flows, the Navier-Stokes Equations . . . . . . . . . . . 23
2.2.4 Variational formulation of the incompressible Navier-Stokes Equations 24
2.3 Structure Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.1 The Conservation Equations in Structure Dynamics . . . . . . . . . . 25
2.3.2 The Piola Transformation . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3.3 Material Laws in Structure Dynamics . . . . . . . . . . . . . . . . . . 29
2.3.4 Linear Models of Elasticity . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.5 Variational formulation of the elastic structure equations . . . . . . . 32

3 The Fluid-Structure Interaction Problem 33

3.1 Fluid Flows in ALE Formulation . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Construction of the ALE-mapping T̂f . . . . . . . . . . . . . . . . . . . . . . 38
3.3 Fluid-structure interaction in ALE coordinates . . . . . . . . . . . . . . . . . 41

4 The Finite Element Method for Continuum Mechanics 43

4.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.1.1 Finite Elements for elliptic partial differential equations . . . . . . . . 44
4.1.2 Discretization of time-depending partial differential equations . . . . . 45
4.1.3 Discretization of nonlinear problems . . . . . . . . . . . . . . . . . . . 46
4.1.4 Discretization of transport-dominant problems . . . . . . . . . . . . . 48
4.2 Finite Elements for the incompressible Navier-Stokes Equations . . . . . . . . 49
4.2.1 Finite Element discretization of the Stokes Equations . . . . . . . . . 49
4.2.2 Finite Element discretization of the Navier-Stokes Equations . . . . . 51
4.2.3 Finite Elements for Navier-Stokes in ALE-formulation . . . . . . . . . 53
4.3 Finite Elements for elastic structure equations . . . . . . . . . . . . . . . . . . 55
4.4 Finite Elements for the fluid-structure interaction problem . . . . . . . . . . . 56

3
Contents

4.5 Some mathematical theory for fluid-structure interaction problems (noch ein
wenig experimentell und mit Vorsicht zu geniessen) . . . . . . . . . . . . . . . 57
4.6 Solution of the monolithic FSI formulation . . . . . . . . . . . . . . . . . . . . 62

5 Partitioned Approaches 65
5.1 Coupling of the subproblems . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.2 Weakly coupled approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3 Strongly coupled partitioned approaches . . . . . . . . . . . . . . . . . . . . . 71
5.4 Acceleration schemes for partitioned fsi-solvers . . . . . . . . . . . . . . . . . 72
5.4.1 Aitken relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.4.2 Steepest descent relaxation . . . . . . . . . . . . . . . . . . . . . . . . 74
5.5 Stability analysis and the added mass effect . . . . . . . . . . . . . . . . . . . 75

6 Alternative Approaches for FSI-problems 81

6.1 Fully Eulerian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.2 A Fixed Mesh Euler-Lagrange Approach . . . . . . . . . . . . . . . . . . . . . 86
6.2.1 The Flow Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.2.2 The Interface Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2.3 The Structure Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2.4 The coupled problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

4
Bibliography

[Ran08] R. Rannacher. Numerische Methoden für Probleme der Kontinuumsmechanik.

Universität Heidelberg, http://numerik.uni-hd.de/˜lehre/notes/, 2008. Vor-
lesungsskriptum.

[Joh87] C. Johnson: Numerical Solution of Partial Differential Equations by the Finite

Element Method. Cambridge University Press, 1987

[RST08] H.-G. Roos, M. Stynes, L. Tobiska. Robust Numerical Methods for Singularly Per-
turbed Differential Equations. Springer Series in Computational Mathematics 24,
Second Edition, Springer Verlag 2008.

[Gal1] G.P. Galdi. An Introduction to the Mathematical Theory of the Navier-Stokes Equa-
tions. Volume I, Linearised Steady Problems. Springer, 1994.

[Gal2] G.P. Galdi. An Introduction to the Mathematical Theory of the Navier-Stokes Equa-
tions. Volume II, Nonlinear Steady Problems. Springer, 1994.

[GR86] V. Girault and P.-A. Raviart. Finite Elements for the Navier Stokes Equations.
Springer, Berlin, 1986.

[SG97] H. Schlichting and K. Gersten. Grenzschicht-Theorie. Springer Verlag, 1997.

[Tem84] R. Temam. Navier-Stokes Equations. North-Holland, 1984.

Grundlagen der Funktionalanalysis

[Alt08] H.W. Alt. Lineare Funktionalanalysis. Eine anwendungsorientierte Einführung.

Springer Verlag, Berlin, 5. Auflage, 2008.

[Wer07] D. Werner. Funktionalanalysis. Springer, Berlin, 6. Auflange, 2007.

Methods for Fluid Structure Interaction

[GeWa08a] A. Gerstenberger, W.A. Wall. Enhancement of fixed-grid methods towars com-

plex fluid-structure interaction applications. International Journal for Numerical
Methods in Fluids, 57(9), pp 1227-1248 (2008)

[GeWa08b] A. Gerstenberger, W.A. Wall. An extended finite element method / Lagrange

multiplier based approach for fluid-structure interaction. Computer Methods in
Applied MEchanics and Engineering, 197(19-20), pp 1699-1714 (2008)

5
Bibliography

[GeWa10] A. Gerstenberger, W.A. Wall. An embedded Dirichlet formulation for 3d continua.

International Journal for Numerical Methods in Engineering, 82 (5), pp 537-563
(2010)

6
1 Introduction

Fluid-structure interaction problems describe the coupled dynamics of fluid mechanics and
structure mechanics. They are classical multi-physics problems. Examples for fluid-structure
interaction exist in many engineering applications:

aerodynamics The simulation of the aerodynamical properties of planes is the classical com-
putational fluid dynamics (CFD) problem. To estimate the forces, drag- or lift-values,
of a given configuration, simulations are inevitable. In particular for new and large
constructions, like the Airbus 380, experimental data from wind tunnel tests are not
available.

For an accurate simulation however, the deformation of the plane under the aerody-
namical forces has to be taken into account, since this deformation alters the shape of
the plane and thus changes the aerodynamical behavior.

bio-engineering For a better understanding of the flow of blood (hemodynamics) in vessels

or in the heart, simulations are the only way to obtain data. The walls of the blood
vessels are highly elastic and deformed under the pulsating blood flow. Further, the
flow itself is driven by the contraction of the heart muscles, that is, by the deformation
of the structure surrounding the flow domain.

The described problems have in common, that the evolving flow acts on the surface of the
structure leading to its deformation. This deformation changes the flow domain itself. We
are looking at two-way coupled systems, where each part influences the other.

Definition 1 (Coupled Systems). A coupled system S is one in which physically or compu-

tationally heterogeneous mechanical components interact dynamically.[?, ?, ?, ?, ?, ?]. By
S 1 and S 2 we denote the subsystems. The coupled system S is called one-way, if there is no
feedback between the subsystems and two-way, if there is feedback between the subsystems.

The fluid-structure interaction examples described above are two-way coupled systems. An
example for a one-way coupled system would be the interaction of a walking human with the
surrounding air: The movement of the human will give rise to an airflow around it, this slow
airflow will however not impose significant forces on the human. The analysis of coupled
systems and especially the numerical methods will depend on the type of coupling between
the subsystems.

The concept of coupled systems can be generalized to multi-coupled systems with subsystems
S 1 , . . . , S N . This can be multi-structure-fluid interaction, fluid-structure-fluid-interaction
(e.g. water-boat-air). We then call a fully coupled system multi-way coupled.

7
1 Introduction

In most problems it cannot be decided if the problem is one-way or two-way. Regarding the
interaction of a very slow driving car with the surrounding air, the influence of the air on
the car can be neglected. At a certain speed however, the aerodynamic resistance plays an
important role.

1.1 Fluid-Structure Interaction

For the following, we introduce a prototypical fluid-structure interaction problem: Figure 1.1
shows a flow domain with an obstacle. We call the common domain Ω, the flow domain Ωf
and the structure domain Ωs :

Ωf

Ωs

Figure 1.1: Fluid-structure interaction domain.

Now assume, that the fluid domain Ωf is filled with air, and Ωs is a rigid moving body of
steel. This movement will set the fluid into motion. The air however will not significantly
act on the obstacle:

Ωf Ωf
Ωf

Ωs Ωs Ωs

Figure 1.2: Fluid motion imposed by moving structure.

This problem is a one-way fluid-structure interaction problem. The movement of the struc-
ture controls the motion of the fluid but the fluid’s motion does not impair the movement
of the structure.

Next assume, that the flow is driven by an inflow condition and the obstacle is an elas-
tic structure. The evolving flow will act on the surface of the structure and will cause a
deformation. This deformation changes the flow domain:

8
 1.1 Fluid-Structure Interaction

Ωf Ωf
Ωf

Ωs Ωs Ωs

Figure 1.3: Fluid-structure interaction.

Due to the deformation of the obstacle, the flow domain is altered. Here, there is a real
feedback between both subsystems and the coupling is two-way.

Both coupled problems in Figure 1.2 and 1.3 have in common, that the problems are formu-
lated on moving domains. Here, the common domain Ω keeps the same, but the subdomains
of the fluid Ωf and the solid Ωs problem change with time: Ω = Ωf (t) ∪ Ωs (t). This is one
of the main difficulties connected with the modeling of fluid-structure interaction problems
as well the design of numerical methods for their solution.

The different degree in coupling and interaction is important for the treatment of the prob-
lems. We will consider all fluid-structure interaction problems as time-dependent problems
S (t). The solution is approximated at time-steps t1 , t2 , . . . :

S( t n ) S(t n+1 )

Figure 1.4: Time-approximation of the coupled problem.

In every time-step tn → tn+1 both problems need to be solved. The solution S n+1 depends
on the state of both subproblems at time tn as well as on the interaction between both
subproblems. The straightforward way for simulation the coupled problem is the monolithic
approach: we simultaneously solve the fluid and the structure problem at the same time:

S(
f tn) S(t
f n+1 )

S(
s tn) S(t
s n+1 )

Figure 1.5: Monolithic solution of the coupled system.

9
1 Introduction

For this approach, we need to formulate both subproblems, fluid and structure, as one
combined problem. Sometimes however the coupling between both problems suggests a
staggering of the solutions. Considering the problem described in Figure 1.2, the flow field
has no influence on the body, which is moved by some external mechanism. Here, it may be
advisable to first solve for the new shape of the structure and flow domain and then for the
flow field:

S(
f tn) S(t
f n+1 )

S(
s tn) S(t
s n+1 )

Figure 1.6: Staggered solution of the coupled system.

This configuration can now be treated with standard methods: the structure deformation
is computable with a structure solver, the flow problem can be computed separately with
a fluid dynamics code. Unfortunately, interesting fluid-structure interaction problems are
mostly two-way coupled.

Still, it is possible to numerically decouple the interaction problem. These methods are
called partitioned approaches:

S(
f tn) S(t
f n+1 )

S(
s tn) S(t
s n+1 )

Figure 1.7: Partitioned solution of the coupled system.

In every time-step tn → tn+s both problems are solved separately. The flow problem S f at
time tn+1 depends on the flow and on the structure problem at time tn , but the interaction
at time tn+1 is not taken into account for. For the structure problem the same approach
is used. In terms of time-stepping methods this approach can be called an semi-explicit
approach: while the fluid and structure dynamics itself is considered in an implicit fashion,
the interaction between both problems is included in an explicit way giving rise to stability
problems and asking for small time steps. An advantage of the partitioned approach is that
different solvers can be used for the different subproblems. The coupling between fluid and

10
 1.1 Fluid-Structure Interaction

structure comes into the problem by means of boundary conditions on the interface between
both subdomains. Aside, this decoupling allows for a parallel solution of the fluid and the
structure problem. However, in most applications the coupling between both problems is to
strong for partitioned approaches.
A further development of the partitioned approaches are the strongly coupled partitioned
approaches. The two subproblems are solved independently in a decoupled way. Every time-
step tn → tn+1 is however iterated yielding approximate solutions S (tn+1 )if and S (tn+1 )is .
To compute the i-th iterative, the solution at time tn and the last approximations S (tn+1 )i−1
f
and S (tn+1 )i−1
s are considered:

S(
f tn) S(t
f n+1 )

S(
s tn) S(t
s n+1 )

Figure 1.8: Strongly coupled partitioned approach for the coupled system.

The strongly coupled partitioned approach still allows for the use of different solvers for the
fluid and structure dynamics subsystem. Due to the outer iteration, stability problems are
damped, even though not removed. If a very large number of sub-iterations is necessary to
find the stable state, the strongly coupled partitioned approach can be less efficient than the
monolithic solution of the coupled problem.

11
1 Introduction

12
2 Fundamentals of Continuum Mechanics

In this chapter we develop the fundamental equations in continuum mechanics.

2.1 Kinematics

V̂ V (t)
û
x̂
x

Figure 2.1: Deformation of a material particle x̂ in a material body V̂ .

Let V̂ be a material body in the reference configuration. Then, let x̂ ∈ V̂ be a material particle
in the material body V̂ . After some time t, the particle x̂ is moved to a new position x. By
û = x − x̂ we denote the deformation of the particle x̂. We assume, that the deformation is
a continuous (and perhaps differentiable) function defining the displacement of all particles
x̂ ∈ V̂ :
û = x − x̂ ⇒ x = x(x̂, t) = x̂ + û(x̂, t). (2.1)
Regarding the displacement of all particles x̂ ∈ V̂ , the reference body V̂ is deformed to V (t),
see Figure 2.1. By x(x̂, t) ∈ V (t) for t ≥ 0 we denote the trajectory of the particle x̂. Physical
evidence tells us that at time t, no two different particles x̂ ∈ V̂ and x̂0 ∈ V̂ will be at the
same position position x ∈ V , thus the relation (2.1) is invertible and we define on V (t)

u = x − x̂ ⇒ x̂ = x̂(x, t) = x − u(x, t). (2.2)

By summing up (2.1) and (2.2) we get

û(x̂, t) = u(x, t).

If we follow the trajectory of a particle x(x̂, t) we can define its velocity v̂(x̂, t) by
ˆ x̂ û + ∂t û(x̂, t) = ∂t û(x̂, t).
v̂(x̂, t) = dt x(x̂, t) = ∂t x̂ · ∇

We distinguish two coordinate frameworks when observing moving bodies and particles. The
Lagrangian framework viewpoint is material-centered: the focus follows a specify particle x̂.
Deformation û(x̂, t) and velocity v̂(x̂, t) denote the deformation and velocity of the particle

13
2 Fundamentals of Continuum Mechanics

x̂ at time t. The Eulerian framework is space-centered: the focus is on a fixed point x ∈ V (t)
in space. Deformation u(x, t) and velocity v(x, t) denote the deformation and velocity of the
particle x̂ = x(x̂, t) which is in location x at time t.
The deformation u(x, t) in the Eulerian framework is defined by (2.2), the velocity v(x, t) in
the Eulerian system is the velocity of the particle x̂ which is at location x at time t:
v(x, t) := ∂t x(x̂, t) = ∂t (x̂ + û(x̂, t)) = ∂t u(x, t).

The Lagrangian framework is always described on the reference system V̂ . Throughout this
work, variables in Lagrangian coordinates are always marked by a “hat”. The Lagrangian
framework is the natural way of describing problems of structure dynamics, where the de-
formation of a material body V̂ into a new equilibrium V is modeled. In the Eulerian
framework, variables and properties are described on the actual deformed (and possibly
moving) body V . Trajectories of single particles are not followed, instead, the point of focus
is on a fixed location in space. Variables in Eulerian coordinates are always given as Roman
letters without further marking.

V̂ V (t)
û

δ x̂ δx

Figure 2.2: Deformation of differential line-segment δ x̂ from the reference system V̂ to δx in

the deformed system V .

Kinematics is the study of motion. Motion includes the movement (velocity and accelera-
tion) of a particle (or body) and also the change of the position relative to the reference
configuration. Apart from the deformation of single particles, for structural mechanics we
need to observe the change in the relation between particles. Let x̂ and x̂δ be two material
points with
x̂δ = x̂ + δ x̂.
By x ∈ V and xδ ∈ V we denote the deformed points given by a differentiable deformation
field û in V̂ via
x = x̂ + û(x̂), ˆ
xδ = x̂δ + û(x̂δ ) = x̂ + δ x̂ + û(x̂) + ∇û(x̂)δ x̂ + O(|δx|2 ).
See Figure 2.2. Thus, we calculate the deformed line-segment δx in first order by
ˆ
δx := xδ − x ≈ δ x̂ + ∇û(x̂)δ ˆ
x̂ = (I + ∇û(x̂)) δ x̂.
| {z }
=:F̂ (x̂)

The second-order tensor F̂ is called the deformation gradient and is one of the measures for
the deformation. We will use F̂ to describe the change of bodies, in terms of volume, strain,
strain-rates. F̂ is a function of x̂ in the reference framework. By
F := F̂ −1 = (I − ∇u),

14
 2.1 Kinematics

we denote the inverse deformation gradient:

δ x̂ = F̂ −1 δx = F δx.

To distinguish these two tensors, we also call F̂ the Lagrangian deformation gradient and F
the Eulerian deformation gradient.

2.1.1 The strain tensor

In continuum mechanics the strain tensor Ê is one of the key quantities. It is used to
describe the relative length-change of a differential line-segment δ x̂ under deformation. Let
kêk = 1 be a unit vector and with δ x̂ := sê for s > 0 we define the two material points x̂
and x̂δ = x̂ + δ x̂ = x̂ + sê. For, the deformed points x and xδ , by

kx − xδ k − kx̂ − x̂δ k
dê (x̂) := lim ,
s→0 kx̂ − x̂δ k

we denote the relative length-change in direction ê for the material point x̂. With the
ˆ we obtain for small s → 0
deformation gradient F̂ = I + ∇û

kF̂ êk − kêk
1

ˆ + ∇û
ˆ T + ∇û
ˆ T ∇û}ê,
ˆ

1
dê (x̂) = = kF̂ êk − 1 = F̂ ê, F̂ ê 2 − 1 = ê + {∇û ê 2 − 1
kêk

We define the strain tensor Ê(x̂) by

1 ˆ ˆ T ˆ T ˆ 1
Ê(x̂) = {∇û(x̂) + ∇û(x̂) + ∇û(x̂) ∇û(x̂)} = {F̂ T F̂ − I},
2 2
to obtain for every direction ê

1
dê (x̂) = 1 + 2(Ê(x̂)ê, ê) 2
− 1.

We call F̂ T F̂ the right Cauchy Green tensor and Ê = 12 (F̂ T F̂ − I) the Green strain tensor
or Lagrangian strain tensor.
ˆ is small, the quadratic term ∇û
If the gradient of the deformation ∇û ˆ T ∇û
ˆ is negligible and
we define the linear strain tensor ˆ by
1 ˆ ˆ T

ˆ := ∇û(x̂) + ∇û(x̂) .
2

For the Cartesian unit vectors êi , the principal elongations in direction of the coordinate
axes are

1 1
dˆ(i) = 1 + 2(Êii êi , êi ) 2 − 1 = (1 + 2Êii ) 2 − 1.
ˆ
For small displacements k∇ûk  1 this is in first order

dˆ(i) = Êii + O(Êii2 )

15
2 Fundamentals of Continuum Mechanics

Examples Let û be a homogenous translation û(x̂, t) = û0 . Then, the strain tensor is zero
Ê = 0. The length of line-segment keeps the same.
Next, let û be a scaling with a scaling parameter s ∈ R+
!
ˆ = s−1 0
û(x̂, t) = sx̂ − x̂, ∇û
0 s−1

Strain tensor and linear strain tensor (in two dimensions) are given by
! !
s−1


(s − 1) 1 + 2 0 s−1 0
Ê = s−1

, ˆ =
0 (s − 1) 1 + 2 0 s−1

For small scalings (s close to 1), the linear tensor is a good approximation. In direction ê,
the relative length-change is
1 1 1
dˆê = 1 + 2(ê, ê) 2 − 1 = 1 + 2(s − 1) 2 − 1 = (2s − 1) 2 − 1 = s − 1 + O(|s − 1|2 )

For large deformations, e.g. s = 1/10 or s = 10 the linear approximation is not accurate.
For the full tensor Ê we have
 1
s−1
 
2
dˆê = 1 + 2(s − 1) 1 + −1=s−1
2

Next, let û be a shearing:

! ! !
sŷ 1 0 s 1 0 s
û(x̂) = , Ê = , ˆ = ,
0 2 s s2 2 s 0

ˆ
with a shearing parameter s ∈ R. Again, for small values of s the gradient k∇ûk  1 is very
small and the linear tensor is a good approximation. In the general case, with a unit vector
ê = (ê1 , ê2 ), kek = 1 we get
 1
dˆê = 1 + 2sê1 ê2 + s2 ê22
2
−1
√
For the Cartesian unit-vectors ê = (1, 0) and ê = (0, 1) we have dˆ(1,0) = 1 − 1 = 0 and
√
dˆ(0,1) = 1 + s2 − 1:

û û

p
1 + s2

Figure 2.3: Deformation of a line-segment under a shearing.

16
 2.1 Kinematics

The shearing can also be used to explain the off-diagonal entries in the strain tensor. Let
ê1 and ê2 be two unit-vectors in a right angle, e.g. (ê1 , ê2 ) = 0. Assume, that the shearing
parameter s is small again. Then, the angle ω between the two vectors in the deformed
system with ei = F̂ êi can be estimated as

(F̂ ê1 , F̂ ê2 ) ˆ ˆ

1 , (I + ∇û)ê 2


(I + ∇û)ê ˆ + ∇û
ˆ T + ∇û
ˆ T ∇û}
ˆ ê1 , ê2 = 2Êij


cos(ω) = = ≈ { ∇û
kF̂ ê1 k kF̂ ê2 k (1 + dˆ(1) )(1 + dˆ(2) ) | {z }
| {z } =2Ê
≈1

û

Figure 2.4: Angular deformation under a shearing. Deformation of a line-segment√under a

shearing for the three directions ê1 = (0, 1), ê2 = (1, 0) and ê3 = (1, 1)/ 2.

For angles ω close to π/2 we use the expansion:

π π
cos (ω) ≈ − ω, ω≈ − 2Êij .
2 2

2.1.2 The rate-of-strain tensor

In fluid-dynamics even very small forces can lead to a arbitrary large ‘deformation’ of fluid-
particles, that is, to a ‘flow’ of the fluid. Thus, not the strain itself, but the rate of change
in strain is of importance. The derivation is similar to the previous section: we again regard
a differential line-element with the two end-points x and xδ = x + sδx, where kδxk = 1. We
follow the two points over a time-span t 7→ t + k and approximate
k2
x(k) = x + u(x, t + k) = x + u(x, t) +k ∂t u(x, t) + ∂tt u(x, t) + O(k 3 ) .
| {z } | {z } 2
=0
| {z }
=v(x,t)
=:R(x,k)

Then,
x(k) = x + kv(x) + R(x, k), xδ (k) = xδ + kv(xδ ) + R(xδ , k).
By δx(k) := xδ (k) − x(k) we denote the differential line-segment after he time-span k. The
length-change dδx is given by
kxδ (k) − x(k)k − kxδ − xk kδx(k)k
dδx (x, k) := lim = lim −1
s→0 kxδ − xk ksδxk
ksδx + k(v(xδ ) − v(x)) + R(xδ , k) − R(x, k)k
= lim −1
s

17
2 Fundamentals of Continuum Mechanics

and the rate of length-change by

dδx (x, k)
d˙δx (x, k) := lim .
k→0 k
Assuming differentiability of v we see

v(xδ ) − v(x) = s∇v(x)δx + O(s2 ), R(xδ , k) − R(x, k) = sO(k 2 ),

to get in first order in s:

 1
2
kδx(k)k = kδx + k∇vδx + O(k 2 )k = 1 + ({k{∇v + ∇v T }δx + O(k 2 )δx, δx) .

