Research Group of Prof. Dr. M. Griebel
Institute for Numerical Simulation
maximize

Microstructure and Shape Memory Alloys

Home:  Home page of Marcel Arndt
Up:  Front page of Microstructure and Shape Memory Alloys.
Back:  Physical background
Next:  Dynamics of Martensitic Microstructure

Introduction to the Mathematical Modeling of Martensitic Microstructure

Let Omega be a bounded domain in R3 describing the austenite phase of a crystal. Let be an admissible deformation and its gradient. We define the stored energy of the deformation by

where describes the elastic energy density, which is assumed to be continuous and coercive in a suitable sense. One might add surface energies, but we shall neglect this here. For physical reasons, is further assumed to be frame indifferent and invariant with respect to the crystal's symmetry group (cf. [3] p. 197). Typically attains its minimum at a few number of different wells modulo rotation:

so and thus E are nonconvex.

As an example we consider the cubic to tetragonal transformation, where M=3 and

Another example is the orthorombic to monoclinic transformation, where M=2 and

We can now formulate the problem in different ways:

The microscopic approach directly minimizes the stored energy. The problem can be stated as:

Find u in the set of all admissible deformations such that E(u) is minimal.

The direct approach gives both a description of both the macroscopic behavior and the concrete formation of microstructure. Due to the nonconvex structure, the energy functional is usually not lower semicontinuous, and thus the limit of minimizing sequences (if existent) is usually not a minimizer. In fact, many problems of this type lack a minimizer at all, although E is coercive and bounded from below. Numerical simulations are challenging because of the large amount of local minima.

To overcome the problems of direct minimization, one often introduces a relaxation. This leads us to the mesoscopic approach. We enrich the space of deformations by the space of Young measures:

Then problem can be stated as

Minimize  
subject to  

The typical oscillations are now intercepted by the Young measure and describe the microstructure in an abstract way. The averaged deformation u describes the macroscopical behavior. The resulting problem is a constrained linear minimization problem. Numerical simulations require elaborate techniques to handle the high dimensionality due to the Young measure. For detailed information and a comprehensive study of relaxation see [5].

Within the macroscopic approach, the original stored energy density is replaced by its quasiconvex hull. This allows for an accurate description of the macroscopic behavior without oscillations. The price is the complete loss of the microstructure. For many problems this approach is only of theoretic relevance, since it is unknown how to compute the quasiconvex hull.

For further details you might want to have a look at [2], [3] and [4].

Bibliography

[1]  M. Arndt, M. Griebel, T. Roubicek. Modelling and numerical simulation of martensitic transformation in shape memory alloys. Contin. Mech. Thermodyn. 15(5):463-485 (2002).
[2]
C. Carstensen. Numerical analysis of microstructure. Lecture notes 10/2001, MPI Leipzig.
[3]  M. Luskin. On the computation of crystalline microstructure . Acta Numerica 5 (1996), pp. 191-257
[4]  S. Müller. Variational models for microstructure and phase transitions. Lecture notes (1998)
[5]  T. Roubicek. Relaxation in optimization theory and variational calculus. De Gruyter (1997)

Home:  Home page of Marcel Arndt
Up:  Front page of Microstructure and Shape Memory Alloys.
Back:  Physical background
Next:  Dynamics of Martensitic Microstructure