diff --git a/src/differential_equations_1.md b/src/differential_equations_1.md index 261674a8229cb7e4836781754961d08606e37216..492d2a0b54938d9c5e61fd9d4ef62438f36f234a 100644 --- a/src/differential_equations_1.md +++ b/src/differential_equations_1.md @@ -465,14 +465,147 @@ solution are: $\dot{**x**}(t) = e^{**A** t} {**x**}_{0}$, with initial condition $**x**(0) = e^{**A** 0}{**x**}_0 = \mathbbm{1} {**x**}_{0} = {**x**}_{0}$. -Next we will discuss how to determine a solution in general, going beyond the +Next we will discuss how to determine a solution in practice, beyond the formal solution just presented. - +### Case 1: **A** diagonalizable ### + +For an $n \times n$ matrix $**A**$, denote the $n$ distinct eigenvectors as +$\{**v_1**, \cdots, **v**_n \}$. By definition, the eigenvectors satisfy the +equation + +$$**A** **v**_i = \lambda_i **v**_i, \forall i \epsilon \{1, \cdots, n \}. $$ + +Here we give consideration to the case of distinct eigenvectors, in which case +the $n$ eigenvectors form a basis for $\mathbb{R}^{n}$. + +To solve the equation $\dot{**x**}(t) = **A** **x**(t)$, define a set of scalar +functions $\{u_{1}(t), \cdots u_{n}(t) \}$ and make the following ansatz: + +$$**\phi**_{i} = u_{i}(t) **v_{i}**.$$ + +Then, by differentiating, + +$$\dot{**{\phi}**_i}(t) = \dot{u_i}(t) **v**_{i}(t).$$ + +The above equation can be combined with the differential equation for +$**\phi**_{i}(t)$, $\dot{**\phi**_{i}}(t)=**A** **\phi**_{i}(t)$, to derive the +following equations, + +$$\dot{u_i}(t) **v**_{i}(t) = **A** u_{i}(t) **v_{i}**$$ +$$\dot{u_i}(t) **v**_{i}(t) = u_{i}(t) \lambda_{i} **v_{i}** $$ +$$**v**_{i} (\dot{u_i}(t) - \lambda_i u_{i}(t)) = 0, $$ + +where in the second last line we mase use of the fact that $**v**_i$ is an eigenvector +of $**A**$. The obtained relation implies that + +$$\dot{u_i}(t) = \lambda_i u_{i}(t).$$ + +This is a simple differential equation, of the type dealt with in the third +example. The solution is found to be + +$$u_{i}(t) = c_i e^{\lambda_i t},$$ + +with $c_i$ some constant. The general solution is found by adding all $n$ of the +solutions $**\phi_{i}**(t)$, + +$$**x**(t) = c_{1} e^{\lambda_1 t} **v**_{1} + c_{2} e^{\lambda_2 t} **v**_{2} + \cdots + c_{n} e^{\lambda_n t} **v**_{n}.$$ + +and the vectors $\{e^{\lambda_1 t} **v**_{1}, \cdots, e^{\lambda_n t} **v**_{n} \}$ +form a basis for the solution space since $det(**v**_1 | \cdots | **v_n**) \neq 0$ +(the $n$ eigenvectors are linearly independent). + +!!! check "Example: Homogeneous first order linear system with diagonalizable + constant coefficient matrix" + + Define the matrix $**A** = \begin{bmatrix} + 0 & -1 \\ + 1 & 0 \\ + \end{bmatrix}$, and consider the DE + + $$\dot{**x**}(t) = **A** **x**(t), **x_0** = \begin{bmatrix} + 1 \\ + 0 \\ + \end{bmatrix}. $$ + + To proceed following the solution technique, we determine the eigenvalues of + $**A**$, + + $$det {\begin{bmatrix} + -\lambda & -1 \\ + 1 & - \lambda \\ + \end{bmatrix}} = \lambda^2 + 1 = 0. $$ + + By solving the characteristic polynomial, one finds the two eigenvalues + $\lambda_{\pm} = \pm i$. + + Focusing first on the positive eigenvalue, we can determine the first + eigenvector, + + $$\begin{bmatrix} + 0 & -1 \\ + 1 & 0 \\ + \end{bmatrix} \begin{bmatrix} + a \\ + b\\ + \end{bmatrix} = i \begin{bmatrix} + a \\ + b \\ + \end{bmatrix}.$$ + + A solution to this eigenvector equation is given by $a=1$, $b=-i$, altogether + implying that $\lambda_1=i, **v**_{1} = \begin{bmatrix} + 1 \\ + -i \\ + \end{bmatrix}$. + + As for the second eigenvalue, $\lambda_{2} = -i$, we can solve the analagous + eigenvector equation to determine $**v**_{2} = \begin{bmatrix} + 1 \\ + i \\ + \end{bmatrix}$ + + Hence two independent solutions of the differential equation are: + + $$**\phi**_{1} = e^{i t}\begin{bmatrix} + 1 \\ + -i \\ + \end{bmatrix}, **\phi**_{2} = e^{-i t} \begin{bmatrix} + 1 \\ + i \\ + \end{bmatrix}.$$ + + To obtain the general solutoin of the equation, the only thing which is + missing is determination of linear combination coefficients for the two + solutions which allow for satisfaction of the initial condition. To this end, + we must solve + + $$c_1 **\phi**_{1}(t) + c_2 **\phi**_{2}(t) = \begin{bmatrix} + 1 \\ + 0 \\ + \end{bmatrix}$$ + $$\begin{bmatrix} + c_1 + c_2 \\ + -i c_1 + i c_2 \\ + \end{bmatrix} = \begin{bmatrix} + 1 \\ + 0 \\ + \end{bmatrix}.$$ + + The second row of the vector equation for $c_1, c_2$ implies that $c_1=c_2$. + The first row then implies that $c_1=c_2=\frac{1}{2}$. + + Overall then, the general solution of the DE can be summarized + + $$\dot{**x**}(t) = \begin{bmatrix} + \frac{1}{2}(e^{i t} + e^{-i t}) \\ + \frac{1}{2 i}(e^{i t} - e^{-i t}) \\ + \end{bmatrix} = \begin{bmatrix} + cos(t) \\ + sin(t) \\ + \end{bmatrix}. $$ + - - -