1
By (1 + z) 2 = 1 + z/2 + O(z 2 ) we have:

k 
kδx(k)k = 1 + {∇v + ∇v T }δx, δx + O(k 2 ).
2
And finally, for the rate of length-change
k
2 ({∇v + ∇v T }δx, δx) + O(k 2 )  1 
d˙δx (x, k) = lim {∇v + ∇v T } δx, δx = δx,


= ˙ δx .
k→0 k |2 {z }
=:˙

We call ˙ the rate-of-strain tensor. It is the fundamental measure to describe strains in

fluid-dynamics. The rate-of-strain tensor is linear and symmetric with the component-wise
representation
 
2∂1 v1 (x) ∂1 v2 (x) + ∂2 v1 (x) ∂1 v3 (x) + ∂3 v1 (x)
1
(x)
˙ = ∂2 v1 (x) + ∂1 v2 (x) 2∂2 v2 (x) ∂2 v3 (x) + ∂3 v2 (x) .

2
∂3 v1 (x) + ∂1 v3 (x) ∂3 v2 (x) + ∂2 v3 (x) 2∂3 v3 (x)

The diagonal entries ˙ii describe the relative rate of strain in direction of the coordinate axes
xi , and the trace of the matrix ˙ describes the relative speed of change in volume
X
tr ()
˙ = ∂i vi = div v.

A single entry ˙ij can be interpreted as the velocity of angular change in the xi − xj plane.

2.2 Fluid Dynamics

We describe the motion of a fluid (that is a liquid or a gas). The fluid flow is modeled by
describing its properties in points x in a volume V . This space-centered focus is the classical
Eulerian view-point. In every point x we model the velocity v and density ρ. Further, by p
we denote the pressure. These set of variables are called the primitive variable in contrast
to the conservative variables ρ the mass-density and ρv the momentum.

18
 2.2 Fluid Dynamics

2.2.1 The Conservation Equations in Fluid Dynamics

In continuum mechanics we do not describe the physical properties of single particles or the
interaction of one particle with another, instead, we describe the behavior of densities in
a continuum. This way we assume, that the observed space is completely filled with the
substance. We ignore the fact, that matter is made of discrete atoms on a very fine scale.
On length scales much greater than the atom-distant, continuum models are very accurate.

Let V (t) ⊂ R3 be a (moving) volume. The material in V (t) has certain distributed properties
like the mass-density ρ, the momentum ρv and the energy-density ρe. The basic equations
of continuum mechanics are derived from the following physical conservation principles

conservation of mass mass is not added nor removed

conservation of momentum change in momentum is equivalent to the acting force

conservation of energy energy is not added nor removed

For instance the first principle means, that the mass of a volume
Z
m(V (t)) = ρ dx,
V (t)

will not change in time

dt m(V (t)) = 0.

Let Φ̂ be some physical value (e.g. density) given for every mass point ξ and for every t.
Then, by
Φ(x, t) = Φ(x(ξ, t), t) := Φ̂(ξ, t),
we can access this value also in Eulerian coordinates (which is necessary to describe it on
the moving volume V (t)). The ‘local derivative’

∂t Φ = ∂t Φ(x, t),

indicates the local change in a fixed point in space x (that is the Eulerian view-point of a
fixed focus). The ‘material derivative’ (or total derivative)

dt Φ = dt Φ(x(ξ, t), t) = ∂t Φ(x, t) + v · ∇x Φ(x, t),

indicates the change of Φ for a certain particle ξ which is at x = x(ξ, t) at time t. By

v(x, t) = ∂t x(ξ, t) we denote the velocity of the particle.

Theorem 1 (Reynolds Transport Theorem). Let Ψ = Ψ(x, t) be a smooth scalar function.

Then, for every volume V (t) it holds

d
Z Z

Φ dx = ∂t Φ + div (Φv) dx.
dt V (t) V (t)

19
2 Fundamentals of Continuum Mechanics

For a value Φ with

d
Z
Φ dx = 0,
dt V (t)

we get with Reynolds transport theorem the following relation:

d
Z Z Z Z
Φ dx = ∂t Φ + div (Φv) dx ⇒ ∂t Φ dx = − n · vΦ do.
dt V (t) V (t) V (t) ∂V (t)

Thus, the change of a value Φ in V (t) over time equals the negative flux over V (t)’s boundary.

Conservation of Mass Physical evidence tells us, that mass is neither created nor destroyed.
We thus assume, that for each volume V it holds
Z
dt m(V ) = dt ρ dx = 0.
V

Using Φ = ρ in Reynolds transport theorem, we obtain

Z
{∂t ρ + ∇ · (ρv) dx = 0.
V

This holds for every volume V . We assume that the integrand is continuous and conclude
in every spatial point x the equation of mass conservation:

∂t ρ + ∇ · (ρv) = 0. (2.3)

Conservation of Momentum (Impuls) Let V be a volume. The forces acting on V can

be split into volume forces in V and surface forces on ∂V . Volume forces will be denoted
by a distributed function f . An example for a volume force is the gravity, which acts on
every material point in the volume. Surface forces only act with contact on the surface of
the body. Examples are pressure or frictional forces. They are described by a tensor σ. The
force acting on a surface with normal n is given by n · σ. Together, the total force F (V )
acting on V is Z Z Z
F (V ) = ρf dx + n · σ do = {ρf + ∇ · σ} dx.
V ∂V V
This stress-tensor σ includes the material properties of the fluid and will be developed later.
The momentum M (V ) of V is given by
Z
M (V ) = ρv dx.
V

Newton’s law tells that the change in momentum is equal to the acting forces
Z Z
dt M (V ) = F (V ) ⇒ dt ρv dx = {ρf + ∇ · σ} dx.
V V

We apply Reynolds transport theorem for the scalar values Φ = ρvi for i = 1, 2, 3 to get
Z Z
dt ρvi dx = {∂t (ρvi ) + ∇ · (ρvi v)} dx
V V

20
 2.2 Fluid Dynamics

which yields the point-wise equation

∂t (ρvi ) + ∇ · (ρvi v) = ρfi + (∇ · σ)i

In a compact vectorial formulation, we have

∂t (ρv) + ∇ · (ρv ⊗ v) = ρf + ∇ · σ,

where v ⊗ v := (vi vj )3i,j=1 is the dyadic product. With the conservation of mass (2.3) we get

∂t (ρv) = ∂t ρv + ρ∂t v = ρ∂t v − v∇ · (ρv).

Further, with
∇ · (ρv ⊗ v) = v∇ · (ρv) + ρv · ∇v,
we get the equation of momentum conservation

ρ∂t v + ρv · ∇v = ρf + ∇ · σ.

Conservation of Angular Momentum (Drehimpuls) The angular momentum (or moment

of momentum) of a volume with regard to the origin is given by
Z
L(V ) = x × (ρv) dx,
V

and the torque (or moment or moment of force, “Drehmoment”):

Z Z
T (V ) = x × (ρf ) dx + x × (n · σ) dx.
V ∂V

Torque and angular momentum are the rotational analogies of force and momentum. Con-
servation of angular momentum says:

dt L(V ) = T (V ).

Using Reynolds transport theorem and after lengthy computation we obtain for the change
of angular momentum:
Z


dt L(V ) = x × ρ∂t v + ρv · ∇v dx.
V

Likewise, the torque can be rewritten as

Z
T (V ) = x × (ρf + div σ) dx.
V

Here, by  = ()ijk=1,2,3 we denote the permutation tensor or Levi-Civita-Tensor, which is

given by

1

 if (i, j, k) is an even permutation of (1, 2, 3)
ijk = −1 if (i, j, k) is an odd permutation of (1, 2, 3)


0 if (i, j, k) if any symbol is repeated, i = j, i = k, j = k.

21
2 Fundamentals of Continuum Mechanics

We conclude with the point-wise equation for conservation of angular momentum

x × (ρ∂t v + ρv · ∇v − div σ − ρf ) =  : σ.

The first part is given by the conservation of momentum. Thus, conservation of angular
momentum implies
 : σ = 0 ⇔ σjk = σkj ⇔ σ = σ T .
Conservation of angular momentum is hence an additional condition requiring the symmetry
of the stress-tensor σ.
The set of conservation equations is now given by:

∂t ρ + ∇ · (ρv) = 0,
(2.4)
ρ∂t v + ρv · ∇v = ρf + ∇ · σ.

2.2.2 Material Laws in Fluid Dynamics

The conservation equations are derived from very basic principles and basically hold for
all different materials. The set of equations (2.4) includes 4 equations for the 10 physical
quantities density (1), velocity (3) and symmetric tensor σ (6). To close this gap, we will
derive constitutive laws for the dependence of the stress tensor σ on the other variables.
These laws have to be understood as modeling.
A so called Stokes fluid has the property, that the stress tensor σ is spherically symmetric if
the fluid is at rest v = 0. That is, for every two normal unit-vectors n1 and n2 we have

n1 · σ|v=0 n1 = n2 · σ|v=0 n2 .

With the symmetry σ = σ T it follows, that the stress-tensor for a fluid at rest is diagonal

σ|v=0 = −pI,

where p is the scalar hydrostatic pressure. For a general moving fluid, we introduce the shear
stress tensor τ to capture the remaining parts:

σ = −pI + τ.

τ = τ T must be symmetric. Further, the trace of τ must be zero

tr (τ ) = 0,

otherwise it would be “hidden” in the pressure-part “−pI”. We link the shear tensor τ to
the strain rate tensor ˙ via a general material law

τ = F ().
˙

For a Stokes flow we assume the following properties

symmetry if T is a symmetric tensor, then F (T ) is also a symmetric tensor. Further,
F (0) = 0.

22
 2.2 Fluid Dynamics

isotropy the material law F in invariant with regard to volume preserving, orthogonal co-
ordinate transformation, i.e.

F (QT QT ) = QF (T )QT ,

for every symmetric tensor T and every orthogonal transformation Q = (Qij )i,j=1,2,3
with QQT = I and det (Q) = 1.
If we further assume, that the material law is linear, we call the fluid a Newtonian fluid and
the stress tensor has to have the form:

σ = −pI + F ()
˙ = −pI + 2µ˙ + λtr ()I,
˙

with the two material constants µ (shear viscosity) and λ (volume viscosity). These two
parameters usually depend on the temperature and the density of the fluid. For isothermal
fluids, the shear and volume viscosity are constant in space and time, thus

div σ = −∇p + div τ, τ := µ{∇v + ∇v T } + λdiv vI.

From physical observation we use the relation

2
3λ + 2µ = 0, µ≥0 ⇒ λ = − µ,
3
between the parameters. This relation holds for one-atom-gases. The stress tensor for a
Stokes fluid is:
2
σ = −pI + µ{∇v + ∇v T } − µdiv vI (2.5)
3
Finally, the conservation equations for a Newtonian Stokes fluid are given by

∂t ρ + ∇ · (ρv) = 0,
2
ρ∂t v + ρv · ∇v − µ{∇v + ∇v T } + µdiv vI + ∇p = ρf.
3

2.2.3 Incompressible Flows, the Navier-Stokes Equations

Even with large forces, fluids are very difficult to compress. Thus, the volume V (t) nearly
does not change with time. If the volume V (t) is constant in time, the flow is called an
incompressible flow:
d
Z
dx = 0.
dt V
Reynolds transport theorem, used on the constant Φ = 1 yields
Z
∇ · v dx = 0 ⇒ ∇ · v = 0.
V

Conservation of mass then simplifies to

∂t ρ + v · ∇ρ = 0.

23
2 Fundamentals of Continuum Mechanics

For the divergence of the stress tensor (2.5) it holds

2
−div σ = ∇p − µ |∇ {z ∇v T} + ∇ · divv
· ∇v} −µ |∇ ·{z | {z } I = −µ∆v + ∇p.
3
=∆v =∇div v=0 =0

If the fluid is homogeneous, the density does not vary in space and time ρ(x, t) = ρ0 and the
incompressible Navier-Stokes equations are given by

1
∂t v + v · ∇v − ν∆v + ∇p = f,
ρ0 (2.6)
∇ · v = 0.

The value ν := µ/ρ0 is called the kinematic viscosity opposed to the dynamic viscosity µ.
This system is closed by boundary and initial values. Let Ω ⊂ Rd be the domain with
boundary Γ = ∂Ω. We split this boundary into Γ = ΓD ∪ ΓN , where ΓD represents parts of
the boundary with Dirichlet condition and ΓN with Neumann or Robin boundary conditions.
Then, we impose the boundary and initial values:

v(x, 0) = v 0 in Ω, v = g D on ΓD , n · σ = g σ on ΓN ,

where g D and g σ are appropriate functions.

2.2.4 Variational formulation of the incompressible Navier-Stokes Equations

We derive the weak formulation of Equation (2.6) by multiplying with suitable test-functions.
We introduce the following function spaces:

H01 (Ω; ΓD ) = {φ ∈ H 1 (Ω)d : v = 0 on ΓD },

L20 (Ω) = {ξ ∈ L2 (Ω) : (ξ, 1)Ω = 0}.

The variational formulation of problem (2.6) is

Problem 1 (Incompressible Navier-Stokes equations). On Ωf (t) find v ∈ v D + Vf , Vf :=
H01 (Ω; ΓD ) and p ∈ Qf , Qf := L20 (Ω), such that

ρ0 (∂t v + (v · ∇)v), φ) + (µ∇v, ∇φ) = (ρ0 f, φ) + hg σ , φiΓN ∀φ ∈ Vf

(2.7)
(∇ · v, ξ) = 0 ∀ξ ∈ Qf ,

where
σf := −pI + ρf νf ∇v + ∇v T .


While the Dirichlet-condition v = g D is embedded into the function space: the function
v D ∈ H 1 (Ω)d is a continuation of g D into the domain Ω. The Neumann-condition on the
stresses n · σ on ΓN is ensured due to integration by parts:
1
(µ∇v, ∇φ)Ω − (p, ·∇φ)Ω = (−µ∆v + ∇pI, φ)Ω + hµn · ∇v − p · nI, φiΓN .
ρ0

24
 2.3 Structure Mechanics

2.3 Structure Mechanics

In structure mechanics, we describe the deformation and movement of a volume V filled with
an elastic material. By V̂ we will denote the volume in the undeformed reference state. Due
to forces acting in the volume or on the surface of the volume, the structure will experience
some deformation û(x̂, t). This deformation can either be a real dynamic process yielding
a time-depending deformed volume V (t) or it can result in a new equilibrium V after some
initial dynamic deformation. In structure dynamics we are interested in the deformation
field û(x̂, t) for every material-point x̂ of the reference volume V̂ . This is the classical,
material-centered Lagrangian viewpoint.

2.3.1 The Conservation Equations in Structure Dynamics

In structure dynamics, the velocity of particles is of lesser interest. Instead, we regard the
transition of a volume V̂ under strains from a reference state to a new equilibrium V (t). We
search for the deformation vector û(x̂, t)

V̂ 3 x̂ = x(x̂, 0) → x(x̂, t) = x̂ + û(x̂, t) ∈ V (t)

For structure mechanics, the Lagrangian viewpoint is natural. We denote all variables and
entities given in the reference framework V̂ by a “hat” and all variables on the deformed
domain V (t) or V without it. By v̂(x̂, t) we denote the velocity of particle x̂ at time t
(wherever this particle may be in V (t)):

v̂(x̂, t) := dt x(x̂, t) = dt u(x̂, t) = ∂t û(x̂, t).

The deformation in the deformed volume V is defined by u(x, t) := û(x̂(x, t), t) or likewise
(since x̂(x, t) is invertible) u(x(x̂, t), t) := û(x̂, t). This is the Eulerian formulation of the
deformation. Then, for the Eulerian velocity we get

v(x, t) := v̂(x̂(x, t), t) = dt x(x̂, t) = ∂t û(x̂, t).

Thus, by this consideration, velocity and displacement can be given in both frameworks via
the easy relation v(x, t) = v̂(x̂, t) and u(x, t) = û(x̂, t) for the couple x̂ = x̂(x, t). However
we note, that the spatial derivatives are not immediately available. In particular, we have
ˆ
∇u 6= ∇û!
We assume the following material properties. The deformation is isothermal: no mechanical
effects are caused by temperature gradients. In the reference configuration V̂ , all physical
parameters (that is the density ρ̂(x̂, 0) = ρ̂0 (x̂), initial deformation û(x̂, 0) = û0 (x̂) and
initial velocity v̂(x̂, 0) = v̂ 0 (x̂) are given.
The behavior of the structure is then determined by the two conservation principles of mass
and momentum valid on the deformed domain V (t). Conservation of mass reads

d
Z
ρ(x, t) dx = 0, (2.8)
dt V (t)

25
2 Fundamentals of Continuum Mechanics

and momentum:
d
Z Z Z
ρ(x, t)v(x, t) dx = ρ(x, t)f (x, t) dx − n(x) · σ(x, t) dox , (2.9)
dt V (t) V (t) ∂V (t)

where ρf is a prescribed volume force acting on V (t) and σ is the stress tensor describing
the surface forces. The conservation principles are first derived on the (moving) deformed
domain V (t).
Here, the difficulty of structural modeling is obvious (this will also be the underlying difficulty
of fluid-structure interaction): we intend to formulate the structure equations in Lagrangian
coordinates on the reference system V̂ . The governing principles however are determined
on the deformed equilibrium system V (t). To derive the set of equations we thus need to
transform the conservation principles to the reference system V̂ . In principle, it would be
possible to formulate the structure equations in the Eulerian framework, then however the
coordinates itself x = x̂ + û are unknowns in the system.

Conservation of Mass The density distribution ρ̂0 (x̂) is known at time t = 0. Conservation
of mass (2.8) means for all t ≥ 0
Z Z
0
ρ̂ (x̂) dx̂ = ρ(x, t) dx.
V̂ V (t)

Transformation via T̂ : V̂ → V (t) yields

Z Z Z
ρ(x, t) dx = 0
ρ̂ (x̂) dx̂ = ρ̂0 (x̂(x, t))Jˆ−1 dx.
V (t) V̂ V (t)

Conservation of mass gives us the constituting law for the density ρ(x, t) in the deformed
framework:
ρ(x, t) = Jˆ−1 ρ̂0 (x̂), (2.10)
ˆ ˆ
where J = det(I + ∇û). In Eulerian formulation we can express the density by
ρ(x, t) = J(x, t)ρ̂0 (x̂(x, t))

Conservation of Momentum We transform the left side of (2.9) and use the conservation
of mass (2.10):
d d
Z Z Z Z Z
ρv dx = ρ(x(x̂, t))Jˆ v̂ dx̂ = 0
ρ̂ ∂t v̂ dx̂ = ρ̂ (x̂(x̂, t))Jˆ−1 ∂t v dx =
0
ρ∂t v dx.
dt V (t) dt V̂ | {z } V̂ V (t) V (t)
ρ̂0 (x̂)

Together with the right side of (2.9) conservation of momentum gets (written in the Eulerian
framework)
ρ∂t v = ρf + div σ in V (t).
With v = ∂t u we can write the conservation equations in structure mechanics as

ρ = J ρ̂0 ,
(2.11)
ρ∂tt u = ρf + div σ.

26
 2.3 Structure Mechanics

Like in the fluid case, we close this general system with initial and boundary values. The
boundary Γ = ∂Ω is again split into a Dirichlet part ΓD and a Neumann part ΓN . Since the
structure problem is described by a wave-equation, we have to specify initial values for the
deformation and its derivative (the velocity):

u(x, 0) = u0 (x) in Ω, v(x, 0) = v 0 (x) in Ω, u = g D on ΓD , n · σ = g σ on ΓN .

These equations are written on the deformed domain V (t) which is determined by the un-
known deformation u(x, t) itself. To close this gap, it is necessary to express the balance
equations (2.11) in the Lagrangian reference system.

2.3.2 The Piola Transformation

We need to build a transformation between the Eulerian space V (t) and the Lagrangian V̂
which maps forces (vectors) in the right way. The transformation of the coordinate system
(e.g. of normal vectors) needs to be considered.
Definition 2 (Piola Transform). Let T̂ : V̂ → V be a diffeomorphism with F̂ := ∇ ˆ T̂ ,
Jˆ := det F̂ . Further, let v ∈ H (Ω) be a differentiable vector field. Then, the Piola transform
1

of v is given by
JˆF̂ −1 v̂.

First, we derive the following Lemma

Lemma 1 (Divergence of the Piola transform). Let T̂ : V̂ → V with T̂ (x̂) = x̂ + û(x̂) be a
C 2 -diffeomorphism and v : V → Rd with d = 2, 3 be a differentiable vector field in V . Then,
for the divergence of the Piola transform JˆF̂ −1 v̂ it holds

Jdiv c (JˆF̂ −1 v̂),

ˆ v = div (2.12)

where v̂ : V̂ → Rd is given by v̂(x̂) = v(T̂ (x̂)).

ˆ T̂ )ij =
Proof: We show the proof in the two-dimensional case. With x(x̂) = T̂ (x̂), F̂ij = (∇
∂ˆj xi and Jˆ := det F̂ , we write
! !
∂ˆ2 x2 −∂ˆ2 x1 ∂ˆ2 x2 v̂1 − ∂ˆ2 x1 v̂2
JˆF̂ −1 = , JˆF̂ −1 v̂ = .
−∂ˆ1 x2 ∂ˆ1 x1 −∂ˆ1 x2 v̂1 + ∂ˆ1 x1 v̂2

The divergence on the right side of (2.12) can be written as

d (JˆF̂ −1 v̂) = ∂ˆ1 (∂ˆ2 x2 v̂1 − ∂ˆ2 x1 v̂2 ) + ∂ˆ2 (−∂ˆ1 x2 v̂1 + ∂ˆ1 x1 v̂2 )
div
= v̂1 (∂ˆ12 x2 − ∂ˆ21 x2 ) + v̂2 (∂ˆ21 x1 − ∂ˆ12 x1 )
+ ∂ˆ2 x2 ∂ˆ1 v̂1 − ∂ˆ2 x1 ∂ˆ1 v̂2 − ∂ˆ1 x2 ∂ˆ2 v̂1 + ∂ˆ1 x1 ∂ˆ2 v̂2 .

Since T̂ is two times differentiable, the partial derivatives commute ∂ˆi ∂ˆj x = ∂ˆj ∂ˆi x) and the
first row in the last term dissappears. Transforming to the Eulerian system using

∂ˆi v̂j = ∂1 vj ∂ˆi x1 + ∂2 vj ∂ˆi x2 ,

27
2 Fundamentals of Continuum Mechanics

yields (if we collect the terms with regard to v̂j )

d (JˆF̂ −1 v̂) = (∂1 v1 + ∂2 v2 )(∂ˆ2 x2 ∂ˆ1 x1 − ∂ˆ1 x2 ∂ˆ2 x1 )
div
+ ∂2 v1 (∂ˆ2 x2 ∂ˆ1 x2 − ∂ˆ1 x2 ∂ˆ2 x2 ) + ∂1 v2 (∂ˆ1 x1 ∂ˆ2 x1 − ∂ˆ2 x1 ∂ˆ1 x1 ).
By noting Jˆ = ∂ˆ1 x1 ∂ˆ2 x2 − ∂ˆ1 x2 ∂ˆ2 x1 the result follows. 
This lemma helps us to get a formula for the transformation of surface integrals:
Lemma 2 (Transformation of Surface Integrals). Let T̂ : V̂ → V be the transformation as
above. Further, let v ∈ H 1 (V ) be a differentiable vector-field. Then,
Z Z
v · n do = (JˆF̂ −1 v̂) · n̂ dô.
∂V ∂ V̂

Proof: With Gauss’ theorem followed by transformation to V̂

Z Z Z
v · n do = div v dx = ˆ v dx̂.
Jdiv
∂V V V̂
Using Lemma 1 and with Gauss’ theorem
Z Z Z
ˆ v dx̂ =
Jdiv d JˆF̂ −1 v̂) dx̂ =
div( (JˆF̂ −1 v̂) · n̂ dô,
V̂ V̂ ∂ V̂
we complete the proof. 
The transformation process is necessary to correctly express the surface forces n · σ in the
reference system. We need to find a corresponding tensor P̂ to write the normal stresses in
the Lagrangian system satisfying
Z Z
n · σ do = n̂ · P̂ dô.
∂V ∂ V̂

Lemma 3 (First Piola-Kirchhoff stress-tensor). Let T̂ : V̂ → V be defined as above. Then,

the surface force on ∂V is transformed to the reference system by
Z Z
n · σ do = n̂ · (Jˆσ̂ F̂ −T ) dô.
∂V ∂ V̂
The tensor
P̂ := Jˆσ̂ F̂ −T ,
is called the first Piola-Kirchhoff stress-tensor.

