Archivo de la etiqueta: matriz fundamental de soluciones

Ecuaciones Diferenciales I: Sistemas lineales no homogéneos – Método de variación de parámetros

Por Omar González Franco

Las ciencias matemáticas exhiben particularmente orden, simetría
y límites; y esas son las más grandes formas de belleza.
– Aristóteles

Introducción

Ya sabemos resolver sistemas lineales homogéneos con coeficientes constantes, en esta entrada estudiaremos el caso no homogéneo.

Como hemos visto en las dos unidades anteriores, el método de variación de parámetros ha sido fundamental para resolver ecuaciones diferenciales en el caso no homogéneo. Éste mismo método es el que desarrollaremos en esta entrada para resolver sistemas lineales no homogéneos con coeficientes constantes.

Cabe mencionar que en esta entrada utilizaremos bastante el concepto de exponencial de una matriz y el de matriz fundamental de soluciones.

Sistemas lineales homogéneos

El sistema lineal que hemos estudiado es

$$\mathbf{Y}^{\prime} = \mathbf{A} \mathbf{Y} \label{1} \tag{1}$$

Donde $\mathbf{A}$ es una matriz con componentes constantes

$$\mathbf{A} = \begin{pmatrix}
a_{11} & a_{12} & \cdots & a_{1n} \\
a_{21} & a_{22} & \cdots & a_{2n} \\
\vdots & & & \vdots \\
a_{n1} & a_{n2} & \cdots & a_{nn}
\end{pmatrix} \label{2} \tag{2}$$

Ahora sabemos que la solución general del sistema lineal homogéneo (\ref{1}) es de la forma

$$\mathbf{Y}_{c}(t) = c_{1} \mathbf{Y}_{1}(t) + c_{2} \mathbf{Y}_{2}(t) + \cdots + c_{n} \mathbf{Y}_{n}(t) \label{3} \tag{3}$$

En donde las funciones $\mathbf{Y}_{1}(t), \mathbf{Y}_{2}(t), \cdots, \mathbf{Y}_{n}(t)$ son soluciones linealmente independientes del mismo sistema. Usamos el subíndice $c$ debido a que, como antes, al resolver el caso no homogéneo será necesario resolver primero el sistema homogéneo asociado y la solución general de dicho sistema será la solución complementaria del sistema no homogéneo.

Recordemos que la matriz que tiene por columnas a las funciones $\mathbf{Y}_{i}(t)$, $i = 1, 2, \cdots, n$ de (\ref{3}) corresponde a la matriz fundamental de soluciones.

$$\hat{\mathbf{Y}}(t) = \begin{pmatrix} \mathbf{Y}_{1}(t) & \mathbf{Y}_{2}(t) & \cdots & \mathbf{Y}_{n}(t) \end{pmatrix} = \begin{pmatrix} y_{11}(t) & y_{12}(t) & \cdots & y_{1n}(t) \\
y_{21}(t) & y_{22}(t) & \cdots & y_{2n}(t) \\
\vdots & & & \vdots \\
y_{n1}(t) & y_{n2}(t) & \cdots & y_{nn}(t)
\end{pmatrix} \label{4} \tag{4}$$

Si definimos el vector de constantes

$$\mathbf{C} = \begin{pmatrix}
c_{1} \\ c_{2} \\ \vdots \\ c_{n}
\end{pmatrix} \label{5} \tag{5}$$

podemos escribir la solución (\ref{3}) como

$$\mathbf{Y}_{c}(t) = \hat{\mathbf{Y}}(t) \mathbf{C} \label{6} \tag{6}$$

Recordemos este resultado para más adelante.

Sistemas lineales no homogéneos

El sistema lineal no homogéneo que intentaremos resolver es de la forma

$$\mathbf{Y}^{\prime} = \mathbf{A} \mathbf{Y} + \mathbf{G} \label{7} \tag{7}$$

Donde $\mathbf{G} = \mathbf{G}(t)$ es una matriz de $n \times 1$ con componentes dependientes de $t$.

$$\mathbf{G}(t) = \begin{pmatrix}
g_{1}(t) \\ g_{2}(t) \\ \vdots \\ g_{n}(t)
\end{pmatrix} \label{8} \tag{8}$$

Dada la forma de la solución general de un sistema lineal homogéneo (\ref{3}), parecería natural pensar que el sistema lineal no homogéneo tiene por solución una función de la forma

$$\mathbf{Y}_{p}(t) = u_{1}(t) \mathbf{Y}_{1}(t) + u_{2}(t) \mathbf{Y}_{2}(t) + \cdots + u_{n}(t) \mathbf{Y}_{n}(t) \label{9} \tag{9}$$

En donde $u_{i}(t)$, $i = 1, 2, \cdots, n$ son funciones escalares de $t$ derivables y las funciones $\mathbf{Y}_{i}(t)$, $i = 1, 2, \cdots, n$ forman una matriz fundamental de soluciones $\hat{\mathbf{Y}}(t)$. Si definimos el vector

$$\mathbf{U}(t) = \begin{pmatrix}
u_{1}(t) \\ u_{2}(t) \\ \vdots \\ u_{n}(t)
\end{pmatrix} \label{10} \tag{10}$$

Entonces la solución propuesta (\ref{9}) adquiere la forma

$$\mathbf{Y}_{p}(t) = \hat{\mathbf{Y}}(t) \mathbf{U}(t) \label{11} \tag{11}$$

El método de variación de parámetros nos permitirá obtener la forma del vector $\mathbf{U}(t)$, una vez obtenida podremos formar la solución general del sistema lineal no homogéneo (\ref{7}) que, como siempre, será la superposición de la solución complementaria del sistema lineal homogéneo asociado $\mathbf{Y}_{c}(t)$ más la solución particular del sistema lineal no homogéneo $\mathbf{Y}_{p}(t)$, esto es

$$\mathbf{Y}(t) = \mathbf{Y}_{c}(t) + \mathbf{Y}_{p}(t) \label{12} \tag{12}$$

Variación de parámetros

Comencemos a desarrollar el método de variación de parámetros, como mencionamos antes, el objetivo es encontrar la forma explícita del vector (\ref{10}) para formar la solución particular del sistema lineal no homogéneo.

Consideremos la función propuesta (\ref{11}) y derivémosla.

$$\mathbf{Y}_{p}^{\prime}(t) = \hat{\mathbf{Y}}^{\prime}(t) \mathbf{U}(t) + \hat{\mathbf{Y}}(t) \mathbf{U}^{\prime}(t) \label{13} \tag{13}$$

Si sustituimos (\ref{11}) y (\ref{13}) en el sistema lineal no homogéneo (\ref{7}), se tiene

$$\hat{\mathbf{Y}}^{\prime}(t) \mathbf{U}(t) + \hat{\mathbf{Y}}(t) \mathbf{U}^{\prime}(t) = \mathbf{A} [\hat{\mathbf{Y}}(t) \mathbf{U}(t)] + \mathbf{G}(t) \label{14} \tag{14}$$

Como $\hat{\mathbf{Y}}(t)$ es una matriz fundamental de soluciones sabemos que satisface el sistema homogéneo, es decir,

$$\hat{\mathbf{Y}}^{\prime}(t) = \mathbf{A} \hat{\mathbf{Y}}(t) \label{15} \tag{15}$$

Si sustituimos en (\ref{14}) la ecuación queda como

$$\mathbf{A} \hat{\mathbf{Y}}(t) \mathbf{U}(t) + \hat{\mathbf{Y}}(t) \mathbf{U}^{\prime}(t) = \mathbf{A} \hat{\mathbf{Y}}(t) \mathbf{U}(t) + \mathbf{G}(t)$$

O bien,

$$\hat{\mathbf{Y}}(t) \mathbf{U}^{\prime}(t) = \mathbf{G}(t) \label{16} \tag{16}$$

La matriz fundamental es no singular, de manera que siempre existe su inversa, esto nos permite establecer que

$$\mathbf{U}^{\prime}(t) = \hat{\mathbf{Y}}^{ -1}(t) \mathbf{G}(t) \label{17} \tag{17}$$

Esta ecuación es matricial y sabemos que es posible integrar sobre matrices, así que integremos la ecuación anterior con el objetivo de hallar la forma de $\mathbf{U}$.

$$\mathbf{U}(t) = \int \hat{\mathbf{Y}}^{ -1}(t) \mathbf{G}(t) dt \label{18} \tag{18}$$

Ahora que conocemos la forma de $\mathbf{U}(t)$, sustituimos en la solución propuesta (\ref{11}), de forma que una solución particular del sistema lineal no homogéneo es

$$\mathbf{Y}_{p}(t) = \hat{\mathbf{Y}}(t) \int \hat{\mathbf{Y}}^{ -1}(t) \mathbf{G}(t) dt \label{19} \tag{19}$$

Por lo tanto, de (\ref{6}) y el resultado (\ref{19}) concluimos que la solución general del sistema lineal no homogéneo es

$$\mathbf{Y}(t) = \hat{\mathbf{Y}}(t) \mathbf{C} + \hat{\mathbf{Y}}(t) \int \hat{\mathbf{Y}}^{ -1}(t) \mathbf{G}(t) dt \label{20} \tag{20}$$

