Problem with eigenvalues - Comparison between matlab and a fortran code values.

Hi,

I am using a Fortran code to solve Orr-Sommerfeld equation with Blasius flow and so I have to solve the general problem:

     A*v = lambda*B*v

Where A and B are two complex matrix and lambda are the eigenvalues (complex too). I would like to get the same lamda values with my code and with matlab.

For a 4*4 matrix (A and B) it is ok. But for a 65*65 there is a problem. The code gives to me (I am interested only by lambda with positive imaginary part): lambda_1 = -0.00010 + 0.000086*i ; lambda_2 = 0.011 + 0.00021*i ; lambda_3 = 0.031 + 0.000023*i

And matlab: lambda_1 = - 89.64 + 505.36*i ; lambda_2 = 0.011 + 0.00025*i

I am using eig(A,B) in matlab to solve the problem.

I think that it is nearly good because lambda_2 code and lambda_2 matlab are (nearly) the same and it is the value that I found in the literature. But I don’t understand why I have Lambda_1 = - 89.64 + 505.36*I with matlab.

I tried to solve in matlab:

     det(A-lambda*B)

To see if I get 0 with the different values of lambda (form matlab and fortran) but it doesn’t work (solutions around e+65 or e+226).

Do you have a idea to solve my problem?

Thank you for your time.