Proof: By Fi we denote the i-th component of the surface force:

Z XZ Z
Fi := (n · σ)i do = nk σki , do = n · σi do,
∂V k ∂V ∂V

where σi is the i-th column-vector of σ. Since σ = σ T is symmetric, σi can also be understood

as the i-th row-vector. With Lemma 2 we get
Z
Fi = n̂ · (JˆF̂ −1 σ̂i ) dô.
∂ V̂
Using the symmetry of the stress-tensor σ we complete the proof. 
We further define

28
 2.3 Structure Mechanics

Definition 3 (Second Piola-Kirchhoff stress-tensor). The symmetric tensor

Ŝ = F̂ −1 P̂ = JˆF̂ −1 σ̂ F̂ −T ,

is the second Piola-Kirchhoff stress-tensor.

We emphasize the differences between the three stress-tensors σ, P̂ and Ŝ. The Cauchy
stress-tensor σ comprises the “real stress” in the current configuration at a point x. The
first Piola-Kirchhoff tensor indicates the stress expressed in the current configuration for a
material point x̂ in reference configuration. The second Piola-Kirchhoff tensor comprises the
stress with regard to the reference configuration.
Finally, we can transform the balance equations in structure dynamics (2.11) to the La-
grangian reference system:
Z Z
ρ̂0 ∂t v̂ dx̂ = d (Jˆσ̂ F̂ −T ) dx̂.
ρ̂0 fˆ + div
V̂ V̂

Likewise, if ρ̂0 , v̂ and σ̂ are smooth:

d (Jˆσ̂ F̂ −T ).
ρ̂0 ∂tt û = ρ̂0 fˆ + div (2.13)

2.3.3 Material Laws in Structure Dynamics

The material law describes the dependence of the stress tensor σ on the strain-tensor Ê
1 ˆ ˆ T + ∇u
ˆ T ∇u},
ˆ
P̂ = C(Ê), Ê = {∇u + ∇u
2
via the elasticity law C. We are interested in elastic materials. Elasticity means, than if
a forces are applied on a body V̂ in reference configuration it will be deformed V (t). If
the forces are driven back to zero, the deformed body will turn back into the reference
configuration V (t) → V̂ . We note that real elasticity is an idealization. In contrast, a
material called plastically, if the deformation (or parts of the deformation) are irreversible.
Finally, a material is called viscoelastic, if it has properties of a viscous fluid as well as of an
elastic structure.
Further we distinguish between isotropic and anisotropic material behavior. Common ex-
amples for an anisotropic materials are blood vessels. They are by more easy to deform in
radial than in axial direction.
Finally, a body is said to be materially homogeneous, if its response to applied loads is
independent of the position within the body.
We call a material incompressible, if it preserves volume:
Z Z Z
t≥0: dx̂ = dx = Jˆ dx̂,
V̂ V (t) V̂

thus if Jˆ = 1. In that case, the density is given by ρ(x, t) = ρ̂0 (x(x̂, t)).

29
2 Fundamentals of Continuum Mechanics

2.3.4 Linear Models of Elasticity

We only consider materials, where the stress to strain relation is linear (in Lagrangian and
Eulerian coordinates):
1
P̂ = Ĉ Ê, Ê = (F̂ T F̂ − I).
2
For a linear, isotropic and symmetric material, the following law can be derived:
Lemma 4 (Hooke’s Law). Under the assumption of a linear, homogenous and isotropic
material law, the stress strain relation must be of the type

Ŝ = 2µÊ + λtr (Ê)I,

with two material parameters µ and λ.

With Definition 3 we can calculate the Cauchy stress-tensor σ̂ by

1
σ̂ = F̂ (2µÊ + λtr(Ê)I)F̂ T .
Jˆ
The first Piola-Kirchhoff tensor is given by

P̂ = F̂ (2µÊ + λtr(Ê)I).

If µ 6= 0 and 3λ + 2µ 6= 0 Hooke’s law is invertible:

1 λ
Ê = Ŝ − tr(Ŝ)I.
2µ 2µ(3µ + 3λ)
We introduce the Poisson’s ratio ν and the Young’s modulus E by
λ µ(3λ + 2µ)
ν := , E := ,
2(λ + µ) λ+µ
to write in compact form
1+ν ν
Ê = Ŝ − tr(Ŝ)I.
E E
We try to identify reasonable condition on the elasticity parameters λ, µ, ν and E. The
deformation is assumed to be small, so that ˆ is a good approximation of Ê. We analyze the
following cases:

1. Uniform pressure: Ŝij = −pδij . Hooke’s law yields the corresponding strains by
1+ν ν 1 − 2ν p
ˆij = Ŝij − tr(ˆ
)δij = −p δij = − δij .
E E E 3λ + 2µ
The diagonal entries ˆii indicate the length-change in directions of the coordinate axes.
The trace tr(ˆ) the relative volume change. A positive pressure p must go along with
a decrease of volume, thus we note

3λ + 2µ > 0.

30
 2.3 Structure Mechanics

2. Pressure in x1 -direction:  
−p 0 0
Ŝ =  0 0 0
 
0 0 0
The strain is estimates as  
−1 0 0
p 
ˆ =  0 ν 0

E
0 0 ν
A pressure in x1 -direction (in negative normal direction) should result in a contraction
in x1 direction and an expansion in x2 and x3 direction, thus ˆ11 < 0 and ˆ22 , ˆ33 > 0.
We note
E > 0, ν > 0.

3. Finally, we apply a shearing in the x1 − x2 plane:

   
0 1 0 0 1 0
1 
Ŝ = 1 0 0 , ˆ = 1 0 0  .
  
2µ
0 0 0 0 0 0

A shearing leads to an angular change. For small deformations, we have found the
relation ˆij = 21 ( 12 π − ωij ), where ωij is the angular change between in the xi − xj
plane. We thus conclude with
µ > 0.

We combine the previous results to find the conditions

3λ + 2µ > 0, µ > 0, ν > 0, E > 0.

The material governed by this linear material law is called the St. Venant Kirchhoff-material.
It considers the full nonlinear strain tensor and full geometric nonlinearities. Just the ma-
terial is assumed to have a linear stress-strain relationship. The problem is to find û, such
that
 
d F̂ (2µÊ + λtr(Ê)I) = ρ̂0 fˆ.
ρ̂0 ∂tt û − div

ˆ
Further simplification can be made by assuming that the deformation is very small k∇ûk 
1. Then, the strain tensor can be approximated by the linear Ê ≈ ˆ and we use the simpli-
fication F̂ ≈ I. We derive the set of equations

ρ̂0 ∂tt û − div )) = ρ̂0 fˆ.

 + λtr(ˆ
g (2µˆ

This equation can be transformed into

µ λ+µ ˆd
∂tt û − 0
∆û + ∇div û = fˆ.
ρ̂ ρ̂0

This fully linear equation is called the Navier-Lame problem.

31
2 Fundamentals of Continuum Mechanics

The equations are closed by boundary and initial values. These have to be transformed onto
the reference coordinates:

û(x̂, 0) = û0 (x̂), v̂(x̂, 0) = v̂ 0 (x̂) in Ω̂, û = ĝ D on Γ̂D , n̂ · (Jˆσ̂ F̂ −T ) = Jˆĝ σ F̂ −T on Γ̂N .

2.3.5 Variational formulation of the elastic structure equations

In the context of fluid-structure interaction, we write the structure-equations as a set of

first-order equations in time, introducing the velocity v̂ = ∂t û as an additional variable.
Then, by
H01 (Ω̂; Γ̂D )d = {φ ∈ H 1 (Ω)d : φ = 0 on Γ̂D },
we denote the test-space. Find û ∈ û0 + H01 (Ω̂; Γ̂D )d , v̂ ∈ v̂ 0 + H01 (Ω̂; Γ̂D )d , such that

ˆ φ̂) = (ρ̂0 fˆ, φ̂) + hJˆĝ σ F̂ −T , φ̂iΓN

(ρ̂0 ∂t v̂, φ̂) + (Jˆσ̂ F̂ −T , ∇ ∀φ ∈ H01 (Ω̂; Γ̂D ),
(2.14)
(dt û, ψ̂) = (v̂, ψ̂) ∀ψ ∈ H01 (Ω̂; Γ̂D ).

The Dirichlet boundary conditions are built into the function space and the Neumann con-
dition on the stresses is given due to integration by parts.

32
3 The Fluid-Structure Interaction Problem

In this chapter we describe the coupled fluid-structure interaction problem. As a prototypical

application we have the configuration from the introduction in mind:

Ω̂f Γ̂wall Ωf (t1 ) Ωf (t2 )

Γ̂out
Γ̂in
Γ̂i Γi (t) Γi (t)
Ω̂s Γbase Ωs (t1 ) Ωs (t2 )

Figure 3.1: Prototypical FSI configuration.

A domain Ω̂ ∈ R2 in reference configuration at time t = 0 is split into a fluid and solid part

Ω̂ = Ω̂f ∪ Ω̂s .

An incompressible fluid is flowing in Ω̂f and an elastic structure is given in Ω̂s which is
sticked to at the boundary of the domain ûs = 0. Due to the coupled dynamics of fluid and
structure, the solid domain will be deformed Ω̂s → Ωs (t) which also imposes a deformation
of the flow domain Ω̂f → Ωf (t). We split the boundary of Ω̂ = Ω̂f ∪ Ω̂s = Ωf (t) ∪ Ωs (t)
into the Dirichlet part ΓD = Γin ∪ Γwall ∪ Γbase as noted in Figure 3.1, and into the outflow
boundary Γout :
∂ Ω̂ = Γin ∪ Γwall ∪ Γbase ∪ Γout .
Since these boundaries do not move with time, Lagrangian and Eulerian coordinates coincide
here. We further denote by Γi the interface between fluid and structure and we define in
Eulerian and Lagrangian coordinates

Γ̂i = Ω̂f ∪ Ω̂s , Γi (t) = Ωf (t) ∪ Ωs (t).

This internal boundary moves in time.

On the moving domain Ωf (t) the incompressible Navier-Stokes equations are valid: find
vf ∈ v D + H01 (Ωf (t); ΓD )d and pf ∈ L20 (Ωf (t)) such that

(ρf (∂t vf + vf · ∇vf ), φ)Ωf (t) + (σf , ∇φ)Ωf (t) = (ρf ff , φ)Ωf (t) + hgfσ , φiΓi (t) + hgfout , φiΓout ,
(∇ · vf , ξ)Ωf (t) = 0,
(3.1)

33
3 The Fluid-Structure Interaction Problem

where σf = νf (∇vf + ∇vfT ) − pf I. On the outflow boundary Γout , we impose the do nothing
boundary condition νf ∂n v −pn = 0 by choosing an appropriate gfout to remove the transposed
part in the stress tensor σf . The boundary function gfσ on the interface Γi (t) will cope with
the stresses from the solid-part.
In the fixed solid reference domain, the elastic structure equation is given. Find û ∈ ûD +
H01 (Ω̂s ; Γbase )d and v̂ ∈ v̂ D + H01 (Ω̂s ; Γbase )d such that

ˆ φ̂) = (ρ̂s fˆs , φ̂) + hJˆĝ σ F̂ −t , φ̂i ,

(ρ̂s ∂t v̂s , φ̂)Ω̂s + (Jˆσ̂s F̂ −T , ∇ Ω̂s Ω̂f s Γ̂i
(3.2)
(∂t û, ψ̂)Ω̂s = (v̂, ψ̂)Ω̂s ,

where σ̂s is the Cauchy stress tensor for a St. Venant-Kirchhoff material (formulated in the
Lagrangian coordinate system)

σ̂s = Jˆs−1 F̂s (2µÊs + λs tr(Ês )I)F̂sT , ˆ s , Jˆs := det(F̂s ), Ês := 1 (F̂sT F̂s − I).
F̂s := I + ∇û
2

On the interface Γ̂i , the function ĝsσ copes with the stresses from the fluid domain.
The interaction of the fluid with the solid is given by the continuity of the velocity and by
a balance of forces:
vf = vs on Γi (t), n · σf = n · σs on Γi (t). (3.3)

With these notations, we can formulate a first version of the fluid-structure interaction
problem

Problem 2 (First Fluid-Structure Interaction Formulation). Find vf ∈ vf0 + Vf , pf ∈ Lf ,

and us ∈ u0s + Vs and vs ∈ vs0 ∈ Vs , such that (3.1, 3.2, 3.3) are fulfilled for all φ ∈ Vf ,
ξ ∈ Lf and φ̂, ψ̂ ∈ Vs .

The formulation of the coupled system is not satisfying:

1. Both systems (3.1) and (3.2) are formulated in different coordinate systems. Ωf (t) is
moving, while Ω̂s is fixed. We will need different triangulations for both subdomains.
2. The coupling conditions (3.3) are formulated on the common interface. However, since
using different coordinate systems, we do not have a common interface. Instead we
need to express both conditions in one of the two coordinate systems, e.g. by using (in
Lagrangian formulation)

x̂ ∈ Γ̂i : vs (x̂) = v̂f (x̂) = vf (x̂ + ûs ).

The transformed coordinate x = x̂ + ûs will most certainly not be a mesh-point in

the fluid domain. The interface conditions thus cannot be fulfilled exactly, instead
we need some interpolation operators. For the normal-stresses condition we further
need to transform the tensors with the Piola transformation onto the same coordinate
system.

34
 3.1 Fluid Flows in ALE Formulation

3. Once the solid is deformed, the domain partitioning will change

Ω̂f ∪ Ω̂s → Ωf (t) ∪ Ωs (t),

thus, the Navier-Stokes equations are formulated on a moving domain. In addition,

the interface-coupling condition is given on the moving interface

Γi (t) = Ωf (t) ∩ Ωs (t).

And since in one time-step tn → tn+1 the domain partitioning is changing, we have
different domains and we thus need different triangulations at the beginning and at
the end of the time-step. How should an appropriate function space look like?

The clue to treat these problems is to formulate both systems in one common coordinate
system. In the following section we introduce the Arbitrary Lagrangian Eulerian Coordi-
nates (ALE) for the Navier-Stokes equations. We express the fluid problem on the fixed
computational domain Ω̂f which matches the solid domain for all times.

3.1 Fluid Flows in ALE Formulation

In fluid-structure interaction, the fluid problem is given on the moving domain Ωf (t). As
discussed above, this gives rise to various problems, most severe, the domain and with it the
computation mesh changes from one time-step to the other tn → tn+1 . To work around this
problem we want to formulate the flow problem on a fixed domain Ω̃f . Let T̃f : Ω̃f → Ωf (t)
be a mapping between this fixed domain Ω̃f and the moving domain Ωf (t). We will use
this mapping to transform the Navier-Stokes equations onto the fixed domain. While this
is not necessarily required, we think of this fixed domain Ω̃f as Ω̂f , the counterpart to the
Lagrangian structure domain Ω̂s . Later on, we will discuss cases, where we will allow for
general reference domains of the flow problem, where Ω̃f does not have to coincide with
Ω̂f . Because the fluid’s reference domain is arbitrary, and the structure will be formulated
in Lagrangian coordinates, the resulting fluid-structure formulation is called the Arbitrary
Lagrangian Eulerian (ALE) formulation.

Let T̂f : Ω̂f → Ωf (t) be a C 2 -diffeomorphism. We call T̂f the ALE mapping. We define the
gradient F̂f and its determinant Jˆf by

T̂f : Ω̂f → Ω̂f (t), ˆ T̂f ,

F̂f := ∇ Jˆf := det(F̂f ).

We define the principle variables velocity v̂ and pressure p̂ on the reference domain Ω̂f via
this transformation:

v̂(x̂, t) = v(Tf (x̂, t), t), p̂(x̂, t) = p(T̂f (x̂, t), t).

The spatial and temporal derivatives of the variables are given by the following Lemma:

35
3 The Fluid-Structure Interaction Problem

Lemma 5 (Derivatives of the ALE mapping). Let T̂f : Ω̂f → Ωf (t) be a C 1 -diffeomorphism,
f ∈ H 1 (Ω(t)) be a differentiable function and v ∈ H 1 (Ω(t))d a differentiable vector-field.
Then it holds for fˆ(x̂, t) := f (x, t) and v̂(x̂, t) = v(x, t)
∂t f = ∂t fˆ − (F̂f−1 ∂t T̂f · ∇)
ˆ fˆ, (3.4)
∇f = F̂f−T ∇
ˆ fˆ, (3.5)
∇v = ˆ F̂ −1 ,
∇v̂ (3.6)
f
−1 ˆ fˆ.
(v · ∇)f = (F̂f v̂ · ∇) (3.7)

Proof: For the partial temporal derivative ∂t f (x, t) we get with the chain rule
∂t f = ∂t f (x, t) = ∂t fˆ(T̂f−1 (x, t), t) = ∂t fˆ(x̂, t) + ∂t T̂f−1 · ∇
ˆ fˆ. (3.8)
For the inverse of the ALE mapping it holds by differentiating
T̂f ◦ T̂f−1 = id ⇒ 0 = ∂t T̂f (T̂f−1 (x, t), t) = ∂t T̂f + ∇
ˆ T̂f ∂t T̂ −1 (x, t) ⇒ ∂t T̂ −1 = −F̂ −1 ∂t T̂f .
f f f
(3.9)
Then, combining (3.8) and (3.9) we derive the first statement (3.4).
For the spatial derivative we get by component-wise calculation
d d
∂i f (x, t) = ∂i fˆ(T̂f−1 (x, t), t) = ∂ˆj fˆ∂i [T̂f−1 ]j = ∂ˆj fˆ[F̂f−1 ]ji ∇f = F̂f−T ∇
ˆ fˆ,
X X
⇒
j=1 j=1

which yields (3.5). Relation (3.6) follows by applying (3.5) to the components of v and by
noting that the Jacobian ∇v is given by the row-vectors ∇viT . For the convective term (3.7)
we get
d d d
v̂i ∂ˆj fˆ[F̂f−1 ]ji = [F̂f−1 v̂]j ∂ˆj fˆ = (F̂f−1 v̂ · ∇)
ˆ fˆ.
X X X
(v · ∇)f = vi ∂i f =
i=1 i,j=1 j=1


Then, the variational formulation of the Navier-Stokes equations (2.7) on the moving domain
Ωf (t) can be mapped onto the reference domain.
We treat the different terms separately, starting with the divergence condition:
Z Z
(∇ · v, ξ)Ωf (t) = div(v)ξdx = ˆ
Jˆf div(v)ξdx̂.
Ωf (t) Ω̂f

We assume that T̂f is a C 2 -diffeomorphism. Then, with Lemma 1, we conclude

d Jˆ F̂ −1 v̂), ξ)
(div(v), ξ)Ωf (t) = (div( ˆ .
f Ω̂f

Using (3.4) and (3.7) to transform the momentum equations, we get

(ρf ∂t v, φ)Ωf (t) = Jˆf ρ̂f (∂t v̂ − (F̂f−1 ∂t T̂f · ∇)v̂),

ˆ


φ̂ Ω̂f
,
= Jˆf ρ̂f (F̂f−1 v̂ · ∇)v̂,
ˆ



ρf (v · ∇v), φ Ωf (t)
φ̂ Ω̂f
,

36
 3.1 Fluid Flows in ALE Formulation

where ρ̂f (x̂, t) = ρf (x, t) is the fluid’s density expressed in the artificial coordinates on Ω̂f .
Combined, we get
= Jˆf ρ̂f (∂t v̂ + (F̂f−1 (v̂ − ∂t T̂f ) · ∇)v̂),
ˆ



ρf (∂t v + (v · ∇)v), φ Ωf (t)
φ̂ Ω̂f
.

Finally, the viscous term in the momentum equation gets

ˆ φ̂F̂ −1 ) = (Jˆσ̂f F̂ −T , ∇
(σf , ∇φ)Ωf (t) = (Jˆσ̂f , ∇ ˆ φ̂) ,
f Ω̂f f Ω̂f

where σ̂f (x̂) = σf (T̂f (x̂, t)) is the fluid’s stress tensor in the artificial coordinate system on
Ω̂f . It holds with (3.6)
ˆ F̂ −1 + F̂ −T ∇v̂
σ̂(x̂) = −p(T̂f (x̂))I + ρ̂f νf ∇v(T̂f (x̂)) + ∇v(T̂f (x̂))T = −p̂I + ρ̂f νf ∇v̂ ˆ T .
 
f f

Finally, the right hand side of the equation is easily transformed by

(ρf f, φ)Ωf (t) = (Jˆf ρ̂f fˆ, φ̂)Ω̂f .

Then, the complete set of equations in ALE coordinates is

Problem 3 (Navier-Stokes equations in ALE coordinates). Let T̂f : Ω̂f → Ωf (t) be a C 2
diffeomorphism. Then, the Navier-Stokes equations in artificial coordinates are given on Ω̂f
by finding v̂ ∈ v̂ 0 + V̂f , V̂f := H01 (Ω̂f ; Γ̂D )d and p̂ ∈ Q̂f , Q̂f := L20 (Ω̂f ) such that

Jˆf ρ̂f (∂t v̂ + (F̂f−1 (v̂ − ∂t T̂f ) · ∇)v̂),

ˆ + (Jˆf σ̂f F̂f−T , ∇
ˆ φ̂) = (Jˆf ρ̂f fˆ, φ̂)


φ̂ Ω̂f Ω̂f Ω̂f
c Jˆ F̂ −1 v̂), ξ)
(div( ˆ ∀φ̂ ∈ V̂f , ξˆ ∈ Q̂f ,
f Ω̂f = 0,
(3.10)

where ρ̂f is the density given on Ω̂f and

ˆ F̂ −1 + F̂ −T ∇v̂
ˆ T .

σ̂f = −p̂I + ρ̂f νf ∇v̂ f f

If Ω = Ω̂f ∪ Ω̂s , we can formulate the fluid-structure interaction problem on one common
domain. The interface Γ̂i = Ω̂f ∩ Ω̂s is just the interface between both computational
domains. Further, by solving the Navier-Stokes equations in ALE coordinates, the domain
is fixed in time. Discretization is straightforward: we generate a triangulation Ω̂f,h of Ω̂f
and can discretize with time-stepping schemes in time and finite elements in space. The
interface condition is given by
v̂f (x̂, t) = v̂s (x̂, t) and n̂f · (Jˆf σ̂f F̂f−T ) + n̂s · (Jˆs σ̂s F̂s−T ) = 0 on Γ̂i .
We note, that even though the formulation looks just like the Lagrangian structure formu-
lation, we stress the fact, that the transformation T̂f is arbitrary and in general not the
transformation to the Lagrangian coordinate system. Comparing the flow equations in ALE
coordinates (Problem 3) with the Eulerian formulation in Problem 1, we observe strong
nonlinearities introduced by the transformation. In particular, its gradient F̂f enters into
the equation in various ways. For the formulation to be well-posed, it has to be at least
invertible. Since this transformation will not be given analytically (because it will have to
depend on the unknown solution itself to fit the structure’s domain), this will give rise to
serious problems.