Si $\hat{\mathbf{Y}}(t)$ es la matriz fundamental de soluciones $e^{\mathbf{A} t}$ y considerando que $\hat{\mathbf{Y}}^{-1}(t) = e^{-\mathbf{A} t}$, el resultado anterior queda como

$$\mathbf{Y}(t) = e^{\mathbf{A} t} \mathbf{C} + e^{\mathbf{A}t} \int e^{-\mathbf{A} t} \mathbf{G}(t) dt \label{21} \tag{21}$$

Problema con valores iniciales

Consideremos el problema con valores iniciales

$$\mathbf{Y}^{\prime} = \mathbf{A} \mathbf{Y} + \mathbf{G}; \hspace{1cm} \mathbf{Y}(t_{0}) = \mathbf{Y}_{0} \label{22} \tag{22}$$

De nuestro desarrollo anterior consideremos la relación (\ref{17}).

$$\mathbf{U}^{\prime}(t) = \hat{\mathbf{Y}}^{ -1}(t) \mathbf{G}(t)$$

En esta ocasión integremos de $t_{0}$ a $t$ y usemos el teorema fundamental del cálculo.

$$\mathbf{U}(t) = \mathbf{U}(t_{0}) + \int_{t_{0}}^{t} \hat{\mathbf{Y}}^{ -1}(s) \mathbf{G}(s) ds \label{23} \tag{23}$$

Notemos que si aplicamos la condición inicial $\mathbf{Y}(t_{0}) = \mathbf{Y}_{0}$ sobre la función (\ref{11}), se obtiene

$$\mathbf{Y}_{p}(t_{0}) = \hat{\mathbf{Y}}(t_{0}) \mathbf{U}(t_{0}) = \mathbf{Y}_{0} \label{24} \tag{24}$$

De donde,

$$\mathbf{U}(t_{0}) = \hat{\mathbf{Y}}^{ -1}(t_{0}) \mathbf{Y}_{0} \label{25} \tag{25}$$

Sustituimos este resultado en la ecuación (\ref{23}).

$$\mathbf{U}(t) = \hat{\mathbf{Y}}^{ -1}(t_{0}) \mathbf{Y}_{0} + \int_{t_{0}}^{t} \hat{\mathbf{Y}}^{ -1}(s) \mathbf{G}(s) ds \label{26} \tag{26}$$

Aquí debemos tener cuidado, si sustituimos la función (\ref{26}) en (\ref{11}), se obtiene

$$\mathbf{Y}(t) = \hat{\mathbf{Y}}(t) \hat{\mathbf{Y}}^{ -1}(t_{0}) \mathbf{Y}_{0} + \hat{\mathbf{Y}}(t) \int_{t_{0}}^{t} \hat{\mathbf{Y}}^{ -1}(s) \mathbf{G}(s) ds \label{27} \tag{27}$$

Pero $\hat{\mathbf{Y}}^{ -1}(t_{0}) \mathbf{Y}_{0}$ es una matriz de constantes, digamos

$$\hat{\mathbf{Y}}^{ -1}(t_{0}) \mathbf{Y}_{0} = \mathbf{C}$$

Entonces el resultado (\ref{27}) queda como

$$\mathbf{Y}(t) = \hat{\mathbf{Y}}(t) \mathbf{C} + \hat{\mathbf{Y}}(t) \int_{t_{0}}^{t} \hat{\mathbf{Y}}^{ -1}(s) \mathbf{G}(s) ds \label{28} \tag{28}$$

Este resultado se parece a la ecuación (\ref{20}), es decir, a pesar de que sustituimos (\ref{26}) en (\ref{11}) esperando obtener la solución particular $\mathbf{Y}_{p}(t)$, en realidad estamos obteniendo la solución general, la solución general del problema de valores iniciales.

Si consideramos nuevamente que $\hat{\mathbf{Y}}(t) = e^{\mathbf{A} t}$, el resultado (\ref{27}) se reduce significativamente.

$$\mathbf{Y}(t) = e^{\mathbf{A} t} e^{-\mathbf{A} t_{0}} \mathbf{Y}_{0} + e^{\mathbf{A}t} \int_{t_{0}}^{t} e^{-\mathbf{A} s} \mathbf{G}(s) ds \label{29} \tag{29}$$

O bien,

$$\mathbf{Y}(t) = e^{\mathbf{A}(t -t_{0})} \mathbf{Y}_{0} + \int_{t_{0}}^{t}e^{\mathbf{A} (t -s)} \mathbf{G}(s) ds \label{30} \tag{30}$$

Por otro lado, si $t_{0} = 0$, de (\ref{27}) se obtiene que

$$\mathbf{Y}(t) = \hat{\mathbf{Y}}(t) \hat{\mathbf{Y}}^{ -1}(0) \mathbf{Y}_{0} + \hat{\mathbf{Y}}(t) \int_{0}^{t} \hat{\mathbf{Y}}^{ -1}(s) \mathbf{G}(s) ds$$

Pero recordemos que

$$e^{\mathbf{A} t} = \hat{\mathbf{Y}}(t) \hat{\mathbf{Y}}^{-1}(0) \label{31} \tag{31}$$

Entonces la solución anterior queda como

$$\mathbf{Y}(t) = e^{\mathbf{A} t} \mathbf{Y}_{0} + \hat{\mathbf{Y}}(t) \int_{0}^{t} \hat{\mathbf{Y}}^{ -1}(s) \mathbf{G}(s) ds \label{32} \tag{32}$$

Si nuestro propósito es determinar la solución general de un sistema lineal no homogéneo (\ref{7}), primero resolvemos el sistema lineal homogéneo asociado para obtener la solución complementaria en la forma (\ref{3}). Con las funciones $\mathbf{Y}_{i}(t)$, $i = 1, 2, \cdots, n$ obtenidas formamos una matriz fundamental $\hat{\mathbf{Y}}(t)$, se calcula su inversa y se sustituyen las matrices correspondientes en la solución particular (\ref{19}). Una vez obtenidas ambas soluciones, la solución general del sistema lineal no homogéneo será

$$\mathbf{Y}(t) = c_{1} \mathbf{Y}_{1}(t) + c_{2} \mathbf{Y}_{2}(t) + \cdots + c_{n} \mathbf{Y}_{n}(t) + \mathbf{Y}_{p}(t) = \mathbf{Y}_{c}(t) + \mathbf{Y}_{p}(t)$$

Sin embargo, si lo que tenemos es un problema de valores iniciales, debemos nuevamente obtener la solución del sistema lineal homogéneo asociado ya que eso es lo que nos permite formar la matriz fundamental de soluciones $\hat{\mathbf{Y}}(t)$, una vez obtenida esta función calculamos su inversa y se sustituyen las matrices correspondientes en la ecuación (\ref{27}), esto nos dará la solución completa del problema de valores iniciales, es decir, no es necesario aplicar las condiciones iniciales en la solución complementaria para obtener los valores de las constantes $c_{1}, c_{2}, \cdots, c_{n}$.

Para concluir con esta entrada realicemos dos ejemplos, en el primero de ellos obtendremos la solución general de un sistema lineal no homogéneo y en el segundo ejemplo resolveremos un problema con valores iniciales. Con estos ejemplos se espera que el método quede claro.

Ejemplo: Obtener la solución general del siguiente sistema lineal no homogéneo.

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
0 & 2 \\ -1 & 3
\end{pmatrix} \mathbf{Y} + \begin{pmatrix}
1 \\ -1
\end{pmatrix} e^{t}$$

Solución: Resolvamos primero el sistema homogéneo asociado.

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
0 & 2 \\ -1 & 3
\end{pmatrix} \mathbf{Y}$$

En este caso la matriz $\mathbf{A}$ es

$$\mathbf{A} = \begin{pmatrix}
0 & 2 \\ -1 & 3
\end{pmatrix}$$

Determinemos los valores y vectores propios de esta matriz.

$$|\mathbf{A} -\lambda \mathbf{I}| = \begin{vmatrix}
-\lambda & 2 \\ -1 & 3 -\lambda
\end{vmatrix} = 0$$

La ecuación característica es

$$\lambda^{2} -3 \lambda + 2 = 0$$

Resolviendo para $\lambda$ se obtiene que los valores propios son

$$\lambda_{1} = 1\hspace{1cm} y \hspace{1cm} \lambda_{2} = 2$$

Determinemos los vectores propios correspondientes a cada valor propio.

Caso 1: $\lambda_{1} = 1$.

Buscamos un vector $\mathbf{K} \neq \mathbf{0}$, tal que

$$(\mathbf{A} -\mathbf{I}) \mathbf{K} = \begin{pmatrix}
-1 & 2 \\ -1 & 2
\end{pmatrix} \begin{pmatrix}
k_{1} \\ k_{2}
\end{pmatrix} = \begin{pmatrix}
0 \\ 0
\end{pmatrix}$$

La ecuación que se obtiene es

$$-k_{1} + 2 k_{2} = 0$$

Es decir, $2 k_{2} = k_{1}$. Elegimos $k_{2} = 1$, entonces $k_{1} = 2$. Por lo tanto el primer vector propio es

$$\mathbf{K}_{1} = \begin{pmatrix}
2 \\ 1
\end{pmatrix}$$

Caso 2: $\lambda_{2} = 2$.

