So I am trying to compile a summary of the procedure one should follow to find the Jordan basis and the Jordan form of a matrix, and I am on the lookout for free resources online where the algorithm to be followed is clearly explained in an amenable way.

I have found some interesting youtube videos but what I am on the lookout for is a written thing.

Alternatively, if you are so kind as to flesh out the procedure I would be happy to accept that as an answer.

  • Find all the eigenvalues of $T$.

  • For each eigenvalue $\lambda$:

    • Let $N = T-\lambda I$.

    • Compute $N^2, N^3, \dots, N^n$.

    • Find the generalized eigenspace $G=G(\lambda,T)$ of solutions $u$ to $N^n u = 0$.

    • Find a temporary basis for $G$.

    • Let $U_0 = G$, $U_n = \{0\}$ and $B_n = \emptyset$.

    • For $k=n-1,\dots,1,0$:

      • Find $U_k = $range$ (N_{|_G})^k$ by applying $N^k$ to the temporary basis of $G$.

      • From the previous step we have a Jordan basis $B_{k+1}$ to $T_{|_{U_{k+1}}}$ given by $N^{d_1}v_1,\dots,N^2 v_1, N v_1, v_1, \dots, N^{d_m}v_m,\dots,N^2 v_m, N v_m, v_m$, with the property that $N^{d_j+1} v_j = 0$ for all $j$.

      • For $j=1,\dots,m$, find one $u_j$ such that $N u_j = v_j$.

      • Let $\tilde B_k = N^{d_1}v_1,\dots,N^2 v_1, N v_1, v_1, u_1 \dots, N^{d_m}v_m,\dots,N^2 v_m, N v_m, v_m, u_m$. Then $\tilde B_k$ is a Jordan basis for $T$ restricted to its span.

      • Find $A_k$ such that $\tilde B_k \cup A_k$ is a basis for $U_k$.

      • For each $w\in A_k$:

        • Find $x \in {\sf span} \tilde B_k$ such that $Nx = Nw$.

        • Let $u=w-x$, so $Nu=0$.

      • Let $\tilde A_k$ be the set of vectors obtained above.

      • Let $B_k = \tilde B_k \cup A_k$. Then $B_k$ is a Jordan basis for $T_{|_{U_k}}$.

    • In the end, $B_0$ is a Jordan basis for $T_{|_G}$.

  • Recollecting all Jordan bases for each $T_{|_{G(\lambda,T)}}$ produces a Jordan basis for $T$.

I found this method myself by digging into the proof in Axler's Done Right book. I hope it is correct.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.