37
3 The Fluid-Structure Interaction Problem

3.2 Construction of the ALE-mapping T̂f

The ALE formulation of the Navier-Stokes equations is useful whenever the problem is given
on a moving domain. Thinking of fluid-structure interaction problems, the domain movement
is stated by the solution of the coupled problem. We will construct the transformation T̂f
similar to the transformation of the structure domain. This is given by:
T̂s (x̂) = x̂ + ûs (x̂) in Ω̂s ,
where the deformation ûs can be regarded as the deformation of the domain. In the fluid
domain, we will follow the same approach. Let Tf : Ω̂f → Ωf (t) be given by
T̂f (x̂) = x̂ + ûf (x̂) in Ω̂f ,
where we denote by ûf the “deformation of the fluid domain”.
A natural choice for the ALE-mapping would be the transformation to material-centered
Lagrangian coordinates. The deformation ûf follows the trajectories of the fluid and is
naturally given by
ûf = ∂t v̂f . (3.11)
Then, with
ˆ f,
F̂f = I + ∇û ∂t T̂f = ∂t uf = vf ,
the acceleration terms in the transformed momentum equation (3.10) get
(Jˆf ρ̂f ∂t v̂f , φ̂)Ω̂f + (Jˆf σ̂f F̂f−T , ∇
ˆ φ̂) = (Jˆρ̂f fˆ, φ̂) ,
Ω̂f Ω̂f

and the equations comes without the convective term. That is natural for the Lagrangian
viewpoint. Since the convective term gives rise to stability problems when dealing with
higher Reynolds numbers, this choice of T̂f looks beneficial.
Using (3.11) however does not yield a mapping between the domains of interest
T̂f : Ω̂f 6→ Ωf (t),
since particles x̂ ∈ Ω̂f can be transported out of the area of interest, see Figure 3.2 for an easy
example, where a Poiseuille flow is modeled using ALE transformation to the Lagrangian
coordinate system. The domain of interest is the fixed rectangle Ωf (t) = [0, X] × [0, Y ] for
all times. The ALE-domain however will get more and more distorted and does not reflect
Ωf (t).

Ω̂f = T̂f (t0 )Ω̂f T̂f (t1 )Ω̂f T̂f (t2 )Ω̂f

Figure 3.2: Using the “natural deformation” to construct the ALE-mapping.

The ALE-mapping should have the following properties:

38
 3.2 Construction of the ALE-mapping T̂f

1. It has to be a bijective mapping of the reference domain Ω̂f onto the domain of interest:

T̂f (t) : Ω̂f → Ωf (t), T̂f−1 (t) : Ωf (t) → Ω̂f .

2. Along the interface, it should coincide with the Lagrange-Euler transformation of the
structure domain:
T̂f (x̂, t) = T̂s (x̂, t) ∀x̂ ∈ Γ̂i .
Introducing a flow-deformation field ûf this means

ûf (x̂, t) = ûs (x̂, t) ∀x̂ ∈ Γ̂i .

3. The transformation should be regular

Jˆf = det(F̂f ) > 0,

or most preferable Jˆf ∼ 1.

We construct ûf as the solution of an artificial partial differential equation to be solved in

Ω̂f :
L(ûf ) = 0 in Ω̂f , ûf = ûs on Γ̂i , ûf = 0 on Γ̂/Γ̂i .
The differential-operator L should be such way, that the resulting solution ûf is as smooth
as possible. The first choice is to consider the Laplace-equation to get a harmonic extension
of ûs to ûf into the fluid domain

ˆ f = 0,
∆û ûf = ûs on Γ̂i , ûf = 0 on Γ̂/Γ̂i .

This additional equation can be formulated and discretized along with the Navier-Stokes
equations. Then, the ALE-mapping is given as a further solution variable in the system. We
note, that this variable ûf is not a deformation in the sense that it fits to the velocity, in
general ∂t ûf 6= v̂f .

One problem of using the harmonic extension to define the ALE mapping is the lack of
regularity. Often, the structure Ω̂s has sharp edges (like in our example, Figure 3.1). These
edges introduce reentrant corners in the flow domain Ω̂f . Here, the solution of the Laplace-
equation locally behaves like
π
|u(r)| ∼ r ω ,
where r is the distance to the corner and ω the outer angle. For ω > π (i.e. for reentrant cor-
ners), u 6∈ H 2 (Ω̂f ) and the derivatives tend to infinity. In ALE-context, the transformation
gradient F̂f looses its regularity.
Jˆf → ∞
Nevertheless, using the harmonic extension for defining the ALE mapping is the method of
choice. If the deformation of the structure (and the flow domain) is small, this method works
very reliable.

39
3 The Fluid-Structure Interaction Problem

For larger deformations, especially close to reentrant edges, using the biharmonic equation
is a good alternative, which leads to more regular solutions:

ˆ 2 ûf = 0 in Ω̂f ,
∆ ûf = ûs , ∂ˆn ûf = 0 on Γ̂/Γ̂i .

Here, the corner singularity is usually about one level better, yielding H 2 solutions even for
the limit case ω = 2π.

While the bi-harmonic equation yields excellent ALE mappings, it is very costly to solve.
One either has to use C 1 -conforming finite elements, or one has to use mixed methods for
its solution introducing an additional set:

ˆ f = ŵf ,
∆û ˆ ŵf = 0.
∆

For very large deformations, using the bi-harmonic equation is one of the view possibilities
to generate a robust ALE method.

A further possibility is to define the ALE-mapping using a elastic-structure equations

ˆ f, 1
F̂f (2µf Êf + λf tr(Êf )I) = 0 in Ω̂f , F̂f := I + ∇û Êf = (F̂fT F̂f − I).
2
One problem of using the structure equation for generating the ALE mapping is the tendency
towards incompressibility. This can lead to a degeneration of elements close to the interface:
a force in x-direction will lead to a compression in this direction which will be balanced
by a stretching in y-direction. This way, the transformed elements can get very anisotropic.
Using a negative Poisson ratio ν < 0 (which is not physical for most materials), this behavior
can be reversed. Compression in x-direction will lead to compression in y-direction and the
transformed elements stay isotropic, see Figure 3.3.

ν>0 ν<0

Figure 3.3: Material behavior with positive and negative Poisson ratio ν.

By choosing the “structure parameters” µf and λf large close to the interface, the “material”
gets stiffer here, yielding smaller transformation gradients F̂f and Jˆf ∼ 1.

40
 3.3 Fluid-structure interaction in ALE coordinates

3.3 Fluid-structure interaction in ALE coordinates

In this section we can finally state the monolithic formulation of the fluid-structure inter-
action problem in ALE-coordinates. The structure equation is formulated in Lagrangian
coordinates on Ω̂s , the Navier-Stokes equations in ALE coordinates on the domain Ω̂f . In
addition we here solve the Laplace equation for the definition of the ALE mapping. The
continuity conditions v̂s = v̂f and ûf = ûs on Γ̂i = Ω̂f ∩ Ω̂s are strongly enforced by including
them in the function space. We search for û ∈ H 1 (Ω̂; Γ̂D )d and v̂ ∈ H 1 (Ω̂; Γ̂D )d , where the
local quantities are defined by restriction v̂f := v̂|Ω̂f , v̂s := v̂|Ω̂s , ûs := û|Ω̂s and so on.

Since there is no pressure variable given in the solid domain, we harmonically extend the
fluid-pressure p̂f to Ω̂s . On all Ω̂ we denote the pressure field by p̂.

We note, that this extension of the pressure is an inconsistency. While the fluid’s pressure
is of low regularity p̂f ∈ L20 (Ω̂f ), the Laplace equation yields p̂s ∈ H 1 (Ω̂s ). This additional
regularity will feed back into the fluid domain, if the extension is not properly decoupled.

Since the ALE-mapping is defined in accordance to the Lagrange-Euler structure mapping

via T̂f := x̂ + ûf , we can define on all Ω:

T̂ := x̂ + û, ˆ
F̂ := I + ∇û, Jˆ := det(F̂ ).

In the structure domain, T̂ takes the place of the Lagrangian-Eulerian coordinate transfor-
mation, while in the fluid domain, T̂ has no physical meaning but serves as ALE mapping.

Problem 4 (ALE formulation of the Fluid Structure Interaction problem). Find v ∈ v D +

H01 (Ω; ΓD ), u ∈ uD + H01 (Ω; ΓD ) and p ∈ L20 (Ω̂), such that

(Jˆρ̂f (∂t v̂ + (F̂ −1 (v̂ − ∂t T̂ ) · ∇)v̂),

ˆ φ̂)Ω̂f + (ρ̂∂t v̂, φ̂)Ω̂s
ˆ φ̂) + (Jˆσ̂f F̂ −T , ∇
+(Jˆσ̂s F̂ −T , ∇ Ω̂s
ˆ φ̂) = (Jˆρ̂f fˆf , φ̂) + (ρ̂s fˆs , φ̂)
Ω̂f Ω̂f Ω̂f
c JˆF̂ −1 v̂), ξ)
ˆ ˆ ˆˆ
(div( Ω̂f + (αp ∇p̂, ∇ξ)Ω̂s = 0,
ˆ ∇
(αu ∇û, ˆ ψ̂) + (∂t û − v̂, ψ̂) = 0,
Ω̂f Ω̂s

where the stress tensors are given as

ˆ F̂ −1 + F̂ −T ∇v̂
σ̂f = −p̂I + ρ̂f νf {∇v̂ ˆ T },
σ̂s = Jˆ−1 F̂ (2µs Ê + λs tr(Ê)I)F̂ T .

In Problem 4, both momentum equations are multiplied by the same continuous test-function
φ̂ ∈ H01 (Ω; ΓD ). Thus, integration by parts in both subdomains yields the boundary term
on Γ̂i :
(n̂f · (Jˆσ̂s F̂ −T ), φ̂)Γ̂i + (n̂s · (Jˆσ̂f F̂ −T ), φ̂)Γ̂i = 0.
The interface condition is hence implicitly included in the formulation.

41
3 The Fluid-Structure Interaction Problem

Integration by parts also yields artificial boundary terms in the equations describing the
extension of the deformation and pressure:
ˆ ∇
(αu ∇û, ˆ ψ̂) + hn̂ · ∇û,
ˆ ψ̂) = (−αu ∆û, ˆ ψ̂i = 0,
Ω̂f Ω̂f Γ̂i

and
ˆ ∇
(αp ∇p̂, ˆ ψ̂) + hn̂ · ∇p̂,
ˆ ψ̂) = (−αp ∆p̂, ˆ ψ̂i = 0,
Ω̂s Ω̂s Γ̂i

On the interface, both equations include homogeneous Neumann boundary values. However,
we want to impose Dirichlet-boundary values to guarantee continuity ûf = ûs on Γ̂i . This
way, the equations are overdetermined and the extension will feedback to the other domain
distorting the results. In the discretization and solution schemes we carefully have to treat
this additional boundary terms.

42
4 The Finite Element Method for Continuum
Mechanics

In this chapter we describe the discretization of the equations with finite elements.

4.1 Basics

Let Ω ∈ Rd with d = 2, 3 be an open domain. The boundary Γ := ∂Ω of the domain is

split into Γ := ΓD ∪ ΓN , a Dirichlet- and Neumann part. We describe the finite element
discretization of an elliptic partial differential equation

L(u) = f in Ω, u = g D on ΓD , B(u) = g N on ΓN . (4.1)

Here, L is the elliptic differential operator, e.g. L := −∆. By g D we denote the prescribed
Dirichlet values on ΓD . B is some boundary operator and g N is the prescribed Neumann-
data. E.g., by
B(u) = ∂n u + αu, g N = 0,

a homogenous Robin boundary condition is indicated.

With the bilinear form a(·, ·) we write Equation (4.1) in the weak formulation

a(u, φ) = (f, φ) ∀φ ∈ V,

derived by multiplying (4.1) with a suitable test-function φ, by integrating over the domain
Ω and applying integration by parts to all second order terms. For the Laplace-equation
with Robin-boundary values on ΓR this bilinear-form is given by

a(u, φ) = (∇u, ∇φ) + hαu, φiΓN .

The weak formulation of (4.1) is to find u ∈ uD + H01 (Ω; ΓD ), such that

a(u, φ) = (f, φ) + hg D , φiΓN ∀φ ∈ H01 (Ω; ΓD ). (4.2)

Here, we understand uD ∈ H 1 (Ω) as an extension of the Dirichlet boundary data into the
domain.

43
4 The Finite Element Method for Continuum Mechanics

4.1.1 Finite Elements for elliptic partial differential equations

The finite element discretization aims at finding a discrete solution uh ∈ Vh of (4.2). If the
discrete function space Vh is a subspace Vh ⊂ H01 (Ω; ΓD ) we call the the resulting method a
conforming finite element method, otherwise a nonconforming finite element method .
For the definition of the discrete space Vh , we introduce:
Definition 4 (Finite Element Mesh). By Ωh = {K} we denote a triangulation of the domain
Ω into open elements K (quadrilaterals in two and hexahedrals in three dimensions):
[
Ω̄ = K̄.
K∈Ωh

We require the mesh to be structurally regular




∅ or

x ∈ Ω

a corner node
i h
∀K, K 0 ∈ Ωh : K 6= K‘ ⇒ K ∩ K 0 = ∅, K̄ ∩ K̄ 0 =
e ∈ Ωh
 a common edge


f ∈ Ωh a common face



and shape regular

π π dK
∀K ∈ Ωh : − cα < αK,i < + cα , ≤ cρ ,
2 2 ρK
where αK,i are the interior angles of K, dK the diameter of K and ρK the radius of the
largest inscribed circle. cα < π/2 and cρ ≥ 2 are constants.

By K̃ = (0, 1)d we denote the reference element. On K̃ we define the space of polynomials
up to degree r:
d
Q̃(r) = {span{(xγ11 · · · xγdd ), 0 ≤ γ1 , . . . , γd ≤ r}, P̃ (r) = {span{(xγ11 · · · xγdd ), 0 ≤
X
γi ≤ r}.
i=1

The first one is the space of polynomials up to degree r in every direction, the second space
has maximum degree r. Every element K can now be written as mapping of a simply
transformation T̃K : K̃ 7→ K.
Remark 1. The mapping T̃K is affine, if K is a parallelogram or parallelepiped (in three
dimensions). For general quadrilaterals and hexahedrals this mapping is nonlinear. If trian-
gular and tetrahedral meshes are used, every element is result of an affine mapping.

Next, we can define the polynomial spaces on every element K by transformation:

−1
Q(r) (K) = {φ : φ ◦ T̃K ∈ Q̃(r) }.

If T̃K ∈ Q̃(r) we call the finite element space Q(r) (K) isoparametric. For higher order finite
elements r > 1 using isoparametric finite elements is essential to approximate domains with
curved boundaries. Then, the edges and faces of the elements are not necessarily straight.

44
 4.1 Basics

On the domain Ω we can finally define continuous and discontinuous finite element spaces
(r) (r),dc
Vh and Vh as

(r)
Vh = {φ : Ω → R : φ ∈ C(Ω), φ|K ∈ Q(r) (K)},
(r),dc (r),dc
Vh = {φ : Ω → R : φ|K ∈ Q(r) (K)} or Vh = {φ : Ω → R : φ|K ∈ P (r) (K)}

(r)
We introduce the Lagrangian nodal basis of Vh in a similar fashion. By {x̃i ∈ K̃, i =
1, . . . , (r + 1)d } we denote equally spaced nodal points of the reference element. Then, by
{φ̃i , i = 1, . . . , (r + 1)d } we denote a basis of Q̃(r) on K̃ with the property

φ̃i (xj ) = δij , i, j = 1, . . . , (r + 1)d .

(r)
This basis is mapped onto Q(r) (K) and combined to a continuous basis of Vh = span{φih , i =
(r)
1, . . . , N }. For vh ∈ Vh we write
N
X
vh = v ih φih .
i=1

Then, the finite element solution of (4.2) is given by:

N
X 
v h ∈ RN : v jh φjh φih = (f, φih ),


a i = 1, . . . , N.
j=1

By introducing the system matrix Ah ∈ RN ×N and the load vector b h ∈ RN :

(Ah )ij = a(φjh , φih ), i, j = 1, . . . , N, b ih = (f, φih ), i = 1, . . . , N.

The solution v h is hence given as the solution of a system of linear equations

Ah v h = b h .

4.1.2 Discretization of time-depending partial differential equations

In this section we describe the discretization of non-stationary partial differential equations.

First, we have parabolic initial value problems in mind. Let I := [0, T ] we a time-interval.
In weak formulation we search u ∈ uD + X0

u(0) = u0 , (ut , φ) + a(u, φ) = (f, φ) ∀φ ∈ H01 (Ω; ΓD ), u(0) = u0 ,

where uD ∈ X are extensions of the Dirichlet values into the domain and u0 ∈ uD (0)+H01 (Ω)
is the initial value.

We discuss the discretization with Rothe’s method. First, we discretize in time by introducing
discrete time-steps

0 = t0 < t1 < . . . , TM = T, Im = (tm−1 , tm ], km := tm − tm−1 ,

45
4 The Finite Element Method for Continuum Mechanics

and by setting for m = 0, . . . , M : um := u(tm ). For approximation of the time-derivative we

use the θ-method with a parameter θ ∈ [0, 1]:

(um − um−1 , φ) + km θa(um , φ) + km (1 − θ)a(um−1 , φ) = km (Fθm , φ), m = 1, . . . , M,

where u0 is the given initial value and

Fθm := θf (tm ) + (1 − θ)f (tm−1 ).

For θ = 1 this method is the implicit backward Euler method, for θ = 0 the fully explicit
forward Euler method. These two methods are first order accurate in time. The explicit
forward Euler method is unstable and not usable for stiff problems, e.g. if a(·, ·) represents
the Laplace-equation. The backward Euler method is unconditionally stable. For θ = 12 the
resulting method is the second order accurate Crank-Nicolson scheme:

km km
(um , φ) + a(um , φ) = km (F1/2
m
, φ) + (um−1 , φ) − a(um−1 , φ)
2 2
m = 1, . . . , M. (4.3)

Every time-step can then be solved separately and is regarded as a stationary partial differ-
ential equation:
(u, φ) + θm km a(u, φ) = (F, φ),
where the right hand side and all the explicit parts of the time-discretization are gathered
in (F, φ). These “stationary” equations can be solved as indicated in the previous section.
It is possible to use different finite element meshes Ωm
h for different time steps tm−1 → tm .

The elastic structure equations are wave-equations with second order temporal derivatives:

u(0) = u0 , ∂t u(0) = v(0) = v 0 , (utt , φ) + a(u, φ) = (f, φ) ∀φ ∈ H01 (Ω; ΓD ).

Instead of using special time-discretization schemes like the Newmark Scheme, we always
write these equations as a set of two equations:
(
0 0 (vt , φ) + a(u, φ) = (f, φ) ∀φ ∈ H01 (Ω; ΓD )
u(0) = u , v(0) = v , .
(ut , φ) = (v, φ) ∀ψ ∈ H01 (Ω; ΓD )

Then, we can apply simple time-stepping methods like the θ-scheme on this system of equa-
tions:
(v m , φ) + θkm a(v m , φ) = (v m−1 , φ) + (Fθm , φ) − (1 − θ)km a(v m−1 , φ),
(um , ψ) − θkm (v m , ψ) = (um−1 , ψ) + (1 − θ)km (v m−1 , ψ).

4.1.3 Discretization of nonlinear problems

Next, we assume that the differential operator contains nonlinearities. We write the weak
formulation of the equations with a bilinear-form a(φ)(φ), which is linear in the second and
possibly nonlinear in the first argument:

u ∈ uD + H01 (Ω; ΓD ) : a(u)(φ) = (f, φ) ∀φ ∈ H01 (Ω).

46
 4.1 Basics

Figure 4.1: Exact and numerical (dashed lines) solution to the convection-diffusion model
problem.

For discretization, we first linearize this equation by using a Newton-iteration to find the
i→∞
solution u(i) −−−→ u ∈ H01 (Ω; ΓD ):

Algorithm 1 (Newton-Iteration). Let u(0) ∈ uD + H01 (Ω; ΓD ) be an initial guess of the

solution, e.g. u(0) = uD . For i ≥ 1 iterate

1. Assemble Newton residual

(r(i) , φ) = (f, φ) − a(u(i) )(φ) ∀φ ∈ H01 (Ω).

2. Check residual
kr(i) k ≤ TOL ⇒ STOP!

3. Solve for w ∈ H01 (Ω; ΓD ):

a0 (u(i−1) )(w, φ) = (r(i) , φ) ∀φ ∈ H01 (Ω; ΓD ).

4. Update
u(i) := u(i−1) + w(i) .

In Step 3 of the algorithm a linear system has to be solved. By a0 (u)(w, φ) we denote the
directional derivative of a(·)(·) at point u in direction of w. It is defined by

d
a0 (u)(w, φ) := a(u + sw)(φ),
ds
and it is linear in the second and third argument. These linear Newton update equations
can be solved as described the two previous sections. Quadratic convergence of the Newton
method is guaranteed, if a(·)(·) is two times differentiable and if the initial guess is good
enough.

47
4 The Finite Element Method for Continuum Mechanics

4.1.4 Discretization of transport-dominant problems

In this section we discuss the finite element discretization of transport diffusion equations of
the type
a(u, φ) = (∇u, ∇φ) + (β · ∇u, φ), (4.4)
where  > 0 is the diffusion constant and β : Ω → Rd is a given transport field. For small
diffusion values   1, the finite element method is known to yield numerically unstable
solution. This is easily seen by analyzing the 1d-example

−u00 + u0 = 0 in (0, 1), u(0) = 1, u(1) = 0, (4.5)

with the exact solution x 1

e − e
u(x) = 1 .
1 − e
This solution behaves like u(x) ≈ 1 for x < 1 −  and quickly turns to zero in a small layer
of size O(). A finite element solution of this problem will yield over- and undershoots if the
mesh-size is too large (to be precise, for h > 2), see Figure 4.1.
A stable solution of transport-diffusion equations can either be reached by using discontinu-
ous finite elements for discretization or by adding stabilization terms to Equations 4.4. The
aim of these additional stabilization terms is to gain control over the derivative in transport
direction kβ · ∇uk. The most common stabilization technique is the streamline diffusion
method. Here, the weak formulation is build with the special test-function:
X X
(−∆u+β·∇u, φ+δh β·∇φ) = (∇u, ∇φ)+(β·∇u, φ)+ (−∆u, δh β·∇φ)K + (β·∇u, δh β·∇φ)K .
K∈Ωh K∈Ωh

For obtaining a consistent discretization, the right hand side of the equation is tested with
the same adapted test-function (f, φ + δh β · ∇φ). The stabilization parameter δh is defined
element-wise by
 h2 h 
K
δh K = δ0 min K ,

,
 |β|∞
with a constant δ0 ≈ 12 . This stabilization scheme is able to find stable solutions for arbi-
trary  > 0. One problem of the streamline diffusion method is that the scheme has to be
implemented in a consistent fashion, multiplying every term in the equation with δh β · ∇φ.
This way, not-natural terms like (∆u, δh β · ∇φ) (which is of third order) appear. We will
refer to this problem when discussing the stabilization of the Navier-Stokes equations and
the full fluid-structure interaction problem.
One alternative stabilization technique is the Local Projection Method (LPS). Again, stability
is gained by adding stabilization terms to the bilinear-form:
X
(∇u, ∇φ) + (β · ∇u, φ) + (δh β · ∇πh u, β · ∇πh φ)K = (f, φ),
K∈Ωh

where the element-wise parameter δh is defined as above. The operator πh : Vh → Vh is

the so called fluctuation operator. The method aims at penalizing only the oscillations (see

48
 4.2 Finite Elements for the incompressible Navier-Stokes Equations

Figure 4.1) on the finest mesh level, i.e. we only want to penalize the derivatives in β-
direction on the finest mesh scale. For this, let Ṽh be a coarser finite element space than
(r) (r)
Vh , e.g. the space V2h of same degree but on a coarser mesh. Then, the fluctuation
operator πh is defined by
πh φh := φh − i2h φh ,
as the difference between the function φh itself and its interpolation to the next coarse
mesh. This LPS-method yields optimal convergence results (like the streamline diffusion
method). It however is not able to stabilize the limit-case  → 0. The great benefit of
this method is its diagonal character: the stabilization term only acts on the value to be
stabilized (the transport-derivative) and it does not introduce additional artificial terms to
the equation. Especially for treating complex problems like fluid-structure interaction, this
is a very valuable property.

4.2 Finite Elements for the incompressible Navier-Stokes

Equations

4.2.1 Finite Element discretization of the Stokes Equations

We first discuss the finite element discretization of the stationary Stokes equation in a domain
Ω ⊂ Rd : find v ∈ v 0 + V0 and p ∈ L, such that

(∇v, ∇φ) − (p, ∇ · φ) = (f, φ), (∇ · v, ξ) = 0, ∀φ ∈ V0 , ξ ∈ L.

