I am solving a ODE system with k = 3 (algebraic multiplicity) and s=2 (geometric multiplicity): $\dot X(t) = AX(t)$, Where A=

\begin{bmatrix} 2 & 1 & 1 \\ 1 & 2 & 1 \\ -2 & -2 & -1 \end{bmatrix}

When I solve det(A−λ)= 0 I get 1 eigenvalue λ =1 with k=3, then I get 2 eigenvectors \begin{bmatrix} -1 \\ 0 \\ 1 \end{bmatrix} \begin{bmatrix} -1 \\ 1 \\ 0 \end{bmatrix}

Now I am lost, I really dont know what to do next or how to find the general form of X(t).


I found what to do in the case of k=3 and s=1: you solve (A−λ)$V_1$=0, (A−λ)$V_2$=$V_1$ and (A−λ)$V_3$ = $V_2$.

Then $X(t)=C_1V_1e^{λ_1}+c_2e^{λ_1}(V_1t+V_2)+ C_3e^{λ_1}(\frac{V_1t^2}{2}+V_2t+V_3)$.

I know that a part of the solution I'm looking is $C_1V_1e^{λ_1}+ C_2V_2e^{λ_1}$, with $V_1$ and $V_2$ the eigenvetors that I already have found.

So I am looking for the last part of the solution.

  • $\begingroup$ The simplest way is to capitalize on the fact that $A-I$ is nilpotent. More conventionally, compute the Jordan normal form of $A$ and go from there, or somewhat less conventionally, use the Cayley-Hamilton theorem to write $e^{tA}=aI+bA+cA^2$ and solve for the unknown coefficients. $\endgroup$ – amd Nov 29 '19 at 0:31

When a matrix $A$ has only one eigenvalue $\lambda$ like this (and it’s not a multiple of the identity), then $N=A-\lambda I$ is nilpotent. Moreover, $\lambda I$ and $N$ commute, therefore $e^{tA}=e^{t(\lambda I+N)}=e^{\lambda tI}e^{tN}=e^{\lambda t}e^{tN}$. Since $N$ is nilpotent, the power series for $e^{tN}$ only has a finite number of terms. In this case, because the geometric multiplicity of the eigenvalue is two, you know that $N^2=0$, so that $e^{tN}=I+tN$.

Putting this all together, the solution to the differential equation is $X(t) = e^t(I+t(A-I))X(0).$ If you’re not given initial conditions, then $X(0)$ is simply three arbitrary constants.

You can use the same method regardless of the geometric multiplicity of the lone eigenvalue. If it were $1$ instead, the only difference would be that $e^{tN}$ has one more term since $(A-\lambda I)^2\ne0$.

  • $\begingroup$ I think that a lambda is missing in the final answer +1 for nice answer amd $\endgroup$ – Aryadeva Nov 29 '19 at 1:17
  • $\begingroup$ Thanks, this was very helpful. But what if I have 2 eigen values with multiplicity higher than 1? Can I use this method? $\endgroup$ – Jose Avalos Nov 29 '19 at 3:07

When the matrix A is 3 by 3, and the geometrical multiplicity is 2 and algebraic multiplicity is 3 the general solution is: $$\vec x(t)=c_1\vec v_1* \exp(\lambda t)+c_2\vec v_2*\exp(\lambda*t)+c_3\exp(\lambda*t)[t(\alpha\vec v_1 +\beta \vec v_2)+\vec u]$$ When: $$(*)\space(A-\lambda*I)vec(u)=\alpha\vec v_1 +\beta \vec v_2$$ $\alpha$ and $\beta$ are determined along the solution, as when you're solving $(*)$ you will have to set a condition which will assure the solution. Example: $$ \left[ \begin{array}{ccc|c} 1&1&1 &\alpha\\ 0&0&0 & \alpha - \beta\\ 0 & 0 &0 & \alpha - \beta \end{array} \right] $$ The condition for a solutoin is: $\space \alpha - \beta = 0 \space$


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.