Buscamos un vector $\mathbf{K} \neq \mathbf{0}$, tal que

$$(\mathbf{A} -2 \mathbf{I}) \mathbf{K} = \begin{pmatrix}
-2 & 2 \\ -1 & 1
\end{pmatrix} \begin{pmatrix}
k_{1} \\ k_{2}
\end{pmatrix} = \begin{pmatrix}
0 \\ 0
\end{pmatrix}$$

La ecuación que se obtiene es

$$-k_{1} + k_{2} = 0$$

Es decir, $k_{1} = k_{2}$. Elegimos $k_{1} = k_{2} = 1$. Por lo tanto, el segundo vector propio es

$$\mathbf{K}_{2} = \begin{pmatrix}
1 \\ 1
\end{pmatrix}$$

Con estos resultados concluimos que la solución general del sistema lineal homogéneo asociado es

$$\mathbf{Y}_{c}(t) = c_{1} e^{t} \begin{pmatrix}
2 \\ 1
\end{pmatrix} + c_{2} e^{2t} \begin{pmatrix}
1 \\ 1
\end{pmatrix}$$

Para determinar la solución particular $\mathbf{Y}_{p}(t)$, formemos, con el resultado anterior, la matriz fundamental de soluciones (\ref{4}).

$$\hat{\mathbf{Y}}(t) = \begin{pmatrix}
2e^{t} & e^{2t} \\ e^{t} & e^{2t}
\end{pmatrix}$$

Como también requerimos de la matriz inversa, verifica que

$$\hat{\mathbf{Y}}^{-1}(t) = \begin{pmatrix}
e^{-t} & -e^{-t} \\ -e^{-2t} & 2e^{-2t}
\end{pmatrix}$$

La matriz $\mathbf{G}$ en este caso es

$$ \mathbf{G}(t) = \begin{pmatrix}
e^{t} \\ -e^{t}
\end{pmatrix}$$

Sustituyamos estas matrices en la solución particular (\ref{19}).

\begin{align*}
\mathbf{Y}_{p}(t) &= \begin{pmatrix}
2e^{t} & e^{2t} \\ e^{t} & e^{2t}
\end{pmatrix} \int \begin{pmatrix}
e^{-t} & -e^{-t} \\ -e^{-2t} & 2e^{-2t}
\end{pmatrix} \begin{pmatrix}
e^{t} \\ -e^{t}
\end{pmatrix} dt \\
&= \begin{pmatrix}
2e^{t} & e^{2t} \\ e^{t} & e^{2t}
\end{pmatrix} \int \begin{pmatrix}
2 \\ -3e^{-t}
\end{pmatrix} dt
\end{align*}

Resolviendo la integral (sin considerar constantes de integración), se obtiene

$$\int \begin{pmatrix}
2 \\ -3e^{-t}
\end{pmatrix} dt = \begin{pmatrix}
2t \\ 3e^{-t}
\end{pmatrix}$$

Entonces,

$$\mathbf{Y}_{p}(t) = \begin{pmatrix}
2e^{t} & e^{2t} \\ e^{t} & e^{2t}
\end{pmatrix} \begin{pmatrix}
2t \\ 3e^{-t}
\end{pmatrix} = \begin{pmatrix}
4t e^{t} + 3e^{t} \\ 2te^{t} + 3e^{t}
\end{pmatrix}$$

Esto es,

$$\mathbf{Y}_{p}(t) =\begin{pmatrix}
4 \\ 2
\end{pmatrix} te^{t} + \begin{pmatrix}
3 \\ 3
\end{pmatrix} e^{t}$$

Por lo tanto, la solución general del sistema lineal no homogéneo es

$$\mathbf{Y}(t) = c_{1} e^{t} \begin{pmatrix}
2 \\ 1
\end{pmatrix} + c_{2} e^{2t} \begin{pmatrix}
1 \\ 1
\end{pmatrix} + \begin{pmatrix}
3 \\ 3
\end{pmatrix}e^{t} + \begin{pmatrix}
4 \\ 2
\end{pmatrix} t e^{t}$$

$\square$

Realicemos ahora un problema con valores iniciales.

Ejemplo: Resolver el siguiente problema con valores iniciales.

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
1 & 0 & 0 \\ 2 & 1 & -2 \\ 3 & 2 & 1
\end{pmatrix} \mathbf{Y} + \begin{pmatrix}
0 \\ 0 \\ e^{t} \cos(2t)
\end{pmatrix}, \hspace{1cm} \mathbf{Y}(0) = \begin{pmatrix}
0 \\ 1 \\ 1
\end{pmatrix}$$

Solución: Primero debemos obtener la solución del sistema lineal homogéneo asociado

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
1 & 0 & 0 \\ 2 & 1 & -2 \\ 3 & 2 & 1
\end{pmatrix} \mathbf{Y}$$

Ello nos permitirá obtener la matriz fundamental de soluciones. En este caso la matriz $\mathbf{A}$ es

$$\mathbf{A} = \begin{pmatrix}
1 & 0 & 0 \\ 2 & 1 & -2 \\ 3 & 2 & 1
\end{pmatrix}$$

Determinemos los valores y vectores propios de esta matriz. La ecuación característica se obtendrá de calcular el determinante

$$|\mathbf{A} -\lambda \mathbf{I}| = \begin{vmatrix}
1-\lambda & 0 & 0 \\ 2 & 1 -\lambda & -2 \\ 3 & 2 & 1 -\lambda
\end{vmatrix} = 0$$

Desarrollando el determinante obtendremos que

$$(1 -\lambda )(\lambda ^{2} -2 \lambda + 5) = 0$$

Resolviendo para $\lambda$ se obtiene que los valores propios de $\mathbf{A}$ son

$$\lambda_{1} = 1 \hspace{1cm} y \hspace{1cm} \lambda_{2} = 1 + 2i, \hspace{1cm} \lambda_{3} = 1 -2i$$

De acuerdo a los valores propios obtenidos, la manera de resolver el sistema homogéneo será aplicando la teoría vista en la entrada sobre valores propios complejos.

Determinemos los vectores propios correspondientes a cada valor propio.

Caso 1: $\lambda_{1} = 1$.

Buscamos un vector $\mathbf{K} \neq \mathbf{0}$, tal que

$$(\mathbf{A} -\mathbf{I}) \mathbf{K} = \left[ \begin{pmatrix}
1 & 0 & 0 \\ 2 & 1 & -2 \\ 3 & 2 & 1
\end{pmatrix} -\begin{pmatrix}
1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1
\end{pmatrix} \right] \begin{pmatrix}
k_{1} \\ k_{2} \\ k_{3}
\end{pmatrix} = \begin{pmatrix}
0 & 0 & 0 \\ 2 & 0 & -2 \\ 3 & 2 & 0
\end{pmatrix} \begin{pmatrix}
k_{1} \\ k_{2} \\ k_{3}
\end{pmatrix} = \begin{pmatrix}
0 \\ 0 \\ 0
\end{pmatrix}$$

El sistema que se obtiene es

\begin{align*}
2 k_{1} -2 k_{3} &= 0 \\
3 k_{1} + 2 k_{2} &= 0
\end{align*}

De este sistema se observa que

\begin{align*}
k_{1} &= k_{3} \\
k_{2} &= -\dfrac{3k_{1}}{2}
\end{align*}

Elegimos $k_{1} = 2 = k_{3}$, de tal manera que $k_{2} = -3$, así el primer vector propio es

$$\mathbf{K}_{1} = \begin{pmatrix}
2 \\ -3 \\ 2
\end{pmatrix}$$

Caso 2: $\lambda_{2} = 1 + 2i$.

Buscamos un vector $\mathbf{K}$, diferente de cero, tal que

$$[\mathbf{A} -(1 + 2i) \mathbf{I}] \mathbf{K} = \begin{pmatrix}
-2i & 0 & 0 \\ 2 & -2i & -2 \\ 3 & 2 & -2i
\end{pmatrix} \begin{pmatrix}
k_{1} \\ k_{2} \\ k_{3}
\end{pmatrix} = \begin{pmatrix}
0 \\ 0 \\ 0
\end{pmatrix}$$

El sistema de ecuaciones que se obtiene es

\begin{align*}
-2i k_{1} &= 0 \\
2k_{1} -2i k_{2} -2k_{3} &= 0 \\
3k_{1} + 2k_{2} -2i k_{3} &= 0
\end{align*}

De este sistema se observa que $k_{1} = 0$ y $k_{3} = -ik_{2}$. Elegimos $k_{2} = 1$, de manera que el segundo vector propio sea

$$\mathbf{K}_{2} = \begin{pmatrix}
0 \\ 1 \\ -i
\end{pmatrix}$$

Caso 3: $\lambda_{2} = 1 -2i$.