The function spaces are

Z
V0 := H01 (Ω; ΓD )d , L := L20 (Ω) 2
:= {ξ ∈ L (Ω) : ξdx = 0}.
Ω

Since the Stokes equations determine the pressure only up to a constant, these are removed
from the space L. If the boundary ∂Ω contains a Neumann part ΓN , integration by parts
yields:
(∇v, ∇φ) − (p, ∇ · φ) = (−∆v + ∇p, φ) + h∂n v − n · p, ξiΓN
This boundary condition
h∂n v − n · p, ξiΓN = 0,
is called the do nothing or outflow boundary condition. The name “do nothing” arises from
the property, that some prototypical flow fields are not tempered with by this condition. This
holds e.g. for the Poiseuille flow or the Couette flow. The do nothing condition includes
a further “hidden boundary condition”. Let ΓN be a connected Neumann boundary with
normal vector n and tangential vector τ . Since v is divergence-free, it holds with vn := n · v
and vτ := τ · v
∇ · v = ∂n vn + ∂τ vτ ⇒ ∂n vn = −∂τ vτ .
Then, with the test function ξ = 1 on ΓN :
Z Z Z Z
n · p ds = ∂n v ds = ∂n vn ds = − ∂τ vτ ds = 0,
ΓN ΓN ΓN ΓN

49
4 The Finite Element Method for Continuum Mechanics

since vτ = 0 on the adjacent Dirichlet boundaries. On every connected Neumann boundary

segment, the pressure has zero average value. If Neumann boundary is present, we can thus
search the pressure in the full space p ∈ L := L2 (Ω).

Stable Finite Element Pairs for the Stokes Equations

For a consistent discretization of the Stokes equations we need discrete subspaces

Vh ⊂ V0 , Lh ⊂ L.

To yield a stable discretization this finite element pair Vh ×Lh has to fulfill the BB-condition:

Lemma 6 (BB-condition). Let Vh ⊂ V0 and Lh ⊂ L be finite element spaces fulfilling the

condition
(ξh , ∇ · φh )
γkξh k ≤ max , ∀ξh ∈ Lh , (4.6)
φh ∈Vh k∇φh k
with a constant γ > 0 (independent of h). Then, there exists a unique solution vh ∈ Vh and
ph ∈ Lh of

(∇vh , ∇φh ) − (ph , ∇ · φh ) = (f, φh ), (∇ · vh , ξh ) = 0 ∀φh ∈ V0 , ξh ∈ Lh .

This condition (4.6) is also called the discrete inf-sup condition.

(r) (r)
The most simple equal order finite element pairs, e.g. Vh = [Vh ]d for the velocity Lh = Vh
for the pressure are known not to fulfill the BB-condition. A common inf-sup-stable Stokes-
elements is the Taylor-Hood Element, which is given by a piecewise bi-quadratic velocity and
a bi-linear pressure:
(2) (1)
Vh := Vh , Lh := Vh .
(r+1) (r)
On quadrilateral (and hexahedral) meshes, the finite elements pairs Vh − Vh are all
stable for r ≥ 1.

Remark 2. If the right hand side is f ∈ H 1 (Ω) and the boundary of the domain is “smooth”,
the following a priori estimate holds for the approximation with the Taylor-Hood element:

k∇(v − vh )k + kp − ph k ≤ Ch2 {k∇3 uk + k∇2 pk}.

Stabilized finite elements for the Stokes equations

Sometimes it is preferable to use simple finite element pairs that are not inf-sup-stable as
the equal-order elements Q1 − Q1 or Q2 − Q2 . This can be for ease of implementation.
Stability is reached by adding certain stabilization terms to the equation. The most well-
known technique is the Pressure Stabilized Petrov Galerkin method (PSPG), similar to the

50
 4.2 Finite Elements for the incompressible Navier-Stokes Equations

streamline upwind method. The weak formulation of the Stokes equations is derived by
using a modified test-function pair (φh + αh ∇ξh , ξh ):

(∇vh , ∇φh ) − (ph , ∇ · φh ) = (f, φh ) ∀φh ∈ Vh .

X X
(∇ · vh , ξh ) + (αh ∇ph , ∇ξh )K + (αh (f − ∆vh ), ∇ξh )K = 0 ∀ξh ∈ Lh .
K∈Ωh K∈Ωh

The stabilization parameter αh is chosen element-wise as

1
αh = α0 h2K , α0 ≈ .
2
The drawbacks of this stabilization technique are similar to those of the streamline diffusion
method. Additional coupling are introduced to the system. In the error analysis the term
kf − ∆vh k appears whose implication is not completely understood.
As an alternative technique we again introduce the Local Projection Stabilization, which
reaches stability by adding local fluctuations of the pressure-gradient:

(∇vh , ∇φh ) − (ph , ∇ · φh ) = (f, φh ) ∀φh ∈ Vh .

X
(∇ · vh , ξh ) + (αh ∇πh ph , ∇πh ξh )K = 0 ∀ξh ∈ Lh .
K∈Ωh

The stabilization parameter αh is defined as for the PSPG method, the operator πh again
measures the fluctuation with respect to a coarse functions space Ṽh . A common projection
space for the LPS-method is a finite element space of lower degree on the same mesh:
(r) (r−1)
Vh = Vh , Ṽh = Vh : πh := id − ir−1 ,

which yields optimal approximation results (since the natural regularity of the pressure is
one order lower than the velocities regularity).
Remark 3. The function space pairing Qr − Qr−1 yields a stable method for transport-
stabilization. We however do not reach optimal order of convergence. For transport stabi-
lization, we need to project to an equal-order space on a coarse mesh Qr − Qr2h .

4.2.2 Finite Element discretization of the Navier-Stokes Equations

The incompressible Navier-Stokes equations on the time-interval I = [0, T ] in weak formu-

lation are given by

(ρ∂t (v + v · ∇v), φ) + (ρν{∇v + ∇v T }, φ) − (p, ∇ · φ) = (ρf, φ),

(∇ · v, ξ) = 0.

For time discretization difference schemes like presented in Section 4.1.2 are used. When
treating flow problems, further demands are made on the time-stepping scheme:
Stability The difference scheme should at least be A-stable. Stronger stability results like
Strong A-Stability are desirable so that oscillations in the data or numerical errors are
damped.

51
4 The Finite Element Method for Continuum Mechanics

Dissipation The scheme should be energy preserving: physical oscillation may not be damped.
First order schemes like backward Euler are not competitive. Further, the backward Euler
scheme is very dissipative, physical oscillations are damped too much. The easy and well-
established Crank-Nicolson scheme is second order, A-stable and exactly energy-preserving.
It however lacks the ability to damp unwanted oscillations. A remedy is given by the im-
plicitly shifted Crank-Nicolson scheme, a θ-scheme with the value θ = 21 + k. This method
has slightly better stability properties and still has low dissipation.
A further method, which combines most preferable properties is the Fractional-Step-Theta
(FST) scheme. Here, one time-step tn−1 → tn is assembled by three sub-steps of the θ-
scheme, using different θ-values and different time-steps for every sub-step:
θ 1−θ θ
tn−1 →
− tn−1+k0 −−→ tn−k0 →
− tn ,

The step-size is k 0 for the first and the last and k − 2k 0 for the middle step. This method is
of second order for
1
 
0
k := 1 − √ k,
2
and strongly A-stable for all values θ ∈ ( 12 , 1]. Further, the FST-scheme has very low
dissipation, making it well-suited for flow problems.
In the incompressible Navier-Stokes equations, the pressure is regarded as a point-wise (in
time) Lagrange-multiplier to guarantee the incompressibility. We thus always treat the
pressure-part fully implicit and one step of the θ-scheme has the form (where v̄ is the velocity
in the old time-step):

(ρv, φ) + kθ(ρv · ∇v, φ) + kθ(ρν{∇v + ∇v T }, ∇φ) − k(p, ∇ · φ) = F (k, θ, v̄)

(4.7)
(∇ · v, ξ) = 0,

where by F (k, θ, v̄) we denote the right hand side containing the data and all explicit terms
evaluated at the old time-step:

F (k, θ, v̄) := (ρv̄, φ) + k(θf + (1 − θ)f¯, φ) − k(1 − θ)(ρv̄ · ∇v̄, φ) − k(1 − θ)(ρν{∇v̄ + ∇v̄ T }, ∇φ).

After time-discretization, the resulting nonlinear equation is solved with Newton’s method.
Let u = (v, p) be the approximation from the last Newton step. Then, the update (w, q) is
given by the linear problem:

(ρw, φ) + kθ(ρ{v · ∇w + w · ∇v}, φ) + kθ(ρν{∇w + ∇wT }, ∇φ) − k(q, ∇ · φ) = F (k, θ, v̄)

−kθ(ρv · ∇v, φ) − kθ(ρν{∇v + ∇v T }, ∇φ) + k(p, ∇ · φ)
(∇ · w, ξ) = −(∇ · v, ξ),
(4.8)

We write this equation in a compact notation by

A0 ({v, p})({w, q}, {φ, ξ}) = F (k, θ, v̄, v, q). (4.9)

52
 4.2 Finite Elements for the incompressible Navier-Stokes Equations

When using higher equal order finite elements, and for problems with dominant convection
(i.e. for high Reynolds numbers), the Galerkin discretization of this equation is not stable
and further stabilization terms are necessary. Let Qh × Vh be the finite element pair. Then,
stability could by reached by combining the PSPG and the SUPG-method with the test-
function:
φ̃ := φ + δw · ∇φ + α∇ξ.
To get a consistent scheme, this test-function is used on both sides of (4.8) and thus also
acts on the explicit terms and in such a way is coupling the old with the new time velocity:
1
(v − v̄, φ + δw · ∇φ + αξ) . . .
k

As an alternative, we use LPS-stabilization for both, convective terms and inf-sup condition.
Instead of (4.9) we solve

A0 ({v, p})({w, q}, {φ, ξ}) + Slps ({v, p})({w, q}, {φ, ξ}) = F (k, θ, v̄, v, q),

where the stabilization form is defined by

X n o
Slps ({v, p})({w, q}, {φ, ξ}) = δK (ρv · ∇πw, v · ∇πφ)K + αK (1/ρ∇πq, ∇πξ)K ,
K∈Ωh

with parameters defined locally on every element K as:

!−1
ν |v|K;∞
δK = αK = α0 2 + .
hK hK

4.2.3 Finite Elements for Navier-Stokes in ALE-formulation

In ALE formulation, the Navier-Stokes equations are strongly nonlinear. We repeat Formu-
lation (3.10) using T̂f := x̂ + û as used in FSI context:

Jˆf ρ̂f (∂t v̂ + (F̂f−1 (v̂ − ∂t û) · ∇)v̂),

ˆ + (Jˆf σ̂f F̂f−T , ∇
ˆ φ̂) = (Jˆf ρ̂f fˆ, φ̂)


φ̂ Ω̂f Ω̂f Ω̂f
d Jˆ F̂ −1 v̂), ξ)
(div( ˆ
f Ω̂f = 0.

For discretization we have to follow the presented course: discretize in time, linearize, dis-
cretize in space. Here we have to face an additional curiosity since the time-derivative
appears in a nonlinear form. Further, when the deformation û is defined implicitly, a second
time derivative appears. It is not obvious, how to derive a proper, in particular a second
order and stable, time-stepping formulation.

A rigorous derivation is done, if Galerkin methods are used for time-discretization. We

analyze the following model-problem:

(∂t u, φ) + (∇u, ∇φ) = (f, φ) on [0, T ] × Ω, u(u) = u0 .

53
4 The Finite Element Method for Continuum Mechanics

We can write this equation in a space-time Galerkin formulation by integrating over I = [0, T ]
and by multiplying with suitable test-functions:
Z n o Z
(∂t u, φ) + (∇u, ∇φ) dt = (f, φ)dt ∀φ ∈ X.
I I

The solution u is found in the space:

u ∈ u0 + X, X = {v : v ∈ L2 (I; H 1 (Ω)), ∂t v ∈ L2 (I; H 1 (Ω)∗ )}.

¯ H 1 (Ω)), thus, these functions are continuous in time.
For this space it holds X ,→ C(I;
We discretize this equation in time by introducing discrete-time steps ti , i = 0, . . . , M and
by defining sub-intervals

0 = t0 < t1 < · · · < tM = T, Im := [tm−1 − tm ], km := tm − tm−1 .

We use a Petrov-Galerkin approach with continuous trial and discontinuous test-functions:

¯ H 1 (Ω)) : v|I ∈ P m (Im ; H 1 (Ω))},
Xktrial = {v ∈ C(I; Xktest = {v ∈ L(I; H 1 (Ω)) : v|Im ∈ H 1 (Ω)},
m

where P m (Im ) is a polynomial space with dimension 2. We write uk ∈ Xktrial as:

M
X
uk = um Φm , um ∈ H 1 (Ω).
m=0

Then, the time-discretized model-problem is given by

M Z
X n o M
X
(∂t uk , φm ) + (∇uk , ∇φm ) dt = (f, φm )dt, ∀φ1 , . . . , φM ∈ H 1 (Ω).
m=1 Im m=1

Since the test-functions are discontinuous, the system is decoupled and can be written as a
time-stepping scheme, with global coupling only due to the continuity of the trial-space Xk .
We now define the space P m (Im ) and indicate the two basis functions for a parameter
θ ∈ (0, 1):

tm − t  1  t − tm−1  1  (t − tm−1 )(tm − t)

Φ0m (t) := − θ− φθ (t), Φ1m (t) := + θ− φθ (t), φθ (t) := 6 2
.
km 2 km 2 km

For Φ0m and Φ1m it holds:

Z Z Z Z
Φ0m dt = (1 − θ)km , Φ1m dt = (θ)km , ∂t Φ1m dt = −1, ∂t Φ1m dt = 1.
Im Im Im Im

In every time-step tm−1 → tm , the solution is known at the left boundary uk (tm−1 ) = um−1
and we need to solve:
Z
(−um−1 + um , φ) + (1 − θ)km (∇um−1 , ∇φ) + θkm (∇um , ∇φ) = (f, φ)dt ∀φ ∈ H 1 (Ω).
Im

Up to an integration error in the right hand side, this scheme is exactly the θ-scheme.

54
 4.3 Finite Elements for elastic structure equations

We can now use this approach to derive a time-stepping scheme for more complex situations
as arising in fluid-structure interaction problems: we discuss the model-problem

(h∂t u, φ) + (∇u, ∇φ) = (f, φ) on [0, T ] × Ω, u(u) = u0 ,

where h(t, x) is some function, also to be discretized via

M
X
hk = hm Φm , hm ∈ H 1 (Ω).
m=0

The Galerkin discretization of the first term with the time derivative now gets:
Z Z  
(hk ∂t uk , φ)dt = (hm−1 Φm−1 + hm Φm )∂t (um−1 Φm−1 + um Φm ), φ dt
Im Im
 hm−1 + hm 
= (um − um−1 ), φ .
2

The full time-stepping scheme for the model problem then reads:

 hm−1 + hm  Z
m m−1 m−1 m
(u −u ), φ +(1−θ)km (∇u , ∇φ)+θkm (∇u , ∇φ) = (f, φ)dt ∀φ ∈ H 1 (Ω).
2 Im

Here, the solution at old and new time-step are always coupled. In the full Navier-Stokes
equations in ALE formulation, the second time-dependent term is discretized likewise.

4.3 Finite Elements for elastic structure equations

In weak formulation, the equations describing the deformation of an elastic structure are
given by

(ρ̂∂t v̂, φ̂) + (Jˆσ̂s F̂ −T , ∇

ˆ φ̂) = (ρ̂fˆs , φ̂)
(∂t û − v̂, ψ̂) = 0,

with the stress-tensor

1
σ̂s = Jˆ−1 F̂ (2µÊ + λs tr(Ê)I)F̂ T , Ê := (F̂ T F̂ − I).
2

Even tho this equation is very nonlinear, the discretization is straight-forward and follows
the presented route: discretize in time (by a Galerkin or time-stepping method), linearize
and discretize in space by finite elements. In Lagrangian formulation there are no additional
stabilization terms necessary, since the equation does not include transport-terms. When
using incompressible material laws however, the function spaces need to fulfill an inf-sup
condition and “pressure-stabilization” may be required for equal-order finite elements.

55
4 The Finite Element Method for Continuum Mechanics

4.4 Finite Elements for the fluid-structure interaction problem

The coupled fluid-structure interaction problem as derived in Chapter 3 reads (compare to

Problem 4):

(Jˆρ̂f (∂t v̂ + (F̂ −1 (v̂ − ∂t T̂ ) · ∇)v̂),

ˆ φ̂)Ω̂f + (ρ̂∂t v̂, φ̂)Ω̂s
+(Jˆσ̂s F̂ −T , ∇
ˆ φ̂) + (Jˆσ̂f F̂ −T , ∇
Ω̂s
ˆ φ̂) = (Jˆρ̂f fˆf , φ̂) + (ρ̂s fˆs , φ̂)
Ω̂f Ω̂f Ω̂f
d JˆF̂ −1 v̂), ξ)
ˆ ˆ ˆˆ
(div( Ω̂f + (αp ∇p̂, ∇ξ)Ω̂s = 0,
ˆ ∇
(αu ∇û, ˆ ψ̂) + (∂t û − v̂, ψ̂) = 0,
Ω̂f Ω̂s

where the stress tensors are given as

ˆ F̂ −1 + F̂ −T ∇v̂
σ̂f = −p̂I + ρ̂f νf {∇v̂ ˆ T },
σ̂s = Jˆ−1 F̂ (2µs Ê + λs tr(Ê)I)F̂ T .

The finite element discretization has to take care of several difficulties:

• Strong nonlinearities
• Time-dependent problem including “non-standard terms”
• Stabilization is necessary
• Different type of equation in different subdomains
• Artificial extension of variables (“uf ” and “ps ”)
We already have dealt with the first three issues. In every subdomain Ω̂f and Ω̂s can be
assumed to be regular, if the data is regular and if the boundary of the domain (and the
shape of the interface Γ̂i ) is regular. Across the interface however, velocity v̂ and deformation
û are continuous, but higher regularity cannot be postulated. If the interface Γ̂i = ∂ Ω̂f ∩∂ Ω̂s
cuts through a mesh-element K ∈ Ωh , we cannot expect a good approximation property:
ˆ − ih û)kK ≤ ch?K k∇
k∇(û ˆ 1+? ûkK .

This additional approximation error is only local along the interface. This region however
plays a decisive role for the overall dynamics of the fluid-structure system. Whenever possi-
ble, the mesh should thus be aligned with the interface Γ̂i !
The coupled formulation includes two parameter αp and αu controlling the harmonic exten-
sion of deformation and pressure to the other domain. The choice of these parameter will
influence the approximation of the discrete schemes as well as the properties of the algebraic
system and thus the convergence of the linear solvers. This will be discussed at some further
point.
Both values, pressure and deformation are extended with the Laplace-equation. On the in-
terface Γ̂i continuity is postulated. Integration by parts yields (exemplarily for the pressure):
d JˆF̂ −1 v̂), ξ)
0 = (div( ˆ Ω + (αp ∇p̂,
ˆ ∇ d JˆF̂ −1 v̂), ξ)
ˆ ψ̂) = (div( ˆ ψ̂) + hn̂s · ∇p̂,
ˆ Ω + (−αp ∆p̂, ˆ ψ̂i .
f Ω̂s f Ω̂s Γ̂i

56
4.5 Some mathematical theory for fluid-structure interaction problems (noch ein wenig experimentell und mit Vo

On the Interface, this formulation includes a Neumann boundary condition:

ˆ ψ̂i = 0.
hn̂s · ∇p̂, Γ̂i

This boundary condition is in addition to the already present Dirichlet condition given
by the continuity postulate “p̂s = p̂f ” and will lead to a spurious feedback to the flow-
domain. We can remove this feedback by adding this unwanted interface-boundary term to
the formulation. A better way to work around this feedback is to decouple both problems
and to enforce the Dirichlet-values strongly by removing the additional basis-functions:

Ω̂f Ω̂s
Γ̂i

Figure 4.2: Delete basis functions on interface seen from the solid-side

The extension of the pressure to the structure is only a numerical necessity since we do not
need a pressure in the structure domain at all. It would be more appropriate to find the
pressure in the space p̂f ∈ L20 (Ω̂f ) and to write the divergence equation

d JˆF̂ −1 ), ξ)
ˆ 2
(div( Ω̂f = 0 ∀ξ ∈ L0 (Ω̂f ),

and not to have an equation in the structure domain at all. Further, it would be more
appropriate to not combine the extension of the deformation to the flow domain and the
condition ∂t û = v̂ in one equation since they are independent. From a theoretical point of
view one should use two different test-functions, one defined only on Ω̂f and the other only
on Ω̂s :

ˆ ∇
(αu ∇û, ˆ ψ̂ f ) + (∂t û − v̂, ψ̂ s ) ∀ψ̂ f ∈ H01 (Ω̂f ; ΓD ), ∀ψ̂ s ∈ H01 (Ω̂s ; ΓD ).
Ω̂f Ω̂s

4.5 Some mathematical theory for fluid-structure interaction

problems (noch ein wenig experimentell und mit Vorsicht zu
geniessen)

In this section we will analyze a simplified model for fluid-structure interaction:

57
4 The Finite Element Method for Continuum Mechanics

• We only consider the stationary case:


∂t v = ∂t u = 0, v = 0.

Γs

• The elasticity model is supposed to be linear:

1
E ≈ {∇u + ∇uT },
2

• We do not consider geometric nonlinearities T̂s = I and we assume the ALE-transformation

to be the identity T̂f = I:
F = I, J = 0.

• We neglect the transposed entries in the stress-tensors, and we neglect the divergence-
condition of the fluid problem:

−∇ · σf = ν∆v, −∇ · σs = µ∆u.

What follows is a linear set of partial differential equations:

) )
−ν∆v = f −µ∆u = f
in Ωf , in Ωs , ν∂n v = µ∂n u on Γi := ∂Ωf ∩ ∂Ωs .
−α∆u = 0 v =0

Since we omit the divergence condition and the transposed parts in the tensors, no couplings
between the different velocity and deformation components remain. We can thus consider
velocity and deformation to be scalar.

With U := (v, u) and Φ := (φ, ψ s , ψ v ) we introduce the bilinear-form

a(U, Φ) := ν(∇v, ∇φ)f + µ(∇u, ∇φ)s + α(∇u, ∇ψ f )f − αh∂n u, ψ f is − (v, ψ s )s ,

where

U = (v, u) ∈ H01 (Ω; ΓD )×H01 (Ω; ΓD ), Φ = (φ, ψ s , ψ v ) ∈ H01 (Ω; ΓD )×H01 (Ωf ; ΓD )×H01 (Ωs ; ΓD ).

The following remark is easily seen:

Lemma 7 (Continuity). With the triple-norm

 1
|||Φ||| := k∇vk2f + k∇uk2s + k∇uk2f + kvk2s 2
,

the bilinear-form a(·, ·) is continuous:

|a(U )(Φ)| ≤ cc |||U ||| |||Φ||| ∀U, Φ ∈ H01 (Ω; ΓD ) × H01 (Ω; ΓD ).

