Stability Properties of Explicit and Implicit Methods.: PDE To ODE After Semi Discretization

Stability Properties of Explicit and Implicit Methods.
PDE to ODE after semi discretization;

Consider the diffusion equation
(1)
After discretization in space using central finite differences, we obtain a

system of ODEs
or
(2)
with
Solving the PDE is now reduced to solving a system of ODEs.
When we consider a periodic solution in space, the system of

ODEs is reduced to a single ODE. Inserting the periodic solution in equation (2),
and using
and
We get,
or
(3)
Where q is a complex coefficient that depends on the reduced wave number,

, and the locus of q in the complex plane is the Fourier footprint of the
discretized equation. Thus solving a PDE is reduced to solving a single ODE.
Stability of the numerical integration of an ODE
When the method is stable, provided the discretization of the equation is

consistent, the method is convergent, which means the numerical solution
tends towards analytical solution for Δt going to zero.
A discretization scheme is said to be consistent, if the discretized equations
converge to the given differential equations for both the time step and grid size
tending to zero.
Stability + Consistency = Convergence. ------ This is called Lax theorem.
Stability properties of methods for integrating ODEs
Considering the linear test equation,
(4)
Let us consider a couple of examples;

The 4th order Runge-Kutta method applied to the test equation (4)
gives
Considering two step explicit midpoint method and applying it to the test
equation (4),
Consequently, g must be root of
The above examples are weakly stable cases because as per the
stability definition we considered, the solution is bounded and stable
when and not for actual computations with finite values of
Δt.
Next we consider the concept of region of absolute stability;
The region of absolute stability of a numerical algorithm for
integrating an ODE is the set of values of z = qΔt , such that the
sequence un of the numerical values remains bounded as 𝑛 → ∞ .
As previous definition of stability required that the sequence un
remain bounded for 𝑛 → ∞, ∆𝑡 → 0, this is equivalent to stating that
the origin lies in the region of absolute stability
Considering the example of forward time difference (forward Euler)

method applied to the test equation (4)
(5)
The amplification factor is computed as
(6)
Hence we deduce that the region of stability is the region of the

complex plane where and this represented in the Figure as
Region of stability of the forward Euler method.

With q = -1, this means that the condition, where z = qΔt,
is valid for cases of ∆𝑡 ≤ 2 to obtain a bounded numerical solution.
This is not a severe restriction of the time step size, since in order to
obtain the analytical solution, one would have to choose the value of
Δt of the order of 0.1 to obtain satisfactory accuracy. That is, in this
case, Δt is limited by accuracy rather than by stability considerations.
Considering the central finite difference (explicit mid-point) method,
Consequently, g must be a root of
Which corresponds to
From the above, we can deduce that the region of stability is the
segment of the imaginary axis between −𝑖 𝑎𝑛𝑑 + 𝑖 as shown in the
Figure.
Region of stability of the mid-point method.
Absolute Stability
Absolute stability is defined as a property by which the numerical
solution of a physically stable problem would be bounded irrespective
of the time step.
In mathematical terms, this means test problems of the type
Considering the backward Euler method, which is an implicit method,
and applying it to the test equation we get,
If Re(q) < 0, then and if 𝑅𝑒(𝑞) ≥ 0 then |1 − 𝑞∆𝑡| ≤ 1

for values of qΔt less than or equal to 1.
In the z = qΔt plane, this is the region outside of a circle of radius 1
centred at 1, as shown in the Figure.
Region of stability of the backward Euler method.

Thus this method is absolutely stable or A–stable.
Considering the 4th order R-K method, we have
Since qΔt has higher order terms, even if Re(q) is less than zero, the
sum of the four terms of qΔt upto the fourth order need not be less
than zero, and therefore it does not lie entirely in the left half plane.
The region of stability of the 4th order R-K method is given in the
Figure and it is clear that it is not an A stable method.
Region of stability of 4th order R-K method.

Stability Properties of Explicit and Implicit Methods.: PDE To ODE After Semi Discretization

Uploaded by

Copyright:

Available Formats

Stability Properties of Explicit and Implicit Methods.: PDE To ODE After Semi Discretization

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Stability Properties of Explicit and Implicit Methods.: PDE To ODE After Semi Discretization

Uploaded by

Copyright:

Available Formats

Stability Properties of Explicit and Implicit Methods.

PDE to ODE after semi discretization;

After discretization in space using central finite differences, we obtain a

Solving the PDE is now reduced to solving a system of ODEs.

When we consider a periodic solution in space, the system of

Where q is a complex coefficient that depends on the reduced wave number,

Stability of the numerical integration of an ODE

When the method is stable, provided the discretization of the equation is

Let us consider a couple of examples;

Consequently, g must be root of

Considering the example of forward time difference (forward Euler)

The amplification factor is computed as

Hence we deduce that the region of stability is the region of the

Region of stability of the forward Euler method.

Considering the central finite difference (explicit mid-point) method,

Consequently, g must be a root of

If Re(q) < 0, then and if 𝑅𝑒(𝑞) ≥ 0 then |1 − 𝑞∆𝑡| ≤ 1

Region of stability of the backward Euler method.

Region of stability of 4th order R-K method.

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.