Sabemos que este caso es el conjugado del caso anterior, por lo que directamente establecemos que el tercer vector propio es

$$\mathbf{K}_{3} = \begin{pmatrix}
0 \\ 1 \\ i
\end{pmatrix}$$

La solución general del sistema lineal homogéneo asociado, en su forma compleja, es

$$\mathbf{Y}_{c}(t) = c_{1} e^{t} \begin{pmatrix}
2 \\ -3 \\ 2
\end{pmatrix} + c_{2} e^{(1 + 2i) t} \begin{pmatrix}
0 \\ 1 \\ -i
\end{pmatrix} + c_{3} e^{(1 -2i) t} \begin{pmatrix}
0 \\ 1 \\ i
\end{pmatrix}$$

Sin embargo esta solución no nos sirve de mucho, pues desearíamos construir la matriz fundamental de soluciones con valores reales. Recordando lo visto en la entrada sobre valores propios complejos, podemos encontrar dos funciones $\mathbf{W}_{1}(t)$ y $\mathbf{W}_{2}(t)$, tal que la solución general sea de la forma

$$\mathbf{Y}_{c}(t) = c_{1} e^{\lambda_{1} t} + c_{2} \mathbf{W}_{1}(t) + c_{3} \mathbf{W}_{2}(t) \label{33} \tag{33}$$

Recordemos que las funciones $\mathbf{W}_{1}(t)$ y $\mathbf{W}_{2}(t)$ están dadas por

$$\mathbf{W}_{1}(t) = e^{\alpha t} [\mathbf{U} \cos(\beta t) -\mathbf{V} \sin(\beta t)] \label{34} \tag{34}$$

y

$$\mathbf{W}_{2}(t) = e^{\alpha t} [\mathbf{U} \sin(\beta t) + \mathbf{V} \cos(\beta t)] \label{35} \tag{35}$$

Consideremos el caso 2 en el que $\lambda_{2} = 1 + 2i$ y

$$\mathbf{K}_{2} = \begin{pmatrix}
0 \\ 1 \\ -i
\end{pmatrix} = \begin{pmatrix}
0 \\ 1 \\ 0
\end{pmatrix} + i \begin{pmatrix}
0 \\ 0 \\ -1
\end{pmatrix} $$

De estos resultados obtenemos que $\alpha = 1$, $\beta = 2$ y

$$\mathbf{U} = \begin{pmatrix}
0 \\ 1 \\ 0
\end{pmatrix} \hspace{1cm} y \hspace{1cm} \mathbf{V} = \begin{pmatrix}
0 \\ 0 \\ -1
\end{pmatrix}$$

Sustituyamos en la funciones (\ref{34}) y (\ref{35}). Por un lado,

\begin{align*}
\mathbf{W}_{1}(t) &= e^{t} \left[ \begin{pmatrix}
0 \\ 1 \\ 0
\end{pmatrix} \cos(2t) -\begin{pmatrix}
0 \\ 0 \\ -1
\end{pmatrix} \sin(2t) \right] \\
&= e^{t} \left[ \begin{pmatrix}
0 \\ \cos(2t) \\ 0
\end{pmatrix} -\begin{pmatrix}
0 \\ 0 \\ -\sin(2t)
\end{pmatrix} \right]
\end{align*}

Esto es,

$$\mathbf{W}_{1}(t) = e^{t} \begin{pmatrix}
0 \\ \cos(2t) \\ \sin(2t)
\end{pmatrix}$$

Por otro lado,

\begin{align*}
\mathbf{W}_{2}(t) &= e^{t} \left[ \begin{pmatrix}
0 \\ 1 \\ 0
\end{pmatrix} \sin(2t) + \begin{pmatrix}
0 \\ 0 \\ -1
\end{pmatrix} \cos(2t) \right] \\
&= e^{t} \left[ \begin{pmatrix}
0 \\ \sin(2t) \\ 0
\end{pmatrix} + \begin{pmatrix}
0 \\ 0 \\ -\cos(2t)
\end{pmatrix} \right]
\end{align*}

Esto es,

$$\mathbf{W}_{2}(t) = e^{t} \begin{pmatrix}
0 \\ \sin(2t) \\ -\cos(2t)
\end{pmatrix}$$

Recordemos que estas funciones también se pueden obtener considerando la identidad de Euler. Del caso 2 la solución que se obtiene es

$$\mathbf{Y}_{2c}(t) = e^{(1 + 2i) t} \begin{pmatrix}
0 \\ 1 \\ -i
\end{pmatrix}$$

Así,

$$\mathbf{W}_{1}(t) = Re \{ \mathbf{Y}_{2c}(t) \} \hspace{1cm} y \hspace{1cm} \mathbf{W}_{2}(t) = Im \{ \mathbf{Y}_{2c}(t) \}$$

Usando la identidad de Euler sobre esta solución obtenemos lo siguiente.

\begin{align*}
e^{(1 + 2i)t} \begin{pmatrix}
0 \\ 1 \\ -i
\end{pmatrix} &= e^{t}[\cos(2t) + i \sin(2t)] \left[ \begin{pmatrix}
0 \\ 1 \\ 0
\end{pmatrix} -i \begin{pmatrix}
0 \\ 0 \\ 1
\end{pmatrix} \right] \\
&= e^{t} \left[ \cos(2t) \begin{pmatrix}
0 \\ 1 \\ 0
\end{pmatrix} + \sin(2t) \begin{pmatrix}
0 \\ 0 \\ 1
\end{pmatrix} \right] + i e^{t} \left[ \sin(2t) \begin{pmatrix}
0 \\ 1 \\ 0
\end{pmatrix} -\cos(2t) \begin{pmatrix}
0 \\ 0 \\ 1
\end{pmatrix} \right] \\
&= e^{t} \left[ \begin{pmatrix}
0 \\ \cos(2t) \\ 0
\end{pmatrix} + \begin{pmatrix}
0 \\ 0 \\ \sin(2t)
\end{pmatrix} \right] + ie^{t} \left[ \begin{pmatrix}
0 \\ \sin(2t) \\ 0
\end{pmatrix} + \begin{pmatrix}
0 \\ 0 \\ -\cos(2t)
\end{pmatrix} \right]
\end{align*}

De donde,

$$\mathbf{W}_{1}(t) = e^{t} \begin{pmatrix}
0 \\ \cos(2t) \\ \sin(2t)
\end{pmatrix} \hspace{1cm} y \hspace{1cm} \mathbf{W}_{2}(t) = e^{t} \begin{pmatrix}
0 \\ \sin(2t) \\ -\cos(2t)
\end{pmatrix}$$

De esta forma, la solución general del sistema lineal homogéneo asociado es

$$\mathbf{Y}_{c}(t) = c_{1} e^{t} \begin{pmatrix}
2 \\ -3 \\ 2
\end{pmatrix} + c_{2} e^{t} \begin{pmatrix}
0 \\ \cos(2t) \\ \sin(2t)
\end{pmatrix} + c_{3} e^{t} \begin{pmatrix}
0 \\ \sin(2t) \\ -\cos(2t)
\end{pmatrix} \label{36} \tag{36}$$

Esta solución es de la forma (\ref{3}) por lo que la matriz fundamental de soluciones, formada por estos vectores linealmente independientes, es

$$\hat{\mathbf{Y}}(t) = \begin{pmatrix}
2e^{t} & 0 & 0 \\ -3e^{t} & e^{t} \cos(2t) & e^{t} \sin(2t) \\ 2e^{t} & e^{t} \sin(2t) & -e^{t} \cos(2t)
\end{pmatrix}$$

Para obtener la solución del problema con valores iniciales usaremos el resultado (\ref{29}) para $t_{0} = 0$.

$$\mathbf{Y}(t) = e^{\mathbf{A} t} \mathbf{Y}_{0} + e^{\mathbf{A}t} \int_{0}^{t} e^{-\mathbf{A} s} \mathbf{G}(s) ds \label{37} \tag{37}$$

Es decir, consideraremos a la matriz $e^{\mathbf{A} t}$ como la matriz fundamental de soluciones. También es posible usar la relación (\ref{32}) usando la matriz $\hat{\mathbf{Y}}(t)$ antes establecida. ¿Por qué son equivalentes ambas formas?.

Determinemos la matriz $e^{\mathbf{A} t}$ usando la relación (\ref{31}). Si evaluamos $t = 0$ en la matriz $\hat{\mathbf{Y}}(t)$ se obtiene la matriz

$$\hat{\mathbf{Y}}(0) = \begin{pmatrix}
2 & 0 & 0 \\ -3 & 1 & 0 \\ 2 & 0 & -1
\end{pmatrix}$$

Comprueba que la matriz inversa es

$$\hat{\mathbf{Y}}^{ -1}(0) = \begin{pmatrix}
2 & 0 & 0 \\ -3 & 1 & 0 \\ 2 & 0 & -1
\end{pmatrix}^{ -1} = \begin{pmatrix}
\dfrac{1}{2} & 0 & 0 \\ \dfrac{3}{2} & 1 & 0 \\ 1 & 0 & -1
\end{pmatrix}$$

Sustituyamos en (\ref{31}).

\begin{align*}
e^{\mathbf{A}t} &= \begin{pmatrix}
2e^{t} & 0 & 0 \\ -3e^{t} & e^{t} \cos(2t) & e^{t} \sin(2t) \\ 2e^{t}& e^{t} \sin(2t) & -e^{t} \cos(2t)
\end{pmatrix} \begin{pmatrix}
\dfrac{1}{2} & 0 & 0 \\ \dfrac{3}{2} & 1 & 0 \\ 1 & 0 & -1
\end{pmatrix} \\
&= e^{t} \begin{pmatrix}
1 & 0 & 0 \\ -\dfrac{3}{2} + \dfrac{3}{2} \cos(2t) + \sin(2t) & \cos(2t) & -\sin(2t) \\ 1 + \dfrac{3}{2} \sin(2t) -\cos(2t) & \sin(2t) & \cos(2t)
\end{pmatrix}
\end{align*}