58
4.5 Some mathematical theory for fluid-structure interaction problems (noch ein wenig experimentell und mit Vo

Let Ψ = (ψ, −ψ) with ψ = 0 on Ωf . Then, with Poincaré inequality

a(Ψ, Ψ) = −µk∇ψk2s + kψk2 ≤ (c2p − µ)k∇vkss ,
which is negative for large values of µ. We thus cannot get ellipticity of the bilinear-form.
We can however formulate the following stability result:
Lemma 8 (Stability of the model equation). The following inf-sup condition holds with
some γ > 0:
a(U, Φ) − µ(∇v, ∇ψ)s
sup ≥ γ (|||U ||| + k∇vks ) ∀U ∈ X.
U ∈X |||Φ|||

Proof: Let U = (v, u) be arbitrary. Diagonal testing with Φ1 := (v, u) yields:

a(U, Φ1 ) = νk∇vk2f + µ(∇u, ∇v)s + αk∇uk2f − (v, u)s − µ(∇v, ∇u)s
 cp
≥ νk∇vk2f + αk∇uk2f − kvk2s − k∇uk2s ,
2 2
where  > 0 is arbitrary and cp is the constant of the Poincaré inequality. Then, with
Φ2 := (u, 0)
a(U, Φ2 ) = ν(∇v, ∇u)f + µk∇uk2s ,
and Φ3 := (0, −v)
a(U, Φ3 ) = −α(∇u, ∇v)f + kvk2s + k∇vk2s ,
we can test with the combined test-function
!
ν v+u
Φ4 := Φ1 + Φ2 + Φ3 =
α u − αν v
to get
cp ν  ν
   
a(U, Φ4 ) = νk∇vk2f + µ− k∇uk2s + αk∇uk2f + kvk2s − + k∇vk2s .
2 α 2 α
For  = cp /ν we have
µ 2µν − αcp ν
 
a(U, Φ4 ) − µ(∇v, ∇ψ)s = νk∇vk2f + k∇uk2s + αk∇uk2f + kvk2s + k∇vk2s .
2 2µα α
2µν
For α < cp we get
µ cp cp 
  
a(U, Φ4 ) − µ(∇v, ∇ψ)s ≥ ν + + + |||U |||2 + k∇vk2s . (4.10)
2 2µν 2µ
Finally, we need to estimate the triple-norm of this test-function Φ4 :
ν ν
|||Φ4 |||2 = νk∇(v + u)k2f + kv + uk2s + µk∇(u − v)k2s + αk∇(u − v)k2f
α α
2ν 2ν
≤ 2νk∇vkf + 2νk∇ukf + 2kvks + 2kukf + 2µk∇uks + k∇vk2s + 2αk∇uk2f + k∇vk2f
2 2 2 2 2
α α
2ν
 
≤ 2ν + k∇vk2f + (2ν + 2cp + 2α) k∇uk2f + 2k∇uk2f + 2µk∇uk2s
α
!
4µν cp  
≤ 4ν + 2µ + 2cp + +2+ |||U |||2 + k∇vk2s
cp µ
(4.11)

59
4 The Finite Element Method for Continuum Mechanics

Thus, combining the constants c1 of (4.10) and c2 of (4.11)

a(U, Φ) − µ(∇v, ∇ψ)s a(U, Φ4 ) − µ(∇v, ∇ψ4 )s |||U |||2 + k∇vk2s c1

sup ≥ ≥ c1 ≥ |||U |||,
Φ |||Φ||| |||Φ4 ||| |||Φ4 ||| c2
|{z}
=:γ

we conclude the proof. 

With this stability result, we can proof the existence of a unique solution:

Lemma 9 (Existence and Uniqueness). For every f ∈ H −1 (Ω), there exists a unique solution
U ∈ H01 (Ω; ΓD ) × H01 (Ω; ΓD ) of

A(U, Φ) − µ(∇v, ∇ψ)s = (f, φ) ∀Φ = (φ, ψ) ∈ H01 (Ω; ΓD ) × H01 (Ω; ΓD ).

For this solution U the following stability estimate holds

1
|||U ||| + k∇vks ≤ kf k−1 ,
γ
where γ > 0 is the constant from Lemma 8 and cp the Poincaré constant.

Proof: (i) Let X := H01 (Ω; ΓD ) × H01 (Ω; ΓD ) and X 0 be its dual space. Then, we introduce
the linear operator L : X → X 0 by

hLU, Φi = a(U, Φ) − µ(∇v, ∇ψ)s ∀Φ ∈ X.

This operator is continuous:

kLU kX 0 ≤ ckU kX ,
due to the continuity of the bilinear form:

kLU kX 0 := sup |hLU, W i| = sup |a(U, W ) − µ(∇v, ∇ψ)s | ≤ cc (|||U ||| + k∇vks ) .
|||W |||=1 |||W |||=1

(ii) Next, we show that this operator is injective. U1 , U2 be given with a(U1 , Φ)−µ(v1 , ψ)s =
a(U2 , Φ) − µ(v2 , ψ)s for all Φ. E.g., U1 and U2 can be two solutions of the model problem.
Then, by Lemma 8

1 a(U1 − U2 , Φ) − µ(∇(v1 − v2 ), ∇ψ)s

|||U1 − U2 ||| ≤ sup = 0.
γ |||Φ|||

Thus, for every f ∈ L(V ) there exists a unique inverse U = L−1 f which is solution of the
model-problem.
(iii) It holds:
1 1 1 1
|||U |||+k∇vks ≤ sup (a(U )(Φ) − (∇v, ∇ψ)s ) = sup (f, φ) ≤ sup ||f ||−1 kφk1 = ||f ||−1
γ |||Φ|||=1 γ |||Φ|||=1 γ |||Φ|||=1 γ

Thus, the operator L and its inverse L−1 are continuous on the range of L.

60
4.5 Some mathematical theory for fluid-structure interaction problems (noch ein wenig experimentell und mit Vo

(iv) Since L and L−1 are continuous, it follows that L(X) is closed. With the closed range
theorem we conclude that L is an isomorphism on
L : X → {Φ ∈ X : a(U, Φ) − (∇v, ∇ψ) = 0, ∀U ∈ X}0 ⊂ X 0 .
The space X 0 := {l ∈ V 0 : hl, φi = 0, ∀φ ∈ X} ⊂ X 0 is the polar of X.
(v) We finally get surjectivity by showing that
{Φ ∈ X : a(U, Φ) − µ(∇v, ∇ψ)s = 0, ∀U ∈ X} = {0},
e.g. by showing that for every Φ ∈ X, there exists a U ∈ X, such that
a(U, Φ) − µ(∇v, ∇ψ)s 6= 0.
This is easily seen by constructing special trial-functions. E.g. if the given Φ = (φ, ψ) fulfills
k∇φks > 0, we choose v = 0 and u = δφ, where δ ≥ 0 is zero in Ωf and kδ∇φks > 0. Then,
a(U, Φ) = µkδ∇uk2s . This approach works likewise for the other terms. 
Next, let Vh × Vh = Xh ⊂ X be a conform discretization of the model problem. Then, we
can proof the following a best-approximation result:
Lemma 10 (Best approximation). Let Xh := Vh × Vh ⊂ H01 (Ω; ΓD )2 be a conform dis-
cretization fulfilling the inf-sup condition
a(Uh , Φh ) − µ(∇vh , ∇ψh )s
sup ≥ γh (|||Uh ||| + k∇vh ks ) ∀Uh ∈ Xh ,
Uh ∈Xh |||Φh |||
with γh ≥ γ0 > 0. Further,
∀Φh ∈ Vh ∃Uh ∈ Vh : a(Uh , Φh ) − (∇vh , ∇ψh )s 6= 0.
Then, for the solution Uh ∈ Vh × Vh of the discretized model-problem it holds:
cc
|||U − Uh ||| + k∇(v − vh )ks ≤ inf (|||U − Wh ||| + k∇(v − wh )ks ) .
γh Wh ∈Xh

Proof: Since Vh is a conform approximation space we have Galerkin orthogonality. Thus

for arbitrary Wh ∈ Vh :
a(Uh − Wh , Φh ) − µ(∇(vh − wh , ∇ψh )s = a(U − Wh , Φh ) − µ(∇(v − w), ψh )s ∀Φh .
With the stability result
1 a(Uh − Wh , Φh ) − µ(∇(vh − wh ), ψh )s
|||Uh − Wh ||| + k∇(uh − wh )ks ≤ sup
γ |||Φh |||
1 a(U − Wh , Φh ) − µ(∇(v − wh ), ψh )s cc
= sup ≤ (|||U − Wh ||| + k∇(v − wh )ks ) ,
γ |||Φh ||| γh
and the triangle inequality the result follows:
|||U − Uh ||| + k∇(v − vh )ks ≤ |||U − Wh ||| + k∇(v − wh )ks + |||Uh − Wh ||| + k∇(vh − wh )ks
cc
 
≤ 1+ (|||U − Wh ||| + k∇(v − wh )ks )
γh


61
4 The Finite Element Method for Continuum Mechanics

4.6 Solution of the monolithic FSI formulation

As in the previous section, we consider first consider the simplified set of equations:

ν(∇v, ∇φ)f + µ(∇u, ∇φ)s = (f, φ)

α(∇u, ∇ψ)f − (v, ψ)s = 0.

This corresponds to a linear system of equations given by

x = b,
Ax

where x = (v, u) and b = (f, 0). We split the matrix into a fluid-part and into a structure-part
written as: ! !
νL 0 0 µL
A = Af + As = + ,
0 αL −M 0
f s

where Lij = (∇φj , ∇φi ) is the discretized Laplace-operator (on regular meshes the usual
9-point-stencil) and Mij = (φj , φi ) the mass-matrix. Integration of the matrices is done only
in the corresponding subdomain. Then, the linear system of equations gets:

νLf v f + µLsu s = f
αLf u f − Msv f = 0.

We solve this linear system with iterative methods. These methods use the norm of the
defect as stopping criteria. For the second equation we have due to scaling reasons

di = kαLf u if k + kMsv if k ∼ αf ku
uif k + h2 kvv if k,

and on finer meshes we lack control over the velocity. vh = 0 will not be satisfied. By choosing
α = h2 we balance the two parts in the norms. Further, by scaling the lower equation by 1/α,
we reach that all norms in the entire set of equations have the same asymptotic behavior
with respect to the mesh size h.
Hence, the set of equations is given by the matrices
! !
νL 0 0 µL
A = Af + As = + .
0 L −h−2 M 0
f s

First, we want to estimate the condition number of the two matrices. We note, that for L
and M we have

λmin (M ) ∼ chd , λmax (M ) ∼ chd , cond2 (M ) ∼ 1

d
λmin (L) ∼ ch , λmax (L) ∼ ch d−2
, cond2 (M ) ∼ h−2 ,

where d is the spacial dimension. The matrix Af is block-diagonal and symmetric. We can
estimate the largest and smallest eigenvalue as:

λmax (Af ) = max{νλmax (L), λmax (L)} = λmax (L) max{ν, 1},
λmin (Af ) = min{νλmin (L), λmin (L)} = λmin (L) min{ν, 1}.

62
 4.6 Solution of the monolithic FSI formulation

Since the viscosity is usually small, we derive

cond2 (Af ) = ν −1 cond2 (L) ∼ ν −1 h−2

The matrix As is not symmetric. Here, the condition number is defined as:
1
1
−1 21 λmax (ATs As ) 2
cond2 (As ) = kAs k2 kA−1
s k2 = λmax (ATs As ) 2 λmax (A−T
s As ) = 1
λmin (As ATs ) 2
For ATs As and As ATs holds due to symmetry of M and L:
!
ν 2 LL 0
ATs As = As ATs = 1
0 α2
M M
Hence,
1
λmax (ATs AS ) 2 = λmax (As ) = max{µλmax (L), h−2 λmax (M )},
1
λmin (ATs AS ) 2 = λmin (As ) = min{µλmin (L), h−2 λmin (M )},
and finally for the condition number of the matrix As :
max{µλmax (L), h−2 λmax (M )} max{µhd−2 , hd−2 } max{µ, 1}
cond2 (As ) = −2
∼ d d−2
=
min{µλmin (L), h λmin (M )} min{µh , h } min{µh2 , 1}
For small mesh sizes, the condition number goes like
cond2 (As ) = h−2 .

Considered separately, both subproblems have a condition number as expected. It is however

essential to scale the second equation by h−2 , otherwise, the condition number would behave
like cond2 (Af ) = h−4 . Due to the coupling over the interface, where both subproblems
fall together, the condition number of the overall system is difficult to estimate. Numerical
estimations of the eigenvalues however show that the overall condition still goes like h−2 .
A geometric multigrid method is thus optimal for solving the system Ax x = b to deal with
the bad conditioning. It is however very difficult to obtain robust multigrid smoothers for
the coupled problem. Methods that are available for either flow- or structure problems will
in general not work on the coupled system. The multigrid smoother however only needs to
solve high frequent error parts which are local. Thus it is advisable to split the system into
the two subproblems for smoothing. On mesh-level l, we write the smoother in general form
as a preconditioned Richardson iteration:
xi = xi−1 + P −1 (bb − Ax
xi−1 ),
where P ∼ A is the preconditioner. For the inverse of P , use an approach from domain
decomposition:
P −1 := A−1 −1
f + As ,
with the goal to split the system into the both subproblems. Then, each step of the iteration
reads:
x i = x i−1 + (A−1 −1
xi−1 ),
b − Ax
f + As )(b
and can be written in two sub-steps:

63
4 The Finite Element Method for Continuum Mechanics

1. r i−1 := b − Ax
xi−1
2. Af w if = r i−1
3. Asw is = r i−1
4. x i := x i−1 + ω(w
w if + w is )
The two systems in Step 2 and 3 of this algorithms can by smoothed using available stan-
dard techniques. For the flow-problem a smoother of Vanka-type or an incomplete LU-
decomposition of the matrix Af is robust. Due to the nonlinear transformations, and ILU-
type smoother is also necessary for the structure equation.
This linear solution process can be directly transferred to the full set of equations.
Finally, we shortly refer to the full set of (stationary) fluid-structure interaction equations:

(Jˆρ̂f F̂ −1 (v̂ · ∇)v̂,

ˆ ˆ φ̂) + (Jˆσ̂f F̂ −T , ∇
φ̂)Ω̂f + (Jˆσ̂s F̂ −T , ∇ Ω̂s
ˆ φ̂) = (Jˆρ̂f fˆf , φ̂) + (ρ̂s fˆs , φ̂)
Ω̂f Ω̂f Ω̂f
d JˆF̂ −1 v̂), ξ)
ˆ ˆ ˆˆ
(div( Ω̂f + (αp ∇p̂, ∇ξ)Ω̂s = 0,
ˆ ∇
(αu ∇û, ˆ ψ̂) + (−v̂, ψ̂) = 0,
Ω̂f Ω̂s

The system is highly nonlinear and we a Newton’s method is used for solving. In every
Newton-Update step the linear system can again be split into the two subdomains and we
solve with a matrix:
   
K(vv , u ) + Σf (u
u) B(u
u) FK,Σf (u
u, v ) 0 0 u)
Σs (u
A = Af + As =  C(u u, v ) 0 u, v )  +  0
FC (u αp L 0  .
   
0 0 αu L f
−M 0 0 s

Here, by K we denote the matrix belonging to the convection, Σf and Σs are the stress-
tensors, B is the pressure-gradient, C the divergence and L the Laplace-matrix. The matrices
F denote the derivatives with regard to the ALE-mapping. These terms cannot easily be
linked to a specific differential operator. The dependence on the solution u and v is due to
the nonlinearity.
For the solution process we again split the system with a domain decomposition approach
using as preconditioner
P −1 = A−1 −1
f + As .

For the approximation of both these subproblems, the matrix-structure can be used to further
simplify, e.g. the solution of the structure-block can be split into three subproblems:

As x δ = r ⇔ −Mvv δ = r v , αp Lppδ = r p , uδ = r u .
u)u
Σs (u

The fluid-problem is more difficult, but it is still possible to split this system into a Laplace
equation and a saddle-point system:
! ! !
δ δ u K(vv , u ) + Σf (u
u) B(u
u) vδ r v − FK,Σf (u uδ
u, v )u
Af x = r ⇔ u =r ,
αu Lu = .
C(u u, v ) 0 pδ p
r − FC (uu, v )u
u δ

64
5 Partitioned Approaches

In this chapter we discuss partitioned approaches for the solution of fluid-structure interaction
problems. Opposed to the previously treated monolithic approach, the system is decoupled
and fluid- and solid-problem are solved independently. One big advantage of a decoupling
is that existing and efficient solution methods can be used for both subproblems. The
equilibrium condition on the interface is no longer implicit part of the formulation, instead
it has to be coupled into the subproblems via boundary values or loads.
In every time-step tn−1 → tn we split the system into the two subproblems, namely the fluid
problem F and the solid problem S:

F F F

S S S

tn−1 tn tn+1

Opposed to monolithic approaches, we aim at solving the subproblems independently of each

other. Thus, the equilibrium condition given on the interface (in Eulerian formulation)

vf = vs , n · σs = n · σf on Γi (t),

is not implicit part of the problem formulation. Instead it must be contributed in form of
boundary conditions and loads. In the following, we describe the two subproblems.

Solution of the fluid problem F Within one time-step tn−1 → tn the fluid domain Ωf (t) can
change. We have already discussed the problem of solving the Navier-Stokes equations on the
time depending domain Ωn−1 f := Ωf (tn−1 ) → Ωf (tn ) =: Ωnf . Key is the ALE-Formulation
of the Navier-Stokes equations. Here we describe a more general approach than used for
monolithic formulations. By Ω̃nf we denote the reference-domain to be used for the time-step
tn−1 → tn and by T̃fn−1 : Ω̃nf → Ωn−1
f and T̃fn : Ω̃nf → Ωnf we denote the ALE-mappings at
the beginning and at the end of the time-interval [tn−1 , tn ]. We note that Ωnf is an unknown
in this time-step. In general, we thus cannot assume that T̃fn is known exactly! At the old
time-step tn−1 , the domain partitioning is given by the structure’s deformation and T̃fn−1 is
known. Without further information, the approximation T̃fn ≈ T̃fn−1 is the only possibility.

One choice for the ALE-mapping would be to use the reference domain Ω̃f := Ω̂f = Ωf (0)
for all time-steps. This would be the ALE-mapping as used in the monolithic approach

65
5 Partitioned Approaches

Ω̂s

Ω̂f

Figure 5.1: Non-matching meshes for the fluid and structure problem.

described before. One can however also use different reference domains in every time step, e.g.
Ω̃f := Ωn−1
f which is the domain at the old time-step. Then, T̃fn−1 = id is the identity. Using
different reference domains in every time-step will ask for a remeshing of the computational
domain after every step. Then, the ALE-mapping only refers to the difference between the
old and the new time-step. We will stress the benefit of – at first glance very expensive –
technique for problems involving large deformations later on.
The coupling to the structure problems is embedded by prescribing Dirichlet conditions for
the fluid’s velocity ṽf on the interface Γ̃i . This velocity has to match the velocity v̂s of the
structure:
ṽf = ṽs on Γ̃i .
Here, we have to face the problem, that the structural velocity is given in Lagrangian co-
ordinates v̂s (x̂) on Ω̂s , while the fluid’s velocity is given in arbitrary ALE coordinate ṽf (x̃),
where the fluid’s reference domain Ω̃f does not need to match the Lagrangian reference
domain Ω̂nf if different ALE-mappings are used in different time-steps. This makes the ac-
curate evaluation of the Dirichlet data difficult. Furthermore, even if one ALE mapping
T̂f : Ω̂f → Ωf (t) is used for all time-steps, it is possible (and most often advisable) to use
different triangulations Ω̂f,h and Ω̂s,h . Then, if these triangulations do not match on the
common interface Γ̂i as shown in Figure 5.1, the interpolation of the Dirichlet values is not
straightforward.
We formulate fluid-problem ALE-formulation using the θ-scheme for time-discretization:
Problem 5 (Fluid Problem in ALE coordinates F − ALE). Given ṽ n−1 , p̃n−1 , T̃ n−1 and
T̃ n find ṽ n ∈ ṽ D,i + H01 (Ω̃f ; Γ̃D,i ) and p̃n ∈ L20 (Ω̃f ), such that
 
ρ̂f J(ṽ n − ṽ n−1 ), φ̃ − ρ̂f J˜F̃ −1 (T̃ n − T̃ n−1 ) · ∇v,
˜ φ̃


Ω̃f Ω̃f

+ θkn (F̃ −1 n
ṽ · ∇ṽ ˜ φ̃)Ω = θkn (ρf J˜n−1 f˜n−1 , φ̃) + (1 − θ)kn (ρf J˜n f˜n , φ̃)
˜ n , φ̃)Ω + θkn (J˜n σ̃ n F̃ −T , ∇
f f f f Ω̃f f Ω̃f
˜ n−1 , φ̃) − (1 − θ)kn (J˜n−1 σ̃ n−1 [F̃ n−1 ]−T , ∇
− (1 − θ)kn ([F̃ n−1 ]−1 ṽ n−1 · ∇ṽ ˜ φ̃) ∀φ̃ ∈ H01 (Ω̃f ; Γ̃D,i ),
Ω̃f f Ω̃f
˜
f J˜n [F̃ n ]−1 ṽ n ), ξ)
(div( ∀ξ˜ ∈ L20 (Ω̃f ). (5.1)
Ω̃f = 0

Here, with the bar we denote the average values over the time-interval:

ṽ n−1 + ṽ n J˜n−1 + J˜n J˜n−1 [F̃ n−1 ]−1 + J˜n [F̃ n ]−1
v := , J := , JF −1 := .
2 2 2

66
Further, for the extension of the boundary values into the domain it holds

ṽ D,i = ṽ D on ∂ Ω̃f /Γ̃i , ṽ D,i = ṽsn on Γ̃i .

The structure’s velocity on the interface is a further prerequisite (that also cannot be given
exactly since it is unknown of the coupled system). Finally, F̃ n := ∇˜ T̃ n and J˜n := det(F̃ n ).

We have already stated, that the ALE-mapping T̃ n : Ω̃nf → Ωnf is not known a priori
and needs to be determined in an additional problem, which will be call M, the mapping
problem. The solution of this problem will be described separately. Here it gets obvious, that
a partitioned approach cannot readily result in the correct solution of the coupled problem,
since the mapping T̃ n must be provided beforehand. The correct mapping however depends
on the domain layout at old and new time-step. By approximating T̃ n ≈ T̃ n−1 the correct
domain is not considered.

Solution of the solid problem S The solid problem is formulated and will be solved in
Lagrangian coordinates on the reference domain Ω̂s . The coupling to the fluid-problem is
given by means of Neumann boundary values on the interface:

n̂ · (Jˆσ̂s F̂ −T ) = n̂ · (Jˆσ̂f F̂ −T ) on Γ̂i ,

where σ̂f is the fluid’s stress tensor on the interface evaluated with the best known approx-
imation of velocity ṽf and pressure p̃f . Again, the proper evaluation of the interface force
is difficult, since it is required in Lagrangian coordinates which do not have to match the
fluid’s coordinate system Ω̃f . We write the structure problem discretized in time with the
θ-scheme:
Problem 6 (Solid Problem S). Given ûn−1 and v̂ n−1 find ûn ∈ ûD + H01 (Ω̂s ; Γ̂D ) and
v̂ n ∈ Ĥ01 (Ω̂s ; Γ̂D ) such that

(ρ̂s (v̂ n − v̂ n−1 ), φ̂)Ω̂s + θkn (Jˆn σ̂sn [F̂ n ]−T , ∇

ˆ φ̂) = θkn (ρ̂s fˆsn , φ̂)
Ω̂s Ω̂s
ˆ n−1 ˆn−1 n−1 n−1 −T ˆ
+ (1 − θ)kn (ρ̂s f s , φ̂) − (1 − θ)kn (J
Ω̂s σ̂ s [F̂ ] , ∇φ̂) Ω̂s
n−1
+ θkn hn̂s · ĝfn , φ̂iΓ̂i
+ (1 − θ)kn hn̂s · ĝf , φ̂iΓ̂i ∀φ̂ ∈ 1 D
H0 (Ω̂s ; Γ̂ )
(ûn − ûn−1 , ψ̂)Ω̂s − θkn (v̂ n , ψ̂)Ω̂s = (1 − θ)kn (v̂ n−1 , ψ̂)Ω̂s ∀ψ̂ ∈ H01 (Ω̂s ; Γ̂D ) (5.2)

where ûD is some extension of the Dirichlet data on Γ̂D into the domain. The interface load
is given by
ĝfn−1 = Jˆn−1 σ̂fn−1 [F̂ n−1 ]−T , ĝfn = Jˆn σ̂fn [F̂ n ]−T .
The fluid’s stresses (again unknown at tn ) need to be given for the solution of the time-step.

Solution of the mapping problem M The third subproblem consists of finding the new
domain partitioning Ω := Ωnf ∪Ωns and of finding an appropriate ALE-mapping T̃f : Ω̃f → Ωnf .

