When a real matrix have repeated eigenvalue (i.e. multiplicity >1) and the matrix is not the identity matrix, then it is called defective eigenvalue. 

For example if the eigenvalue is repeated 2 times, there will be only one eigevector associated with it.

There is an algorithm to generate another (linearly independent) eigenvector from this same eigenvalue called the defective eigenvalue method.

My question is: Does Maple have support for finding such additional eigenvectors?  The standard Eigenvectors does not seem to do this. Here is an example.

restart;
A :=Matrix([[1,-2],[2,5]]);
(eigen_values, eigen_vectors) := LinearAlgebra:-Eigenvectors(A);

By hand, using the defective eigenvalue method it is possible find second eigenvector for this eigenvalue, which is linearly independent of the one returned by Maple. Such as 

The method to find these additional eigenvector is described in number of places such as 

https://en.wikipedia.org/wiki/Defective_matrix

And the pdf at the bottom show a more detailed example.

I can code this method by hand to find complete set of L.I. eigenvectors for this defective eigenvalue.  

But before doing that, I thought to ask if Maple have buildin support for this, such as a single command to do this or some package.

I looked at help for LinearAlgebra and I am still not able to see anything, It only says this:

With an eigenvalue of multiplicity  k>1 , there may be fewer than  k linearly independent eigenvectors. In this case, the matrix is called defective.  By design, the returned matrix always has full column dimension.  Therefore, in the defective case, some of the columns that are returned are zero.  Thus, they are not eigenvectors.  With the option, output=list, only eigenvectors are returned.  For more information, see LinearAlgebra[JordanForm] and LinearAlgebra[SchurForm].
 

This PDF also describes using Maple, how to do it for one example

https://wps.prenhall.com/wps/media/objects/884/905485/chapt5/proj5.4/proj5-4.pdf

This PDF also describes how to do it by hand without using Maple, starting at end of page 3

http://www.math.utah.edu/~zwick/Classes/Fall2013_2280/Lectures/Lecture23_with_Examples.pdf

It is not too hard to code this method in Maple. But if Maple allready can do it, using some command, it will be better ofcourse.

Using latest Physics, I found case where it is not giving the Latex.

Please see attached worksheet.  This is on windows 10, Maple 2020.1.1

By trial and error, I found that the error was introduced in Physics 797. Since in Physics 796 it did work.

Here is screen shot

Download latex_issue_11.mw

When I give symgen a HINT, using functional form f(x),g(x)*y it does not generate the infinitesimals of the Lie group for this ODE.

But from the answer given using way=abaco1 it is clear they have this form, where f(x)=-1/x and g(x)=1/x^2

From help, it says

HINT=[e1,e2], indicates to the solver that it should take e1 and e2 as the infinitesimals, where e1 and e2 can contain a maximum of two indeterminate functions. The solver tries to determine the infinitesimals to solve the problem.

And I am using only two indeterminate functions. These are f(x) and g(x)

Am I making a mistake somewhere? Please see worksheet below.

Download symgen_issue_2.mw

Any idea why symgen hangs when using formal algorithm on this first order ode? I was just comparing the Lie symmetries generated for this ode using different algorithms when I noticed this hang on formal.

The textbook gives this result btw

From help on symgen it says

The first algorithms, called formal, formulates a linear PDE system for the infinitesimals [xi, eta], then formally triangularize this system - using differential algebra techniques - finally tacking the resulting uncoupled system. When successful, this algorithm returns the complete set of point symmetries admitted by a given ODE of order 2 or higher. NOTE: this algorithm is advantageous mainly for 2nd and higher order ODEs. The algorithm works as well in the case of first order ODEs, but in this case the subproblems it will need to solve to find the symmetries are as difficult to solve as the first order ODE itself.
 

But on Maple 2020.1 server seems to hang, taking 100% CPU. I had to terminate mserver.exe manually since clicking on interreupt button from worksheet has no effect. This only happens with formal algoritm. All others work very fast. 

I know help says this is meant for second order ODE's, but it also says it works  well for first order.  It seems to be stuck in solve call somewhere. Is this to be expected sometimes when using formal algo in symgen for first order ode's?

Maple can solve the ODE very quickly otherwise.

The following ode also hangs in symgen

ode:=diff(y(x),x)=x*y(x)^2-2*y(x)/x-1/x^3;

Download symgen_issue_1.mw

The following transfer function has zero/pole cancelation. I am trying to create a transfer function object, but Maple automatically simplifies the transfer function before it gets to the DynamicSystem call, which result in different output than what I expected.

I do set the cancellation=false option, even though this is the default. The problem is Maple does pole/zero cancelation before the call.

I tried to add '' around it to delay evaluation, but it did not work.  

restart;
alias(DS=DynamicSystems):
DS:-SystemOptions(cancellation=false,complexfreqvar=s):
tf:=DS:-TransferFunction('-(s - 1)/((-2 + s)*(s - 1))'):
DS:-PrintSystem(tf)

You can see it did pole/zero cancelation.

In Matlab and Mathematica, this does not happen. For example

Clear["Global`*"];
sys = TransferFunctionModel[-(s - 1)/((-2 + s) (s - 1)), s]

Same with Matlab:

>> s=tf('s');
>> sys_tf =-(s - 1)/((-2 + s)*(s - 1))

sys_tf =
 
     -s + 1
  -------------
  s^2 - 3 s + 2
 
Continuous-time transfer function.

What do I need to do in Maple to keep the transfer function without pole/zero cancelation (this affects the state space realization later on when this cancelation happens)

I am using Maple 2019 at this moment as Maple 2020 is busy.