Problem with plotting eigenvalues

Question

I want to plot the eigenvalues of a matrix which is dependant on a parameter (well, actually I want to plot a tight-binding electronic band structure). The dimension of the hamitonian matrix is greater than 5, so of course there is no analytical solution. But anyway, when I use Eigenvalues directly, Mathematica will give results containing Root. So I just use these Roots to plot. I thought that theoretically the plot will be fine. Because it is just a root searching process, so if Mathematica searches the roots little by little the eigenvalue curve will be fine. But it turns out that things are not that simple. I give all the code that I have written as follows:

n = 5;
h = Table[0, {i, 1, 2 n}, {j, 1, 2 n}];
a[m_] := 2 m - 1
b[m_] := 2 m
Do[
 h[[b[i], a[i]]] = E^((I k)/2) + E^((-I k)/2);
 If[i + 1 <= n, h[[b[i + 1], a[i]]] = 1, Null];
 If[i - 1 > 0, h[[a[i - 1], b[i]]] = 1, Null];
 h[[a[i], b[i]]] = E^((I k)/2) + E^((-I k)/2)
 , {i, 1, n}]

the above code is to generate the hamiltonian matrix with dimension 2n

eigen = Eigenvalues[h];

Plot[x /. x -> eigen[[1]], {k, 0, \[Pi]}]

the plot is as follows:

It is quite good at the beginning, but totally messed up in the end.So the first question is: What was wrong with the ugly tail?

Finally, about plotting eigenvalues, I only have one more method, that is to calculate the eigenvalues at the discrete point within the range of the parameter and then ListPlot them. But this method has a disadvantage, Because all the eigenvalues are in random order. So which point belongs to which curve can only be judged by eyes, and using this method, I can not plot different eigenvalue curve with different color especially when two curves intersect. So is there better solutions?

Answer 1

Looks like you get small imaginary residuals that are not chopped in V8 (but are in V9).

Plot[eigen[[1]] // Im, {k, 0, \[Pi]}]

Adding a Re (or Chop or similar) gets rid of those for good:

Plot[eigen[[1]] // Re, {k, 0, \[Pi]}]