In the structure domain, the deformation describes the layout of the domain via T̂sn = id+ ûn .
This mapping also indicates the position of the interface Γni and thus describes the boundary

67
5 Partitioned Approaches

of the fluid domain Ωnf . To define the mapping T̃f : Ω̃nf → Ωnf we need to provide a fluid-
deformation ũnf . This can be done by either interpolating the boundary values into the
domain or by solving an partial differential equation. Here, we describe the easy method of
using a harmonic extension:

Problem 7 (Mapping Problem M). On Ω̃f with given ũns on Γ̃i , find ũf , such that

˜ n = 0,
−∆ũ ũnf = ũns on Γ̃i , ũnf = 0 on ∂ Ω̃f /Γ̃i .
f

Again, due to different coordinate systems Ω̃nf and Ω̂s the boundary values need to be
interpolated.

5.1 Coupling of the subproblems

If these three problems where all solved simultaneously, ṽfn , p̃nf , ûns , v̂s and ũnf would solve
the coupled fluid-structure interaction problem as given in the monolithic formulation.

Every time-step of a partitioned solution approach will consist of the following three sub-
steps. These can be combined in various order or solved parallel.

F Solve the flow problem

ṽfn−1 , p̃n−1
f , v̂sn−1 , v̂sn , T̃fn−1 , T̃fn ⇒ ṽfn , p̃nf

S Solve the structure problem

ûn−1
s , v̂sn−1 , σ̃fn−1 , σ̃fn ⇒ ûns , v̂sn

M Solve the mapping problem

ũn−1
f , ûns ⇒ ũnf

In every subproblem some input-values are not known exactly but part of the problem to be
solved. We will distinguish between two different kinds of coupling algorithms: staggered or
weakly coupled on the one and iterative staggered or strongly coupled on the other side.

Weakly coupled approaches solve every subproblem only once per time-step. These ap-
proaches are generic explicit approaches. The interface condition will not be fulfilled exactly.
In strongly coupled approaches, the three subproblems are solved in an iterative manner.
This way, the unknown data (like the interface velocity ûns in the fluid problem) will be
approximated in an outer iteration.

68
 5.2 Weakly coupled approaches

5.2 Weakly coupled approaches

Here, we describe so called weakly coupled or staggered solution approaches. We suppose

the general case, where a different reference domain Ω̃nf := Ωn−1
f is used in every time-step
tn−1 → tn . The following sub-tasks need to be solved:

1. Create the new fluid-domain and mesh for tn−1 → tn :

ũn−1
f
ũn−1
f
Ω̃n−1
f −−−→ Ω̃nf , Ω̃n−1 n
f,h −−−→ Ω̃f,h .

2. Transfer the fluid variables onto the new mesh

ũn−1
f
ũn−1
f
ṽfn−1 Ω̃n−1 −−−→ ṽfn−1 Ω̃n , p̃n−1  n−1 −−−→ p̃n−1  n
   
f f
f Ω̃ f
f Ω̃ f

3. Transfer the structure deformation and velocity on the interface to the new fluid-
coordinates:
 ũn−1
f
  ũfn−1 
ûn−1 −−−→ ũsn−1  , v̂sn−1  −−−→ ṽsn−1  .
   
s 
Γ̂i Γ̃i Γ̂i Γ̃i

4. Solve the mapping problem to match the last structure-velocity

˜ n = 0,
−∆u ũnf = ũn−1 on Γ̃i .
f s

5. Solve the fluid problem

ũn
f
ṽfn−1 , p̃n−1
f −→ ṽfn , p̃nf .

6. Transfer the interface-load to the structure domain

ũn
f
σ̃fn −→ σ̂fn

7. Solve the structure problem

σ̂fn
ûn−1
s , v̂sn−1 −→ ûns , v̂sn .

If the standard ALE approach with just one reference domain Ω̃nf = Ω̂f is used for all
time-steps n, the first two steps of the algorithm do not need to be considered, since the
computational mesh stays the same. Along the interface however, the structure’s velocity
and deformation needs to be transferred since possibly non-matching meshes are used.

In the following, we describe the different solution-steps of this basic partitioned algorithm
in detail:

69
5 Partitioned Approaches

Step 1: mesh-movement We discuss the case, where the reference domain Ω̃nf for the
time-step tn−1 → tn is given by the deformed domain of the previous time-step, e.g. by the
transformation
Ω̃nf := T̃[tn−1
n−2 ,tn−1 ]
(Ω̃fn−1 ).

There are basically two possibilities for generating the new mesh Ω̃nf,h . First, based on the
boundary Ω̃nf a new mesh can be generated with a mesh-generator. The advantage of this
method is that in every step, a good mesh quality fulfilling shape regularity can be assured.
As a drawback, by generating a new mesh there would be no topological connection between
the old and new meshes. This would complicate the transfer of the solution variables from
Ω̃n−1 n
f,h to Ω̃f,h in Step 2 of the algorithm.

Alternatively, the domain deformation ũn−1

f can be used to move the mesh nodes to the new
mesh:
Ω̃n−1 n−1
f,h 3 x̃i ⇒ x̃in−1 + ũn−1
f (x̃in−1 ) ∈ Ω̃nf,h .
This way, the new mesh is generated by moving the nodes (and elements) of the old one.
As indicated in Figure ??, moving the mesh nodes can lead to strongly distorted elements.
Thus, in a second step, the new mesh needs to be locally remeshed by adding or removing
elements to produce a valid, shape-regular mesh. In the following, we assume, that the
meshes are modified by this second approach.

Step 2: solution transfer Velocity and pressure need to be transferred to the new mesh
Ω̃n−1 n
f,h → Ω̃f,h . If the mesh is generated by moving with ũf
n−1
, one could simply interpolate
via
ṽfn−1 Ω̃n (x̃ni ) := ṽfn−1 Ω̃n−1 (x̃in−1 ), where x̃ni = x̃n−1 + ũn−1 (x̃in−1 ).
 
i f
f,h f,h

This however is not conforming in a Galerkin-sense. Instead we need to L2 -project the

n−1
solution onto the new mesh. Let vf,h be the discrete solution from the last time-step in
n−1
ALE-coordinates and vf,h the old solution on the new mesh Ω̃nf,h to be given in Eulerian
coordinates. Then with the basis-functions {φi } on the new mesh it holds by transformation
with T̃[tn−2 ,tn−1 ] :
(v n−1 , φi )Ω̃n = (J˜ṽf,h
n−1
, φ̃i )Ω̃n−1 ,
f f

n−1
where φ̃i are the corresponding test-function on Ω̃f,h
given in ALE-coordinates. To properly
project the old solution onto the new mesh, we first need to evaluate the right hand side on
the old mesh, followed by solving the mass-matrix on the new mesh:

i = 1, . . . , N : ri := (J˜ṽf,h
n−1
, φ̃i )Ω̃n−1 , n−1
(vf,h , φi ) = ri ∀i = 1, . . . , N.
f

n−1 n−1 
This L2 -projection vf,h

is to be used as ṽf,h Ω̃n
on the new mesh as initial value.
f,h

If the new mesh would be generated by a complete remeshing operation, there would be no
link between the old and new test-functions. In this case, the evaluation of the right hand
side ri is very expensive.

70
 5.3 Strongly coupled partitioned approaches

Step 4: solution transfer In this step, using the Laplace equation is only one alterna-
tive. For problems with large deformation, an elastic structure equation or the bi-harmonic
equation yields better results.
For partitioned approaches, it is even possible to use a simple interpolation for constructing
the deformation of the flow-domain. In contrast to monolithic models, this mesh-deformation
equation is solved separately and does not introduce couplings in the flow-systems system
matrix.

Step 5, Step 6: the fluid and structure problem Here, the two subproblems need to be
solved. This is done separately and standard solvers can be used.
In weakly coupled solution approaches, the interface conditions vs = vf and n · (σf − σs ) = 0
are not fulfilled exactly. These approaches are only successful if the coupling between the
two problems is very weak.

5.3 Strongly coupled partitioned approaches

Strongly coupled partitioned approaches try to satisfy the interface conditions by iterating
in every time-step until the solution does not change any more. For simplicity, we introduce
some notation for the different sub-steps in the algorithm:
Mesh Solve for fluid-deformation
(uf,(i) ) = M (uΓ,(i) ),

Fluid Solve the flow problem

(vf,(i) , pf,(i) ) = F (vfn−1 , uf,(i) , vΓ,(i) )

Solid Solve the solid problem

(vs,(i) , us,(i) ) = S(vsn−1 , un−1

s , σΓ,(i) )

Fluid-Transfer Transfer the solid’s velocity to the flow domain

(vΓ,(i) ) = TF (vs,(i−1) )

Solid-Transfer Transfer the fluid-forces to the solid

(σΓ,(i) ) = TS (vf,(i) , pf,(i) )

Interface-Restrict Restrict solid’s deformation to the interface

(uΓ,(i) ) = TR (us,(i) )

71
5 Partitioned Approaches

Then, the basic iterative algorithm for every time-step is as follows: Get an initial guess of
the interface displacement uΓ,(0) and iterate for i ≥ 1
1. Solve mesh equation (uf,(t) ) = M (uΓ,(i−1) )
2. Transfer interface-velocity (vΓ,(i) ) = TF (vs,(i−1) )
3. Solve flow problem (vf,(i) , pf,(i) ) = F (uf,(i) , vΓ,(i) )
4. Transfer interface loads (σΓ,(i) ) = TS (vf,(i) , pf,(i) )
5. Solve structure problem (vs,(i) , us,(i) ) = S(σΓ,(i) )
6. Restrict new interface deformation uΓ,(i) = TR (us,(i) )
7. Check residual |uΓ,(i) − uΓ,(i−1) | < tol
The fix-point iteration can be written in a compact form:

uΓ,(i) = TR ◦ S ◦ TS ◦ F ◦ TF ◦ M (uΓ,(i−1) ).

And if we combine the mesh problem M and the flow problem F together with the fluid-
transfer TF into one combined problem, denoted by F̃ , and if we include TS , TR and the
structure problem S into S̃, the fix-point iteration gets:

uΓ,(i) = S̃ ◦ F̃ (uΓ,(i−1) ). (5.3)

For convergence control we introduce the interface residual as

rΓ,(i) = uΓ,(i) − uΓ,(i−1) , (5.4)

which is used in Step 7 of the algorithm as stopping criteria.

The Dirichlet-Neumann setup of the iteration assumes, that the coupling from the solid to
the fluid problem is stronger than the other way around. If the structure has a nearly rigid
wall – like the surface of an airplane – the fluid’s forces have little influence on the structure’s
displacement, whereas a deformation of the structure immedeatly alters the flow-problem.
Then, the proposed iteration seems reasonable and it can be assumed that the predicted
structure displacement will be accurate. If on the other hand the structure is very flexible
and the fluid’s density is large compared to the structure’s density, the influence of the flow
field can be significant. Then, extensive iteration is necessary, and for some problems the
scheme will not converge at all.

5.4 Acceleration schemes for partitioned fsi-solvers

Written as (5.3), the partitioned fluid-structure interaction scheme is given as a fix-point

iteration on the position of the interface. For acceleration of this iteration we use relaxation
methods by first estimating a prediction of the new displacement and then adding only a
share of it

ũΓ,(i) = S̃ ◦ F̃ (uΓ,(i−1) ), rΓ,(i) := ũΓ,(i) − uΓ,(i−1) , uΓ,(i) := uΓ,(i−1) + ωi rΓ,(i) ,

72
 5.4 Acceleration schemes for partitioned fsi-solvers

(uΓ,(i−1) , rΓ,(i) )

(uΓ,(i) , rΓ,(i+1) )

(uΓ,(i+1) , 0)

Figure 5.2: Secant method for convergence acceleration with Aitken relaxation.

with a variable relaxation parameter ωi ∈ R. The simple choice of one fixed relaxation
parameter ωi := ω for all steps leads to a robust convergent scheme, if ω  1 is small
enough. The convergence rate however will be very low. In the following we describe
different methods for choosing a dynamic relaxation parameter.

5.4.1 Aitken relaxation

Aitken relaxation is best explained by assuming, that uΓ,(i) is a scalar value. Using the
last to iterates, the new deformation is approximated. Let uΓ,(i−1) , ũΓ,(i) = S̃ ◦ F̃ (uΓ,(i−1) )
and uΓ,(i) , ũΓ,(i+1) = S̃ ◦ F̃ (uΓ,(i) ) be given. Then, the residual rΓ,(i) = ũΓ,(i) − uΓ,(i−1) and
rΓ,(i+1) = ũΓ,(i+1) − uΓ,(i) are known.

Then, we can figure rΓ,(i) as function of uΓ,(i−1) and using the last given values we try to
find uΓ,(i+1) such that rΓ,(i+2) (uΓ,(i+1) ) = 0. For finding this root we use secants method,
see Figure 5.2, describing uΓ,(i+1) as:

rΓ,(i+1) − rΓ,(i)
0 = rΓ,(i+1) + (u − uΓ,(i) )
uΓ,(i) − uΓ,(i−1) Γ,(i+1)
uΓ,(i) − uΓ,(i−1)
⇔ uΓ,(i+1) = uΓ,(i) − rΓ,(i+1)
rΓ,(i+1) − rΓ,(i)
uΓ,(i−1) − uΓ,(i)
= uΓ,(i) + (ũ −u ).
uΓ,(i−1) − uΓ,(i) + ũΓ,(i+1) − ũΓ,(i) | Γ,(i+1){z Γ,(i)}
| {z } rΓ,(i+1)
ωi+1

We can derive a recursion formula for the relaxation parameter ωi+1 :

uΓ,(i−1) − uΓ,(i)
ωi+1 =
uΓ,(i−1) − uΓ,(i) + ũΓ,(i+1) − ũΓ,(i)
uΓ,(i−1) − (uΓ,(i−1) + ωi rΓ,(i) )
=
rΓ,(i+1) − rΓ,(i)
rΓ,(i)
= −ωi .
rΓ,(i+1) − rΓ,(i)

73
5 Partitioned Approaches

This result cannot directly be transferred to the vector-valued case, since dividing by rΓ,(i+1) −
rΓ,(i) is not possible. Hence, to transform the vector entities into scalars, we project all values
in the rΓ,(i+1) − rΓ,(i) -direction to get:

(rΓ,(i) , rΓ,(i+1) − rΓ,(i) )

ωi+1 = −ωi .
krΓ,(i+1) − rΓ,(i) k2

Aitken’s method works very well for a large range of fluid structure interaction problems. It
is very cheap in a sense, that no additional information than the last two approximations
are necessary. In particular, we do not need derivatives or sensitivities from the solution. To
start the iteration, ω0 must be suitably chosen, either by using ωin from the last time-step
or by taking a small enough value.

5.4.2 Steepest descent relaxation

We try to find the best relaxation parameter ωi in the given search direction rΓ,(i) . For
this purpose, we introduce a merit function φ which is minimal at the searched interface
displacement uΓ . The relaxation parameter is given by

ωi := arg min φ(uΓ,(i−1) + ωi rΓ,(i) ).

ωi

Assuming that φ is differentiable, the condition to determine ωi is given by

dφ ∂φ(uΓ,(i−1) + ωi rΓ,(i) )
0= = rΓ,(i) .
dωi ∂uΓ
Thus, the condition to specify ωi is:

φ0 (uΓ,(i−1) + ωi rΓ,(i) ) = 0.

We Taylor expand φ around the last iteration uΓ,(i−1) to get

 T ωi2 T
φ(uΓ,(i−1) + ωi rΓ,(i) ) ≈ φ(uΓ,(i−1) ) + ωi φ0 (uΓ,(i−1) ) rΓ,(i) + r φ00 (uΓ,(i−1) )rΓ,(i) .
2 Γ,(i)
And for determining ωi we derive the relation

φ(uΓ,(i−1) + ωi rΓ,(i) ) − φ(uΓ,(i−1) )

φ0 (uΓ,(i−1) + ωi rΓ,(i) ) ≈
ωi
ωi T
= φ0 (uΓ,(i−1) )T rΓ,(i) + rΓ,(i) φ00 (uΓ,(i−1) )rΓ,(i) = 0,
2
2φ0 (uΓ,(i−1) )T rΓ,(i)
⇔ ωi = − T .
rΓ,(i) φ00 (uΓ,(i−1) )rΓ,(i)

At this point, we need to specify the merit function φ. We assume, that

φ0 (uΓ,(i−1) ) = rΓ,(i) ,

74
 5.5 Stability analysis and the added mass effect

which stands for the assumption, that the residual rΓ,(i) is the direction of the steepest
descent. Further, since φ is a scalar function, this assumption implies, that the residual can
only be the gradient of a scalar function. This assumption is not fulfilled, however it leads
to a symmetric Jacobian:

∂rΓ (uΓ,(i−1) )
JΓ = φ00 (uΓ,(i−1) ) = = φ00 (uΓ,(i−1) ).
∂uΓ
Finally, the best search step ωi is given by
T
2rΓ,(i) rΓ,(i)
ωi = − T
.
rΓ,(i) JΓ rΓ,(i)

The evaluation of this relation requires the evaluation of the matrix vector product with the
Jacobian, JΓ rΓ,(i) . This Jacobian is not available, however the product can be approximated.
A simple way to approximate the Jacobian is by finite differences:

d  rΓ (dΓ,(i−1) + δrΓ,(i) ) − rΓ,(i)

JΓ rΓ,(i) = rΓ (dΓ,(i−1) + srΓ,(i) ) ≈

ds s=0 δ
To evaluate the residual at dΓ,(i−1) + δrΓ,(i) one step of the fsi-iteration (5.3) needs to be
added:
1 
JΓ rΓ,(i) ≈ S̃ ◦ F̃ (uΓ,(i−1) + δrΓ,(i) ) − uΓ,(i−1) − δrΓ,(i) − rΓ,(i) .
δ
The computational cost of this relaxation method is one additional cycle of the fsi iteration.
Further, due to the finite differences, this method is very sensitive.
Other ways of expressing the Jacobian JΓ are based on approximating the derivative of (5.3):

JΓ rΓ,(i) = S̃ 0 ◦ F̃ (uΓ,(i−1) )F̃ 0 (uΓ,(i−1) ).

The derivative of the structure operator S̃ is available, if a Newton’s method is used for
solving the structure problem. The derivative of the fluid field however is not given, since
we here need the derivative with respect to the interface derivation uΓ,(i−1) . This needs to
be approximated. We do not give details on this methods and refer to the literature.

5.5 Stability analysis and the added mass effect

Here, we discuss stability issues of partitioned approaches for fluid-structure interaction. An

analysis of the full set of equations including all nonlinearities is very difficult. Hence, we
strongly simplify the set of equations by assuming:
• There are no external forces or Dirichlet conditions
• The deformation is small and we neglect all geometric nonlinearities in the ALE map-
ping and in the structural displacement
• The convective term in the Navier-Stokes equations is neglected

75
5 Partitioned Approaches

• Viscid effects are neglected

The set of equations is then simplified to

(ρf ∂t v, φ)f − (p, ∇ · φ)f = hfΓfi , φiΓi , (∇ · v, ξ)f = 0

for the flow system and

(ρs ∂tt u, φ)s + ((u), φ)s = h−fΓsi , φiΓi ,

for the solid part. By fΓfi and fΓsi we denote the interface forces. In a iterative partitioned
scheme, the equilibrium shows fΓfi − fΓsi → 0.

Semi-discrete System We first discretize the system only in time. With the usual notation
of Mf for the fluid’s and Ms solid’s mass matrix scaled with the densities, and K for the
solid’s diffusive matrix, B for the gradient operator on the pressure and the divergence
matrix BT , we write the system as:

ρf Mf v 0 + Bp = fΓfi ,
BT v = 0,

and
ρs Ms u00 + Ku = fΓsi .
By neglecting the ALE-mapping and assuming linearity, all coefficients matrices do not
change in time. Thus, for the divergence we get:
d T
B v = BT v 0 = 0,
dt
and the flow system is given as
" #" # " #
ρ f Mf B v0 ff
= Γi .
BT 0 p 0

Next, we split the systems for flow and structure into the interior part and into the interface
part. Then, with v = (vΩ , vΓ ) and u = (uΩ , uΓ ) (similar for the matrices) have

ρf MfΩΩ ρf MfΩΓ BΩ vΩ
    
0 0
f f  0   f 
 ρf MΓΩ ρf MΓΓ BΓ   vΓ  = fΓi  .


BTΩ BTΓ 0 p 0

In a final simplification, we use lumped versions of the mass-matrices. M fl . We assume, that

the mesh is regular and that a linear discretization is used. Then, we can write the lumped
mass matrix as
M fl = |K|I f ,
where I f is the identity and |K| the volume of the support of a fluid node.

76
 5.5 Stability analysis and the added mass effect

Then, the off-diagonal entries MfΓΩ and MfΩΓ are zero and the overall flow system gets

f
    
ρf |K|IΩΩ 0 BΩ vΩ 0 0
f  0   f 
0 ρf |K|IΓΓ BΓ   vΓ  = fΓi  . (5.5)


BΩT BTΓ 0 p 0

In a partitioned scheme, the acceleration on the interface vΓ0 is given by the structural
prediction. Assuming that vΓ0 is known, we can express the remaining fluid variables as

p = ρf |K|(BTΩ BΩ )−1 BTΓ vΓ0

−1
0
vΩ = −ρ−1 T −1 T 0
f |K| BΩ p = −BΩ (BΩ BΩ ) BΓ vΓ .

Next, we can use the second line of system (5.5) to express resulting boundary force as
n o
fΓfi = ρf |K|IΓΓ
f 0 f
vΓ + BΓ p = ρf |K| IΓΓ + BΓ (BTΩ BΩ )−1 BTΓ vΓ0
| {z }
=:MA

We call MA the added mass operator. It maps the interface acceleration onto a force-vector
on the interface. MA is a dimensionless and symmetric operator. If the flow system has a
unique solution, the matrix BTΩ BΩ is positive. Thus, the eigenvalues of MA are larger than
one.

The added mass operator describes, how the prediction of the interface acceleration vΓ0 relates
to the new interface forces for the structure problem:

fΓsi = ρf |K|MA vΓ0 . (5.6)

Hence, we can write the next structure-step with the estimated interface force as:
" #" # " #" # " #
MsΩΩ MsΩΓ u00Ω KsΩΩ KsΩΓ uΩ 0
ρs 00 + = . (5.7)
MsΓΩ MsΓΓ uΓ KsΓΩ KsΓΓ uΓ −ρf |K|MA vΓ0

On the interface we can identify the fluid’s acceleration vΓ0 with the second derivative of the
solid’s displacement u00Γ . This points up the relation
s
ρs MΓΓ u00Γ ∼ −ρf |K|MA vΓ0 ,

and by understanding ρf |K| as a unit-mass of the fluid in the proximity of a fluid node, the
added mass operator acts as additional mass on the degrees of freedom on the interface.

Time discretization schemes For a stability analysis of the partitioned scheme, we apply
time-discretization of the two sub-problems. For simplicity, we use lowest order, fully implicit
time stepping schemes for the step tn → tn+1 , ∆t = tn+1 − tn .

77
5 Partitioned Approaches

In the flow problem, the interface acceleration vΓ0 is given as a predictor and needs to be
estimated using the solid’s deformation. With u00Γ = vΓ0 we approximate using the last known
time-steps:
n+1 n+1 1
vΓ0 = u00Γ = (un − 2un−1 + uΓn−2 ).
∆t2 Γ Γ

Then, the right hand side fΓsi of the solid system is given by (5.6) as

1
fΓsi n+1 = ρf |K|MA (unΓ − 2un−1
Γ + un−2
Γ ).
∆t2
We further simplify the structure system (5.7) by assuming small time-steps. Then, the
stiffness terms can be neglected and we use K s = 0. Further, with the lumped mass matrix
M s = |S|I s the system is reduced onto the interface:

ρs |S|I s u00Γ = −ρf |K|MA vΓ0 .