Por lo tanto, la matriz que consideraremos como matriz fundamental de soluciones es

$$e^{\mathbf{A}t} = e^{t} \begin{pmatrix}
1 & 0 & 0 \\ -\dfrac{3}{2} + \dfrac{3}{2} \cos(2t) + \sin(2t) & \cos(2t) & -\sin(2t) \\ 1 + \dfrac{3}{2} \sin(2t) -\cos(2t) & \sin(2t) & \cos(2t)
\end{pmatrix}$$

Como también requerimos de la inversa de esta matriz, verifica que

$$e^{-\mathbf{A}t} = e^{-t} \begin{pmatrix}
1 & 0 & 0 \\ -\dfrac{3}{2} + \dfrac{3}{2} \cos(2t) -\sin(2t) & \cos(2t) & \sin(2t) \\ 1 -\dfrac{3}{2} \sin(2t) -\cos(2t) & -\sin(2t) & \cos(2t)
\end{pmatrix}$$

En este caso la matriz $\mathbf{G}(t)$ es

$$\mathbf{G}(t) = \begin{pmatrix}
0 \\ 0 \\ e^{t} \cos(2t)
\end{pmatrix}$$

Sustituyamos todos estos resultados en la solución (\ref{37}).

$$\mathbf{Y}(t) = e^{\mathbf{A} t} \begin{pmatrix}
0 \\ 1 \\ 1
\end{pmatrix} + e^{\mathbf{A} t} \int_{0}^{t} e^{-s} \begin{pmatrix}
1 & 0 & 0 \\ -\dfrac{3}{2} + \dfrac{3}{2} \cos(2s) -\sin(2s)
& \cos(2s) & \sin(2s) \\ 1 -\dfrac{3}{2} \sin(2s) -\cos(2s) & -\sin(2s) & \cos(2s)
\end{pmatrix} \begin{pmatrix}
0 \\ 0 \\ e^{s} \cos(2s)
\end{pmatrix}ds$$

Por un lado,

$$e^{\mathbf{A} t} \begin{pmatrix}
0 \\ 1 \\ 1
\end{pmatrix} = e^{t} \begin{pmatrix}
1 & 0 & 0 \\ -\dfrac{3}{2} + \dfrac{3}{2} \cos(2t) + \sin(2t) & \cos(2t) & -\sin(2t) \\ 1 + \dfrac{3}{2} \sin(2t) -\cos(2t) & \sin(2t) & \cos(2t)
\end{pmatrix} \begin{pmatrix}
0 \\ 1 \\ 1
\end{pmatrix} = e^{t} \begin{pmatrix}
0 \\ \cos(2t) -\sin(2t) \\ \cos(2t) + \sin(2t)
\end{pmatrix}$$

De tarea moral, determina las constantes $c_{1}$, $c_{2}$ y $c_{3}$ aplicando los valores iniciales sobre la solución complementaria (\ref{36}). ¿Qué relación tiene tu resultado con la operación anterior?.

Por otro lado,

$$e^{-s} \begin{pmatrix}
1 & 0 & 0 \\ -\dfrac{3}{2} + \dfrac{3}{2} \cos(2s) -\sin(2s)
& \cos(2s) & \sin(2s) \\ 1 -\dfrac{3}{2} \sin(2s) -\cos(2s) & -\sin(2s) & \cos(2s)
\end{pmatrix} e^{s} \begin{pmatrix}
0 \\ 0 \\ \cos(2s)
\end{pmatrix} = \begin{pmatrix}
0 \\ \sin(2s) \cos(2s) \\ \cos^{2}(2s)
\end{pmatrix}$$

Sustituimos estas matrices en $\mathbf{Y}(t)$.

$$\mathbf{Y}(t) = e^{t} \begin{pmatrix}
0 \\ \cos(2t) -\sin(2t) \\ \cos(2t) + \sin(2t)
\end{pmatrix} + e^{\mathbf{A} t} \int_{0}^{t} \begin{pmatrix}
0 \\ \sin(2s) \cos(2s) \\ \cos^{2}(2s)
\end{pmatrix} ds$$

Resolvamos la integral.

\begin{align*}
\int_{0}^{t} \begin{pmatrix}
0 \\ \sin(2s) \cos(2s) \\ \cos^{2}(2s)
\end{pmatrix} ds &= \left. \begin{pmatrix}
0 \\ -\dfrac{1}{8} \cos(4s) \\ \dfrac{s}{2} + \dfrac{\sin(4s)}{8}
\end{pmatrix} \right|_{t} – \left. \begin{pmatrix}
0 \\ -\dfrac{1}{8} \cos(4s) \\ \dfrac{s}{2} + \dfrac{\sin(4s)}{8}
\end{pmatrix} \right|_{0} \\
&= \begin{pmatrix}
0 \\ -\dfrac{1}{8} \cos(4t) \\ \dfrac{t}{2} + \dfrac{\sin(4t)}{8}
\end{pmatrix} -\begin{pmatrix}
0 \\ -\dfrac{1}{8} \\ 0
\end{pmatrix} \\
&= \begin{pmatrix}
0 \\ \dfrac{1 -\cos(4t)}{8} \\ \dfrac{4t + \sin(4t)}{8}
\end{pmatrix}
\end{align*}

Entonces,

$$\mathbf{Y}(t) = e^{t} \begin{pmatrix}
0 \\ \cos(2t) -\sin(2t) \\ \cos(2t) + \sin(2t)
\end{pmatrix} + e^{\mathbf{A}t} \begin{pmatrix}
0 \\ \dfrac{1 -\cos(4t)}{8} \\ \dfrac{4t + \sin(4t)}{8}
\end{pmatrix}$$

Ahora realicemos el producto del segundo sumando.

\begin{align*}
e^{\mathbf{A}t} \begin{pmatrix}
0 \\ \dfrac{1 -\cos(4t)}{8} \\ \dfrac{4t + \sin(4t)}{8}
\end{pmatrix} &= e^{t} \begin{pmatrix}
1 & 0 & 0 \\ -\dfrac{3}{2} + \dfrac{3}{2} \cos(2t) + \sin(2t) & \cos(2t) & -\sin(2t) \\ 1 + \dfrac{3}{2} \sin(2t) -\cos(2t) & \sin(2t) & \cos(2t)
\end{pmatrix} \begin{pmatrix}
0 \\ \dfrac{1 -\cos(4t)}{8} \\ \dfrac{4t + \sin(4t)}{8}
\end{pmatrix} \\
&= e^{t} \begin{pmatrix}
0 \\ \cos(2t) \left( \dfrac{1 -\cos(4t)}{8} \right) -\sin(2t) \left( \dfrac{4t + \sin(4t)}{8} \right) \\ \sin(2t) \left( \dfrac{1 -\cos(4t)}{8} \right) + \cos(2t) \left( \dfrac{4t + \sin(4t)}{8} \right)
\end{pmatrix} \\
&= e^{t} \begin{pmatrix}
0 \\ -\dfrac{t \sin(2t)}{2} + \dfrac{\cos(2t) -\cos(4t) \cos(2t) -\sin(4t) \sin(2t)}{8}
\\ \dfrac{t \cos(2t)}{2} + \dfrac{\sin(2t) + \sin(4t) \cos(2t) -\cos(4t) \sin(2t)}{8}
\end{pmatrix}
\end{align*}

Así,

$$\mathbf{Y}(t) = e^{t} \begin{pmatrix}
0 \\ \cos(2t) -\sin(2t) \\ \cos(2t) + \sin(2t)
\end{pmatrix} + e^{t} \begin{pmatrix}
0 \\ -\dfrac{t \sin(2t)}{2} + \dfrac{\cos(2t) -\cos(4t) \cos(2t) -\sin(4t) \sin(2t)}{8}
\\ \dfrac{t \cos(2t)}{2} + \dfrac{\sin(2t) + \sin(4t) \cos(2t) -\cos(4t) \sin(2t)}{8}
\end{pmatrix}$$

Haciendo las operaciones correspondientes se obtiene finalmente que la solución al problema con valores iniciales es

$$\mathbf{Y}(t) = e^{t} \begin{pmatrix}
0 \\ \cos(2t) -\left( 1+ \dfrac{1}{2}t \right) \sin(2t) \\ \left( 1 + \dfrac{1}{2}t \right) \cos(2t) + \dfrac{5}{4} \sin(2t)
\end{pmatrix}$$

$\square$

Vemos que este método puede ser bastante largo y complicado, de hecho se puede volver una tarea imposible de hacer a mano si se tienen sistemas con matriz $\mathbf{A}$ de $3 \times 3$ o más. Se recomienda, en la medida de lo posible, usar algún programa computacional para llevar a cabo algunas de las operaciones, lo importante es entender como llevar a cabo el método.

Con esto concluimos lo que corresponde al estudio de los distintos métodos para resolver sistemas lineales. Prácticamente hemos concluido con la unidad 3 del curso.

En las siguientes dos entradas de esta unidad trataremos de justificar los teoremas de existencia y unicidad en el caso de los sistemas lineales, esto con el propósito de justificar toda la teoría desarrollada a lo largo de la unidad.