By discretizing with simple backward second order differences, and by inserting the right
hand side approximated above, we get:
1 1
ρs |S| (un+1 − 2unΓ + un−1
Γ )=− ρf |K|MA (unΓ − 2un−1 + uΓn−2 ).
∆t2 Γ ∆t2 Γ

For a closer analysis, we note that the matrix MA is positive and symmetric. We thus
have an orthogonal system of eigenvectors vi of MA with corresponding eigenvalues µi . The
deformation on the interface unΓ can be developed in this eigenvector system by unΓ = i uni vi .
P

For the coefficients it must hold

ρf |K|
(un+1
i − 2uni + un−1
i )+ µi (uni − 2uin−1 + un−2
i ) = 0. (5.8)
ρs |S|

With the ansatz un+1

i = λi uni the amplification factors λi need to be roots of the characteristic
polynomial:
ρf |K|
(λ3i − 2λ2i + λi ) + µi (λ2i − 2λi + 1) = 0,
ρs |S|
which has a double root at λ1 = 1 and by
 ρf |K| 
(λi − 1)2 λi + µi = 0,
ρs |S|
we find the third root at
ρf |K|
λ3 = − µi .
ρs |S|
For a scheme to be stable, these amplification factors all need to be smaller than one. Thus,
for
ρf |K|
µi > 1,
ρs |S|
the partitioned scheme is unstable, since the added mass effect on the interface is dominant.
Since µi ≥ 1, note that stability of the scheme is governed by the ratio fluid-mass to solid-
mass.

78
 5.5 Stability analysis and the added mass effect

Here, the great success of partitioned scheme in aerodynamics is obvious, where the density
of the air ρf ≈ 1kg m−3 is small in proportion to the density of e.g. aluminum of ρs ≈
3000kg m−3 . This ratio allows for a stable solution with even very simple iterative schemes.
For problems like hemodynamics, where ρf ∼ ρs ∼ 103 kg m−3 , partitioned schemes fail to
converge. Here, monolithic approaches are more efficient.
lat

79
5 Partitioned Approaches

80
6 Alternative Approaches for FSI-problems

Partitioned approaches are successful, if the coupling between fluid and structure is weak.
When combined with remeshing, partitioned approaches are very efficient and can handle
problems with large deformation. If the coupling between the subproblems is strong, parti-
tioned approaches fail to converge.
The monolithic ALE formulation on the other hand is very robust and also suitable for
problems, where partitioned approaches break down. They however have the problem of
modeling large deformation since then, the ALE-mapping T̂f : Ωf → Ωf (t) will loose its
regularity. In some applications, even the contact of the elastic structure with the boundary
of the domain or with another structural part can be given. Then, the topology of the domain
Ωf (t) changes, see Figure 6.1. Here, it is not possible to indicate an invertible ALE-mapping.
In this chapter we discuss alternative approaches for fsi-problems. First, we consider the
Fully Eulerian formulation, where both flow-problem and structure-problem are given in
Eulerian formulation.

6.1 Fully Eulerian Coordinates

For the ALE-formulation we need a mapping T̂f : Ω̂f → Ωf (t) of the flow reference domain
to the actual flow domain Ωf (t). This mapping is used to transform the Navier-Stokes
equations onto the fixed reference domain. For the ALE formulation to work, the mapping
needs to be invertible. If this property is lost, the ALE-formulation breaks down.
Here we try to use the opposite approach: the flow problem is left in the natural Eulerian
coordinates and the structure problem is transformed onto the moving domain Ωs (t). Here,
this transformation is given by the structure’s deformation itself:

T̂s : Ω̂s → Ωs (t), T̂s (x̂, t) = x̂ + ûs (x̂, t).

Ω̂s
Ωs (t)

Ω̂f Ωf (t)

Figure 6.1: Fluid-structure interaction with contact of structure with another structure do-
main leading to a topology change of the domain.

81
6 Alternative Approaches for FSI-problems

The difference between the Fully Eulerian and the ALE approach is best demonstrated
with Figure 6.1: for ALE, the computational domain is the left sketch in the figure, for
the Eulerian approach, the computation is done on the domain on the right half. In both
formulations, one of the two systems needs to be mapped. In ALE, this mapping T̂f is
arbitrary in a sense, that it has no physical relevance. This is the reason, why this mapping
can deteriorate, det(∇ˆ T̂f ) → ∞. In the Fully Eulerian approach, the mapping is given by
the structural displacement itself and it holds:

ˆ := det(∇
det(J) ˆ T̂s ) = det(I + ∇û
ˆ s ).

For incompressible materials it holds det(J) ˆ = 1 and for compressible materials, the limits
ˆ ˆ
det(J) → 0 or det(J) → ∞ are prohibited due to physical reasons. They would imply that
either different material points are at the same location or that holes would appear.
In Eulerian coordinates we define on Ωs (t)

u(x, t) = û(x̂, t), v(x, t) = v̂(x̂, t), x = x̂ + û(x̂, t) ⇔ x̂ = x − u(x, t).

With this definition, the inverse mapping T̂s is given by

Ts : Ωs (t) → Ω̂s , Ts (x, t) = x − u(x, t).

For simplicity, we write:

F := ∇Ts = (I − ∇u), J := det(F ) = det(I − ∇u).

While velocity and deformation are easily available in Eulerian coordinates, their gradient
∇u and ∇v is not. By chain rule, we get:

x̂ = Ts (T̂s (x̂, t), t) ⇒ ˆ

I = (I − ∇u)(I + ∇û).

And thus
ˆ = (I − ∇u)−1 − I.
∇û (6.1)
For the deformation gradient it holds

F̂ = F −1 , Jˆ = J −1 (6.2)

For the temporal derivative we have:

dt v̂(x̂, t) = ∂t v̂(x̂, t) = ∂t v(x, t) + v · ∇v. (6.3)

Then, the equations for an elastic material in Lagrangian formulation (see (2.14)) are given
by:

(ρ̂0 ∂t v̂, φ̂)Ω̂s + (Jˆσ̂ F̂ −T , ∇

ˆ φ̂) = (ρ̂0 fˆ, φ̂)
Ω̂s Ω̂s
(dt û, ψ̂)Ω̂s = (v̂, ψ̂)Ω̂s

82
 6.1 Fully Eulerian Coordinates

We remember, see (2.10) that for the density we have:

ρ(x, t) = Jˆ−1 ρ̂0 (x̂) ⇔ ρ̂0 (x̂) = J −1 ρ(x, t).

Transformation to Ωs (t) and using (6.1), (6.2) and (6.3) yields

(J ρ̂0 (∂t v + v · ∇v), φ)Ωs (t) + (σs , ∇φ)Ωs (t) = (J ρ̂0 f, φ)Ωs (t)
(J(∂t u + v · ∇u), ψ)Ωs (t) = (Jv, ψ)Ωs (t) .

For St. Venant-Kirchhoff materials, the stress-tensor is

1
σ̂s = Jˆ−1 F̂ (2µÊ + λtr(Ê)I)F̂ T , Ê = (F̂ T F̂ − I),
2
In Eulerian coordinates, this transforms to
1
σs = JF −1 (2µE + λtr(E)I)F −T , E = (F −T F −1 − I).
2
For fluid-structure interaction, the interface condition is again continuity of the velocity and
the stress condition:
n · σf = n · σs .

In Fully Eulerian coordinates, we do not need a transformation of the flow domain, thus the
fluid’s deformation uf is not required. In this sense, the Eulerian formulation appears to
be easier than the ALE formulation. The big drawback of the Eulerian approach however
is that the domain partitioned into Ωf (t) and Ωs (t) is not known a priori but is itself an
unknown in the system. By writing the full set of fsi-equations the gap gets obvious

(ρf (∂t v + v · ∇v), φ)Ωf (t) + (σf , ∇φ)Ωf (t) = (ρf f, φ)Ωf (t)
(div v, ξ)Ωf (t) = 0
n · σs = n · σf , vf = vs on Γi (t)
(J ρ̂0s (∂t v + v · ∇v), φ)Ωs (t) + (σs , ∇φ)Ωs (t) = (J ρ̂0s f, φ)Ωs (t)
(J(∂t u + v · ∇u), ψ)Ωs (t) = (Jv, ψ)Ωs (t) .

As in ALE, on different parts of the domain Ω different equations are valid. Opposed to the
ALE formulation, this partitioning is however not known. A certain point x ∈ Ω can either
be in the fluid domain x ∈ Ωf (t) or in the solid domain x ∈ Ωs (t). For the assembly of the
system it is of course required to know which is the domain of influence. This information
however is given by the solution itself: if x − u(x, t) ∈ Ω̂s then x ∈ Ωs (t).
To obtain a simular relation in the flow problem, we need to find a displacement uf of the
fluid-domain. If we denote by u the displacement variable on all Ω, with u = us on Ωs (t)
and u = uf on Ωf (t), we can decide:
(
x ∈ Ωs (t) x − u ∈ Ω̂s
x∈Ω ⇒
x ∈ Ωf (t) x − u 6∈ Ω̂s .

83
6 Alternative Approaches for FSI-problems

Figure 6.2: Non-matching mapping of the flow domain with artificial deformation uf .

We have chosen to check x − u 6∈ Ω̂s instead of x − u ∈ Ω̂f on purpose. This way, we have
more freedom with defining the extension of u to the flow-domain. x − u|Ωf (t) does not need
to assemble a mapping Ωf (t) → Ω̂f . See Figure 6.2 for an example of such a non-matching
deformation. Is is only important that x − u is not mapping into Ω̂s for x ∈ Ωf (t).
For the extension of u to the fluid domain Ωf (t) the only important condition is, that a
u
fluid-particle x ∈ Ωf (t) will be transported within the fluid domain x − → x0 ∈ Ωf (t0 ). Here,
the obvious choice is to use the deformation onto Lagrangian coordinates and to extend u
by:
(J(∂t u + v · ∇u), ψ)Ωf (t) = (Jv, ψ)Ωf (t) .
This extension was discussed and rejected in the context of ALE coordinates: first, fluid
particles can be transported out of the computational domain Ω, whenever there are outflow
or inflow conditions. Second, the deformation u can awkward, e.g. if a rotational flow is pre-
scribed. Here however, this deformation is not used to define a transformation, derivatives,
or the inverse is not required. u is only needed for lookup of the domain of influence. We
can write the implicit system for the Fully Eulerian fsi-problem:
Problem 8. Fluid Structure Interaction in Fully Eulerian Coordinates Find v ∈ v D +
H01 (Ω; ΓD )d , p ∈ L20 (Ω) and u ∈ H01 (Ω; ΓD )d such that

(ρf (∂t v + v · ∇v), φ)Ωf (t) + (Jρ0s (∂t v + v · ∇v), φ)Ωs (t)
+(σf , ∇φ)Ωf (t) + (σs , ∇φ)Ωs (t) = (ρf f, φ)Ωf (t) + (J ρ̂0s f, φ)Ωs (t) ,
(∇ · v, ξ)Ωf (t) + (∇p, ∇ξ)Ωs (t) − hns · ∇p, ξiΓi (t) = 0,
(∂t u + v · ∇u, ψ)Ω = (v, ψ)Ω ,

where ρf is the fluid’s density and ρ̂0s (x̂) is the solid’s density in reference state and with the
stress-tensors as mentioned above. For a point x ∈ Ω we decide:
(
x ∈ Ωs (t) ⇔ x − u ∈ Ω̂s
x∈Ω ⇒
x ∈ Ωf (t) ⇔ x − u 6∈ Ω̂s

This formulation goes without arbitrary transformation. In ALE coordinates, this trans-
formation was the reason for most numerical problems. In particular the flow problems is
always well defined in Fully Eulerian coordinates.

84
 6.1 Fully Eulerian Coordinates

The big disadvantage of the Eulerian scheme is the implicit dependence on the domain. Even
if the material laws would be considered to be linear, the overall problem is nonlinear due to
this dependence. For solving the weak formulation with a Newton’s method, the Jacobian
has to be computed. Derivatives with respect to the deformation include derivatives with
respect to the domain-partitioning. We discuss a simplified fluid tensor given by σ̃f = ∇v.
Then, the viscous term gets:

(σ̃f , ∇φ)Ωf (t) = (∇v, ∇φ)Ωf (t) .

For evaluating the Jacobian, we need to consider derivatives in direction of v and u. The
v-derivative is straightforward:

d 
(∇(v + sη), ∇φ)Ωf (t) s=0 = (∇η, ∇φ)Ωf (t) ,
ds
whereas for the u-derivative the deformation dependence is within the integral boundaries.
We write Ωf (t) = Ωf (u) to get:

d d
 Z 
(∇v, ∇φ)Ωf (u+sη)  = ∇v : ∇φdx .
 
ds s=0 ds Ωf (u+sη) s=0

These derivatives appear in shape-optimization problems and are called shape gradients. For
their evaluation we introduce the characteristic function:
(
1 x̂ ∈ Ω̂f
χ̂(x̂) = ,
0 x̂ 6∈ Ω̂f

and derive the characteristic function of the moving Eulerian domain by:

χ(x, u) = χ̂(x − u).

Then, we can write Z Z

f (x)dx = χ̂(x − u)f (x)dx,
Ωf (u) Ω

to evaluate the derivative as:

d d
Z  Z 
f (x)dx = χ(x, u + sη)f (x)dx .
 
ds Ωf (u+sη) s=0 ds Ω s=0

The characteristic function χ is not differentiable. By Hadamard structure theorem we can

write this derivative as:
d 1
Z  Z Z  Z
χ(x, u+sη)f (x)dx = lim χf (x, u+η)f (x)dx− χf (x, u)f (x)dx = nf ·ηf (x)− dx,

ds Ω s=0 →0  Ω Ω Ωi (u)

where f (x)− is the value of f (x) on the interface Γi (u) as seen from the inside of Ωf (t).
Some terms in the Jacobian of the system include integrals over the interface of the domain.
This interface is moving and crossing mesh elements, see Figure 6.3.

85
6 Alternative Approaches for FSI-problems

Figure 6.3: Left: Mesh partitioning in the Fully Eulerian formulation. The interface Γi (t)
crosses the mesh elements. Right: implicit transformation of the solid domain.
This transformation is well-defined.

A careful integration of these terms is essential. Since the interface crosses through elements
K of the mesh, in different parts of the elements, different equations are valid. Usual Gauss
rules are not appropriate for evaluating the integrals at the interface, since the bi-linear form
to be integrated is not differentiable across the interface. Instead special integration ruled
need to be applied. In the vicinity of the interface, we split the quads into triangles and
perform the integration separately on every triangle, which is now either part of the fluid or
of the structure domain. For further increasing the accuracy, we allow these triangles to be
curved by introducing additional degrees of freedom on the interface. See Figure 6.4 for a
sketch.
In theory, the Fully Eulerian approach is able to handle very large deformation, even contact
of the structure, and this in a fully implicit monolithic formulation. The dependence of the
computational domain partitioning on the solution itself however leads to big problems when
solving the algebraic systems.

Ks Ks
Ks Ks

Kf Kf Kf
Kf

Figure 6.4: Integration close to the interface by using specially adapted integration rules.
Marked points: degrees of freedom eliminated from system.

6.2 A Fixed Mesh Euler-Lagrange Approach

Finally, we present a (possibly monolithic) approach, where both subsystems fluid and struc-
ture are left in their natural coordinate systems. The idea is to formulate the fluid problem
on moving domain in an Eulerian mesh and the structure problem on a fixed Lagrangian
mesh. The coupling on the interface is realized with the help of Lagrange-multipliers. This

86
 6.2 A Fixed Mesh Euler-Lagrange Approach

Ωf (t) Γi Ωs

Figure 6.5: Three subdomains for fluid Ωf , interface Γi and the structure Ωs .

approach has recently been described by Gerstenberger and Wall ([GeWa08a], [GeWa08b],
[GeWa10]).
We split the problem into three sub-problems. On the (moving) fluid domain Ωf (t) we
formulate the Navier-Stokes equations:
)
∂t vt + v · ∇v − ∇ · σf = f
in Ωf (t)
∇·v = 0 (6.4)
v = vΓ i on Γi (t),
where vΓi is the velocity of the interface.
The solid equation is first formulated on the moving Eulerian domain Ωs (t). Here, we find
the deformation ûs as:
ρs dtt us − ∇ · σs = f in Ωs (t)
(6.5)
n · σs = n · σf on Γi (t).
Later on, this equation will be solved in Lagrangian coordinates on a fixed mesh. Finally,
on the interface Γi (t) we have an equation for the velocity of the interface vΓi :
vΓi = dt us on Γi (t). (6.6)
By treating Equations (6.4), (6.5) and (6.6) together, the coupled fsi-problem is solved.
The interface condition is realized as a Dirichlet-condition for the fluid-field, a Neumann-
condition for the interface stresses in the solid equation and lower-dimensional equation for
the interface velocity.
For discretization, the flow problem is discretized in Eulerian formulation on a fixed mesh
Ωh with Ωf (t) ⊂ Ωh . This means, that the interface is moving and is not resolved by the
mesh elements. The structure equation will be transformed to Lagrangian coordinates and
then solved on a fixed Lagrangian mesh Ω̂s,h . The equation on the interface is discretized
on an own mesh Γi,h . Altogether, three different meshes and discretizations are used. The
coupling between the different fields is realized by Lagrange-multipliers in an implicit way.

6.2.1 The Flow Problem

It is necessary to solve the Navier-Stokes equations on a moving domain Ωf (t). Therefore,

the fluid domain is embedded in a larger surrounding, e.g. Ωf (t) ⊂ Ω, the complete com-
putational domain. This domain is triangulated with Ωh . The Navier-Stokes equations are

87
6 Alternative Approaches for FSI-problems

Figure 6.6: Navier-Stokes equations on a moving domain. Shaded part: ficticious fluid do-
main, where solid equations are given. Crossed degrees of freedom: interface
nodes with special test-functions. Boxed degrees of freedom: removed from
system.

only valid on those degrees of freedom inside Ωf (t). See Figure 6.6. The remainder Ωf (t)/Ω
is called the ficticious flow domain. While the next degrees of freedom behind the interface
are included in the computation, all the boxed degrees of freedom (in the figure) are removed
from the computation and velocity and pressure are extended with zero.
On the interface, the velocity is required to fulfill the condition

v = vΓ i on Γi (t),

and in the ficticious domain, we extend v = 0. To allow for this discontinuity in a finite
element simulation we add additional degrees of freedom in the interface nodes (the ones
crossed in Figure 6.6):
Nf NΓi
v i φfi
X X Γ Γ
vh = + v i i φi i ,
i=1 i=1

where the special test-functions φΓi i are discontinuous across the interface and given by
(
 1 in K ∩ Ωf (t)
φΓi i  =

K3xi 0 in K ∩ Ω/Ωf (t).

This concept, of adding specially adapted basis functions to the computation is called XFEM
Xtended F inite E lement M ethod.
for eX
The velocity condition on the interface is realized by means of an Lagrange multiplier λ. We
assume that the interface velocity vΓn+1
i
is known at time tn+1 . Then, discretized in time
with the one-step θ-scheme, the new flow field is given by
n o
(ρf v, φ)Ωf + ∆tθ (ρf v · ∇v, φ)Ωf + (σf , ∇φ)Ωf − hλ, φiΓi = (ρf v n , φ)Ωf + ∆t(θ − 1)An
(∇ · v, ξ) = 0
hv, χλ iΓi = hvΓn+1
i
, χλ iΓi .

where for simplicity by An we describe all the explicit parts evaluated at the old time-step tn .
The Lagrange multiplier λ and the corresponding test-function χλ are living on the interface
Γi only and will be discretized on a lower-dimensional mesh. To understand the meaning of

88
 6.2 A Fixed Mesh Euler-Lagrange Approach

the Lagrange multiplier, we need to integrate the term including the stress-tensor by parts
to get:
· · · − (divσf , φ)Ωf + hn · σf , φiΓi = hλ, φiΓi .
The value of λ is exactly the interface-stress.
The introduction of the Lagrange multiplier is best understood with an easy model problem:
assume, homogenous Dirichlet conditions for the Laplace equation are to be included by
means of a Lagrange multiplier. Then, we aim at minimizing:
1
J(u) = k∇uk2 where u = 0 on Γ.
2
The Lagrange functional is given by

L(u, λ) = J(u) + (λ, u),

and the optimality condition says:

(∇u, ∇φu ) + (λ, φu ) = 0,

(
0 = ∇L ⇒
(u, φλ ) = 0.

6.2.2 The Interface Problem

The interface velocity is given by the simple equation

vΓi = dt uΓi ,

as the temporal derivative of the interfaces deformation. We discretize by the one-step

θ-method to get
un+1
Γi − uΓi
n
= θvΓn+1 + (1 − θ)vΓni .
∆t i

6.2.3 The Structure Problem

The structure problem is likewise coupled to the interface with an additional Lagrange
multiplier µ:

(ρs dtt us , ψ)Ωs + (σs , ∇ψ) − hµ, ψiΓi = 0,

hus , χµ iΓi = huΓi , χµ iΓi .

As mentioned before, the value of the Lagrange multiplier µ is the interface stress and the
balance of forces condition for the coupled problems states

µ = λ.

This system can be discretized in time by splitting into a first-order system and applying
the θ-scheme or by Newmark’s method.

89
6 Alternative Approaches for FSI-problems

6.2.4 The coupled problem

Combining the three subproblems, the system can be solved in a monolithic approach by a
Newton’s method. In every Newton step a linear system of the following type is to be solved:

fv
    
Fvv Fvp Xvλ 0 0 0 v
 Fpv 0 0 0 0 0   p   fp 
    
  λ   fλ 
X 0 0 Xλui 0 0     
 λv
=
 0 0 Xui λ 0 Xui µ 0  uΓi  fuΓi 
  
    
 0 0 0 Xµui 0 Xµus   µ   fµ 
0 0 0 0 Xus µ Sus us us fus

Here, F describes the Navier-Stokes system with couplings between velocity and pressure
unknown, S the structure system. The different X-terms describe the couplings to and from
the Lagrange multipliers on the interface. The equation using the χui test-function of the
interface displacement is used to assure the balance of forces by requiring that the Lagrange
multipliers coincide.
This system can either be solved in a monolithic way or partitioned by splitting into a
Dirichlet-Neumann system. The solution of the coupled system is difficult since it is a
saddle-point system in terms of the flow system (pressure) and in terms of the interface
couplings λ and µ. The resulting method however is able to handle very large deformation,
and has the benefit of formulating the subproblems in their natural coordinates without the
need of transformation.
For the numerical solution the construction of appropriate spaces for the Lagrange multipliers
is a difficult task. As with the inf-sup condition for the Navier-Stokes equations certain
constraints must be fulfilled.

90
Index

added mass operator, 77 material particles, 13

ALE, 35
ALE mapping, 35 Navier-Stokes equations
angular momentum, 21 incompressible, 24
Arbitrary Lagrangian Eulerian, 35 Newtonian fluid, 23

backward Euler, 46 outflow boundary condition, 49

conservative variables, 18 partitioned

iterative staggered, 68
deformation, 13 staggered, 68
deformation gradient, 14 strongly coupled, 68
do nothing boundary condition, 34, 49 weakly coupled, 68
dyadic product, 21 permutation tensor, 21
poisson’s ratio, 30
elasticity law, 29 primitive variables, 18
Eulerian framework, 14 PSPG, 50
finite element method rate-of-strain tensor, 18
conforming, 44 right Cauchy Green tensor, 15
nonconforming, 44
forward Euler, 46 shear stress tensor, 22
fractional-step-theta scheme, 52 St. Venant Kirchoff, 31
Fully Eulerian, 81 stabilization parameter, 48
Stokes Flow, 22
Green strain tensor, 15 strain tensor, 15
homogeneous fluid, 24 linear, 15
hydrostatic pressure, 22 streamline diffusion stabilization, 48
surface forces, 20
incompressible flow, 23 system matrix, 45
isoparametric, 44
isothermal fluid, 23 Taylor-Hood Element, 50
torque, 21
Lagrangian framework, 13 trajectory, 13
Lagrangian strain tensor, 15
Levi-Civita-Tensor, 21 velocity, 13
Local Projection Method, 48 viscosity
LPS, 48 dynamic, 24
kinematic, 24
material body, 13 volume forces, 20

91
Index

XFEM, 88

92