Tarea moral

Los siguientes ejercicios no forman parte de la evaluación del curso, pero servirán para entender mucho mejor los conceptos vistos en esta entrada, así como temas posteriores.

  1. Obtener la solución general de los siguientes sistemas lineales no homogéneos.
  • $\mathbf{Y}^{\prime} = \begin{pmatrix}
    1 & -1 \\ 1 & 1
    \end{pmatrix} \mathbf{Y} + \begin{pmatrix}
    \cos(t) \\ \sin(t)
    \end{pmatrix} e^{t}$
  • $\mathbf{Y}^{\prime} = \begin{pmatrix}
    1 & 2 \\ -\dfrac{1}{2} & 1
    \end{pmatrix} \mathbf{Y} + \begin{pmatrix}
    \csc(t) \\ \sec(t)
    \end{pmatrix} e^{t}$
  • $\mathbf{Y}^{\prime} = \begin{pmatrix}
    1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
    \end{pmatrix} \mathbf{Y} + \begin{pmatrix}
    e^{t} \\ e^{2t} \\ te^{3t}
    \end{pmatrix}$
  1. Resolver los siguientes problemas con valores iniciales.
  • $\mathbf{Y}^{\prime} = \begin{pmatrix}
    2 & 0 & 1 \\ 0 & 2 & 0 \\ 0 & 1 & 3
    \end{pmatrix} \mathbf{Y} + \begin{pmatrix}
    1 \\ 0 \\ 1
    \end{pmatrix} e^{2t}, \hspace{1cm} \mathbf{Y}(0) = \begin{pmatrix}
    1 \\ 1 \\ 1
    \end{pmatrix}$
  • $\mathbf{Y}^{\prime} = \begin{pmatrix}
    -1 & -1 & -2 \\ 1 & 1 & 1 \\ 2 & 1 & 3
    \end{pmatrix} \mathbf{Y} + \begin{pmatrix}
    1 \\ 0 \\ 0
    \end{pmatrix} e^{t}, \hspace{1cm} \mathbf{Y}(0) = \begin{pmatrix}
    0 \\ 0 \\ 0
    \end{pmatrix}$

Más adelante…

En la siguiente entrada demostraremos los teoremas de existencia y unicidad para el caso de los sistemas lineales de primer orden con coeficientes constantes homogéneos y no homogéneos y posteriormente, en la última entrada de esta unidad, justificaremos el teorema de existencia y unicidad en el caso general, es decir, para sistemas lineales y no lineales.

Entradas relacionadas

Agradecimientos

Trabajo realizado con el apoyo del Programa UNAM-DGAPA-PAPIME PE104522 «Hacia una modalidad a distancia de la Licenciatura en Matemáticas de la FC-UNAM – Etapa 2»

Ecuaciones Diferenciales I: Exponencial de una matriz y matriz fundamental de soluciones

Por Omar González Franco

Las matemáticas son el arte de dar el mismo nombre a diferentes cosas.
– Henri Poincare

Introducción

Ya conocemos las propiedades de los sistemas de ecuaciones diferenciales lineales de primer orden, así como las de sus soluciones. Mucho de lo que vimos en las dos entradas anteriores es bastante similar a lo que desarrollamos en las dos primeras entradas de la unidad 2, sin embargo en esta entrada aprenderemos algo completamente nuevo, definiremos una herramienta matemática que nos será de bastante utilidad. ¡Se trata de la exponencial de una matriz!.

En esta entrada definiremos lo que significa $e^{\mathbf{A}t}$, donde $\mathbf{A}$ es una matriz de $n \times n$ con componentes reales constantes.

$$\mathbf{A} = \begin{pmatrix}
a_{11} & a_{12} & \cdots & a_{1n} \\
a_{21} & a_{22} & \cdots & a_{2n} \\
\vdots & & & \vdots \\
a_{n1} & a_{n2} & \cdots & a_{nn}
\end{pmatrix} \label{1} \tag{1}$$

Así mismo, estudiaremos algunas de sus propiedades y su utilidad en la resolución de sistemas lineales.

Operaciones sobre componentes de matrices

Muchas de las operaciones que se pueden hacer hacía una matriz son aplicables sobre cada una de las componentes que conforman a dicha matriz. Para comprender este hecho es conveniente definir lo que significa la derivada e integral de una matriz, esto nos permitirá ganar intuición.

Consideremos por un momento una matriz de $n \times n$ compuesta de funciones.

$$\mathbf{A}(t) = \begin{pmatrix}
a_{11}(t) & a_{12}(t) & \cdots & a_{1n}(t) \\
a_{21}(t) & a_{22}(t) & \cdots & a_{2n}(t) \\
\vdots & & & \vdots \\
a_{n1}(t) & a_{n2}(t) & \cdots & a_{nn}(t)
\end{pmatrix} \label{2} \tag{2}$$

donde $a_{i, j}(t)$, $i, j \in \{1, 2, 3, \cdots, n \}$ son funciones sobre algún intervalo común $\delta$. Comencemos por definir la derivada de una matriz.

Algunas reglas de derivación se enuncian a continuación.

De manera equivalente se puede definir la integral de una matriz.

Ejemplo: Calcular la derivada de la matriz

$$\mathbf{A}(t) = \begin{pmatrix}
\cos(t) & \sin(t) \\ \sin^{2}(t) & \cos^{2}(t)
\end{pmatrix}$$

Solución: Aplicamos la derivada sobre cada componente de la matriz.

$$\dfrac{d}{dt} \mathbf{A}(t) = \begin{pmatrix}
\dfrac{d}{dt} \cos(t) & \dfrac{d}{dt} \sin(t) \\ \dfrac{d}{dt} \sin^{2}(t) & \dfrac{d}{dt} \cos^{2}(t)
\end{pmatrix} = \begin{pmatrix}
-\sin(t) & \cos(t) \\ 2 \sin(t) \cos(t) & -2 \cos(t) \sin(t)
\end{pmatrix}$$

$\square$

De manera similar se puede hacer la integral de una matriz.

Definamos lo que es una serie de matrices. En este caso consideremos matrices constantes.

Si tenemos series de matrices es claro que algunas pueden o no converger. A continuación definimos la convergencia en matrices.

Con esto en mente puede resultar más comprensible la definición de la exponencial de una matriz.

Exponencial de una matriz

Recordemos que la función escalar de la exponencial se define como

$$e^{\alpha t} = 1 + \alpha t + \alpha^{2} \dfrac{t^{2}}{2!} + \cdots + \alpha^{k} \dfrac{t^{k}}{k!} = \sum_{k = 0}^{\infty} \alpha^{k} \dfrac{t^{k}}{k!} \label{9} \tag{9}$$

Con las definiciones anteriores podemos extender la serie de la exponencial anterior a una serie de matrices.

Se puede demostrar que la serie (\ref{10}) converge, sin embargo se requiere de un poco más de teoría que queda fuera de nuestro interés.

Veamos un ejemplo en el que determinemos la exponencial de una matriz.

Ejemplo: Determinar la matriz $e^{\mathbf{A}}$, en donde

$$\mathbf{A} = \begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix}$$

Solución: Para determinar la matriz $e^{\mathbf{A}}$ usemos directamente la definición (\ref{10}). Sabemos que

$$\mathbf{A}^{0} = \begin{pmatrix}
1 & 0 \\ 0 & 1
\end{pmatrix} = \mathbf{I} \hspace{1cm} y \hspace{1cm} \mathbf{A}^{1} = \begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix} = \begin{pmatrix}
2^{0} & 2^{0} \\ 2^{0} & 2^{0}
\end{pmatrix}$$

Ahora bien,

$$\mathbf{A}^{2} = \mathbf{AA} = \begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix} \begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix} = \begin{pmatrix}
2 & 2 \\ 2 & 2
\end{pmatrix} = \begin{pmatrix}
2^{1} & 2^{1} \\ 2^{1} & 2^{1}
\end{pmatrix}$$

$$\mathbf{A}^{3} = \mathbf{AA}^{2} = \begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix} \begin{pmatrix}
2 & 2 \\ 2 & 2
\end{pmatrix} = \begin{pmatrix}
4 & 4 \\ 4 & 4
\end{pmatrix} = \begin{pmatrix}
2^{2} & 2^{2} \\ 2^{2} & 2^{2}
\end{pmatrix}$$

$$\mathbf{A}^{4} = \mathbf{AA}^{3} = \begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix} \begin{pmatrix}
4 & 4 \\ 4 & 4
\end{pmatrix} = \begin{pmatrix}
8 & 8 \\ 8 & 8
\end{pmatrix} = \begin{pmatrix}
2^{3} & 2^{3} \\ 2^{3} & 2^{3}
\end{pmatrix}$$

$$\vdots$$

$$\mathbf{A}^{k} = \mathbf{AA}^{k -1} = \begin{pmatrix}
2^{k -1} & 2^{k -1} \\ 2^{k -1} & 2^{k -1}
\end{pmatrix}$$

$$\mathbf{A}^{k + 1} = \mathbf{AA}^{k} = \begin{pmatrix}
2^{k} & 2^{k} \\ 2^{k} & 2^{k}
\end{pmatrix}$$

$$\vdots$$

Sustituimos en (\ref{10}).

\begin{align*} e^{\begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix}} &= \dfrac{1}{0!} \mathbf{A}^{0} + \dfrac{1}{1!} \mathbf{A}^{1} + \dfrac{1}{2!} \mathbf{A}^{2} + \dfrac{1}{3!} \mathbf{A}^{3} + \dfrac{1}{4!} \mathbf{A}^{4} + \cdots + \dfrac{1}{k!} \mathbf{A}^{k} + \cdots \\
&= \dfrac{1}{0!} \begin{pmatrix}
1 & 0 \\ 0 & 1
\end{pmatrix} + \dfrac{1}{1!} \begin{pmatrix}
2^{0} & 2^{0} \\ 2^{0} & 2^{0}
\end{pmatrix} + \dfrac{1}{2!} \begin{pmatrix}
2^{1} & 2^{1} \\ 2^{1} & 2^{1}
\end{pmatrix} + \dfrac{1}{3!} \begin{pmatrix}
2^{2} & 2^{2} \\ 2^{2} & 2^{2}
\end{pmatrix} + \dfrac{1}{4!} \begin{pmatrix}
2^{3} & 2^{3} \\ 2^{3} & 2^{3}
\end{pmatrix} + \cdots + \dfrac{1}{k!} \begin{pmatrix}
2^{k -1} & 2^{k -1} \\ 2^{k -1} & 2^{k -1}
\end{pmatrix} \\
&= \begin{pmatrix}
\dfrac{1}{0!}(1) & \dfrac{1}{0!}(0) \\ \dfrac{1}{0!}(0) & \dfrac{1}{0!}(1)
\end{pmatrix} + \begin{pmatrix}
\dfrac{1}{1!}(2^{0}) & \dfrac{1}{1!}(2^{0}) \\ \dfrac{1}{1!}(2^{0}) & \dfrac{1}{1!}(2^{0})
\end{pmatrix} + \begin{pmatrix}
\dfrac{1}{2!}(2^{1}) & \dfrac{1}{2!}(2^{1}) \\ \dfrac{1}{2!}(2^{1}) & \dfrac{1}{2!}(2^{1})
\end{pmatrix} + \begin{pmatrix}
\dfrac{1}{3!}(2^{2}) & \dfrac{1}{3!}(2^{2}) \\ \dfrac{1}{3!}(2^{2}) & \dfrac{1}{3!}(2^{2})
\end{pmatrix} + \cdots + \begin{pmatrix}
\dfrac{1}{k!}(2^{k -1}) & \dfrac{1}{k!}(2^{k -1}) \\ \dfrac{1}{k!}(2^{k -1}) & \dfrac{1}{k!}(2^{k -1})
\end{pmatrix} \\
&= \begin{pmatrix}
\dfrac{1}{0!}(1) + \dfrac{1}{1!}(2^{0}) + \dfrac{1}{2!}(2^{1}) + \dfrac{1}{3!}(2^{2}) + \cdots + \dfrac{1}{k!}(2^{k -1}) & \dfrac{1}{0!}(0) + \dfrac{1}{1!}(2^{0}) + \dfrac{1}{2!}(2^{1}) + \dfrac{1}{3!}(2^{2}) + \cdots + \dfrac{1}{k!}(2^{k -1}) \\ \dfrac{1}{0!}(0) + \dfrac{1}{1!}(2^{0}) + \dfrac{1}{2!}(2^{1}) + \dfrac{1}{3!}(2^{2}) + \cdots + \dfrac{1}{k!}(2^{k -1}) & \dfrac{1}{0!}(1) + \dfrac{1}{1!}(2^{0}) + \dfrac{1}{2!}(2^{1}) + \dfrac{1}{3!}(2^{2}) + \cdots + \dfrac{1}{k!}(2^{k -1})
\end{pmatrix} \\
&= \dfrac{1}{2}\begin{pmatrix}
\dfrac{1}{0!}(2) + \dfrac{1}{1!}(2^{1}) + \dfrac{1}{2!}(2^{2}) + \dfrac{1}{3!}(2^{3}) + \cdots + \dfrac{1}{k!}(2^{k}) & \dfrac{1}{0!}(0) + \dfrac{1}{1!}(2^{1}) + \dfrac{1}{2!}(2^{2}) + \dfrac{1}{3!}(2^{3}) + \cdots + \dfrac{1}{k!}(2^{k}) \\ \dfrac{1}{0!}(0) + \dfrac{1}{1!}(2^{1}) + \dfrac{1}{2!}(2^{2}) + \dfrac{1}{3!}(2^{3}) + \cdots + \dfrac{1}{k!}(2^{k}) & \dfrac{1}{0!}(2) + \dfrac{1}{1!}(2^{1}) + \dfrac{1}{2!}(2^{2}) + \dfrac{1}{3!}(2^{3}) + \cdots + \dfrac{1}{k!}(2^{k})
\end{pmatrix}
\end{align*}

¡Uff!. En la última igualdad lo que hicimos fue multiplicar por un uno en la forma $1 = \dfrac{2}{2}$, esto nos permitió hacer que la potencia de los $2$ sea la misma que en el factorial.

Escribamos la última matriz como series infinitas.

$$e^{\begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix}} = \dfrac{1}{2} \begin{pmatrix}
\dfrac{1}{0!}(2) + \sum_{k = 1}^{\infty} \dfrac{2^{k}}{k!} & \dfrac{1}{0!}(0) + \sum_{k = 1}^{\infty} \dfrac{2^{k}}{k!} \\ \dfrac{1}{0!}(0) + \sum_{k = 1}^{\infty} \dfrac{2^{k}}{k!} & \dfrac{1}{0!}(2) + \sum_{k = 1}^{\infty} \dfrac{2^{k}}{k!}
\end{pmatrix}$$

Nos gustaría hacer que las series comiencen en $k = 0$. Notemos que, de acuerdo a la forma de la serie, el termino $k = 0$ daría como resultado un $1$, considerando esto la expresión anterior la podemos escribir como

$$e^{\begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix}} = \dfrac{1}{2} \begin{pmatrix}
1 + \sum_{k = 0}^{\infty} \dfrac{2^{k}}{k!} & \sum_{k = 1}^{\infty} \dfrac{2^{k}}{k!} \\ \sum_{k = 1}^{\infty} \dfrac{2^{k}}{k!} & 1 + \sum_{k = 0}^{\infty} \dfrac{2^{k}}{k!}
\end{pmatrix}$$

Para las componentes de la matriz en las que la serie aún no comienza en $k = 0$ sumamos y restamos un $1$, así el $+1$ puede ser incluido en la serie.

$$e^{\begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix}} = \dfrac{1}{2} \begin{pmatrix}
1 + \sum_{k = 0}^{\infty} \dfrac{2^{k}}{k!} & -1 + \sum_{k = 0}^{\infty} \dfrac{2^{k}}{k!} \\ -1 + \sum_{k = 0}^{\infty} \dfrac{2^{k}}{k!} & 1 + \sum_{k = 0}^{\infty} \dfrac{2^{k}}{k!}
\end{pmatrix}$$

Ahora todas las series comienzan en $k = 0$. Sabemos que la serie converge a un número.

$$\sum_{k = 0}^{\infty} \dfrac{2^{k}}{k!} = e^{2}$$

Por definición de convergencia en matrices, se tiene

$$e^{\begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix}} = \dfrac{1}{2} \begin{pmatrix}
1 + e^{2} & -1 + e^{2} \\ -1 + e^{2} & 1 + e^{2}
\end{pmatrix}$$

Por lo tanto, la matriz que buscamos es

$$e^{\begin{pmatrix}
1 & 1 \\ 1 & 1
\end{pmatrix}} = \begin{pmatrix}
\dfrac{e^{2} + 1}{2} & \dfrac{e^{2} -1}{2} \\ \dfrac{e^{2} -1}{2} & \dfrac{e^{2} + 1}{2}
\end{pmatrix}$$

$\square$

Como se puede notar, calcular la exponencial de una matriz usando la definición puede ser una tarea bastante tediosa. Por su puesto existen métodos que nos permiten calcular este tipo de matrices de forma más sencilla, más adelante revisaremos uno de ellos.

Algunas propiedades de la exponencial de una matriz se enuncian a continuación.

No demostraremos este teorema ya que nuestro principal interés está en conocer como estos conceptos y propiedades se pueden aplicar en nuestro estudio sobre sistemas lineales.

A continuación mostraremos un resultado importante e interesante y es que la función (\ref{11}) ¡es solución del sistema lineal homogéneo $\mathbf{Y}^{\prime} = \mathbf{AY}$!.

La exponencial de una matriz y los sistemas lineales

Demostración: Consideremos la función $\mathbf{Y}(t) = e^{\mathbf{A} t}$. Apliquemos la derivada término a término de la definición (\ref{11}).

\begin{align*}
\dfrac{d}{dt} e^{\mathbf{A}t} &= \dfrac{d}{dt} \left[ \mathbf{I} + \mathbf{A}t + \mathbf{A}^{2} \dfrac{t^{2}}{2!} + \cdots + \mathbf{A}^{k}\dfrac{t^{k}}{k!} + \cdots \right] \\
&= \dfrac{d}{dt} \mathbf{I} + \dfrac{d}{dt} \left( \mathbf{A}t \right) + \dfrac{d}{dt} \left( \mathbf{A}^{2} \dfrac{t^{2}}{2!} \right) + \cdots + \dfrac{d}{dt} \left( \mathbf{A}^{k}\dfrac{t^{k}}{k!} \right) + \cdots
\end{align*}

Como las matrices $\mathbf{I}$ y $\mathbf{A}$ son constantes, entonces se tiene lo siguiente.

\begin{align*}
\dfrac{d}{dt} e^{\mathbf{A}t} &= \mathbf{0} + \mathbf{A} + \mathbf{A}^{2}t + \dfrac{1}{2!}\mathbf{A}^{3}t^{2} + \cdots + \mathbf{A}^{k} \dfrac{t^{k -1}}{(k -1)!} + \mathbf{A}^{k + 1}\dfrac{t^{k}}{k!} + \cdots \\
&= \mathbf{A} \left[ \mathbf{I} + \mathbf{A}t + \mathbf{A}^{2} \dfrac{t^{2}}{2!} + \cdots + \mathbf{A}^{k}\dfrac{t^{k}}{k!} + \cdots \right] \\
&= \mathbf{A} e^{\mathbf{A}t}
\end{align*}

Con esto hemos mostrado que

$$\dfrac{d}{dt} e^{\mathbf{A}t} = \mathbf{A} e^{\mathbf{A}t} \label{12} \tag{12}$$

Es decir,

$$\mathbf{Y}^{\prime} = \mathbf{AY}$$

$\square$

La ecuación (\ref{12}) no sólo prueba que es solución del sistema lineal, sino que además muestra cuál es la derivada de la matriz $e^{\mathbf{A} t}$.

Veamos un problema de valores iniciales.

Demostración: Consideremos la función

$$\mathbf{Y}(t) = e^{\mathbf{A} t} \mathbf{Y}_{0}$$

con $\mathbf{Y}_{0}$ un vector constante, si la derivamos obtenemos lo siguiente.

$$\mathbf{Y}^{\prime} = \dfrac{d}{dt}(e^{\mathbf{A}t} \mathbf{Y}_{0}) = \left( \mathbf{A} e^{\mathbf{A}t} \right) \mathbf{Y}_{0} = \mathbf{A} \left( e^{\mathbf{A}t} \mathbf{Y}_{0} \right) = \mathbf{AY}$$

En donde se ha hecho uso del resultado (\ref{12}). Esto muestra que la función $\mathbf{Y}(t) = e^{\mathbf{A} t} \mathbf{Y}_{0}$ es solución del sistema $\mathbf{Y}^{\prime} = \mathbf{AY}$.

Si tomamos $t = 0$ y considerando que $e^{\mathbf{A}0} = e^{\mathbf{0}} = \mathbf{I}$, se tiene

$$\mathbf{Y}(0) = e^{\mathbf{A}0} \mathbf{Y}_{0} = \mathbf{IY}_{0} = \mathbf{Y}_{0}$$

Es decir, se satisface la condición inicial $\mathbf{Y}(0) = \mathbf{Y}_{0}$.

$\square$

Nota: Es claro que la matriz $e^{\mathbf{A} t}$ es una matriz de $n \times n$, mientras que el vector constante $\mathbf{Y}_{0}$ es un vector de $n \times 1$, así que es importante el orden de las matrices, la función del teorema anterior es $\mathbf{Y}(t) = e^{\mathbf{A} t} \mathbf{Y}_{0}$ la cual esta bien definida y el resultado es una matriz de $n \times n$, mientras que la operación $\mathbf{Y}_{0} e^{\mathbf{A} t}$ no está definida de acuerdo al algoritmo de multiplicación de matrices. Cuidado con ello.

Para concluir esta entrada retomaremos el concepto de matriz fundamental de soluciones visto en la entrada anterior.

Matriz fundamental de soluciones

Retomemos la definición de matriz fundamental de soluciones.

En esta sección denotaremos por $\hat{\mathbf{Y}}(t)$ a una matriz fundamental de soluciones.

El objetivo de esta sección es mostrar que se puede determinar directamente la matriz $e^{\mathbf{A}t}$ a partir de cualquier matriz fundamental de soluciones. Antes de llegar a este resultado veamos unos resultados previos.

Demostración:

$\Rightarrow$) Por demostrar: $\hat{\mathbf{Y}}(t)$ satisface al sistema y se cumple que $W(0) \neq 0$.

Supongamos que $\hat{\mathbf{Y}}(t)$ es una matriz fundamental de soluciones del sistema $\mathbf{Y}^{\prime} = \mathbf{AY}$ conformada por los vectores solución $\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}$, es decir,

$$\hat{\mathbf{Y}}(t) = \begin{pmatrix} \mathbf{Y}_{1} & \mathbf{Y}_{2} & \cdots & \mathbf{Y}_{n} \end{pmatrix} = \begin{pmatrix}
y_{11}(t) & y_{12}(t) & \cdots & y_{1n}(t) \\
y_{21}(t) & y_{22}(t) & \cdots & y_{2n}(t) \\
\vdots & & & \vdots \\
y_{n1}(t) & y_{n2}(t) & \cdots & y_{nn}(t)
\end{pmatrix}$$

Si $\mathbf{A}$ es la matriz de coeficientes (\ref{1}), entonces

$$\mathbf{A} \hat{\mathbf{Y}} = \begin{pmatrix}
a_{11} & a_{12} & \cdots & a_{1n} \\
a_{21} & a_{22} & \cdots & a_{2n} \\
\vdots & & & \vdots \\
a_{n1} & a_{n2} & \cdots & a_{nn}
\end{pmatrix} \begin{pmatrix}
y_{11}(t) & y_{12}(t) & \cdots & y_{1n}(t) \\
y_{21}(t) & y_{22}(t) & \cdots & y_{2n}(t) \\
\vdots & & & \vdots \\
y_{n1}(t) & y_{n2}(t) & \cdots & y_{nn}(t)
\end{pmatrix} \label{15} \tag{15}$$

La $i$-ésima columna del resultado de multiplicar estas matrices es

$$\mathbf{A} \mathbf{Y}_{i} = \begin{pmatrix}
a_{11}y_{1i}(t) + a_{12}y_{2i}(t) + \cdots + a_{1n}y_{ni}(t) \\
a_{21}y_{1i}(t) + a_{22}y_{2i}(t) + \cdots + a_{2n}y_{ni}(t) \\
\vdots \\
a_{n1}y_{1i}(t) + a_{n2}y_{2i}(t) + \cdots + a_{nn}y_{ni}(t)
\end{pmatrix} \label{16} \tag{16}$$

Identificamos que esta matriz corresponde a la derivada $\mathbf{Y}^{\prime}_{i}$ ya que cada $\mathbf{Y}_{i}$ es solución del sistema, es decir,

$$\mathbf{Y}^{\prime}_{i} =\mathbf{AY}_{i}, \hspace{1cm} i = 1, 2, \cdots, n \label{17} \tag{17}$$

Entonces $\mathbf{A\hat{Y}}(t)$ tiene por columnas a los vectores $\mathbf{Y}^{\prime}_{1}, \mathbf{Y}^{\prime}_{2}, \cdots, \mathbf{Y}^{\prime}_{n}$. Por lo tanto

$$\hat{\mathbf{Y}}^{\prime} = \mathbf{A\hat{Y}}$$

es decir, la matriz fundamental $\hat{\mathbf{Y}}(t)$ satisface al sistema.

Como $\hat{\mathbf{Y}}(t)$ es una matriz fundamental de soluciones, entonces los vectores que la componen $\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}$ son linealmente independientes, de manera que $\forall$ $t \in \mathbb{R}$ el Wronskiano es distinto de cero.

$$W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n})(t) \neq 0 \label{18} \tag{18}$$

en particular se cumple para $t = 0$, es decir,

$$W( \mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n})(0) \neq 0$$

Demostremos el regreso.

$\Leftarrow$) Por demostrar: $\hat{\mathbf{Y}}(t)$ es una matriz fundamental de soluciones.

Sea $\hat{\mathbf{Y}}(t)$ una matriz compuesta por los vectores $\mathbf{Y}_{1}(t), \mathbf{Y}_{2}(t), \cdots, \mathbf{Y}_{n}(t)$ y cuya derivada es

$$\hat{\mathbf{Y}}^{\prime} = \mathbf{A \hat{Y}}$$

entonces las columnas satisfacen

$$\mathbf{AY}_{i} = \mathbf{Y}^{\prime}_{i}$$

así $\mathbf{Y}_{i}(t)$ es solución del sistema

$$\mathbf{Y}^{\prime} = \mathbf{AY}$$

para $i = 1, 2, \cdots, n$.

Por otro lado, por hipótesis

$$W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n})(0) \neq 0$$

Como es distinto de cero en un punto, entonces lo debe ser en todo el dominio, es decir, $\forall$ $t \in \mathbb{R}$ se cumple que

$$W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n})(t) \neq 0$$

lo que significa que $\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}$ son linealmente independientes.

De ambos resultados concluimos que la matriz $\hat{\mathbf{Y}}(t)$ es una matriz fundamental de soluciones.

$\square$

Veamos un resultado interesante. Ya vimos que la matriz $e^{\mathbf{A} t}$ es solución del sistema $\mathbf{Y}^{\prime} = \mathbf{AY}$, pero no solo resulta ser solución, sino que además ¡es una matriz fundamental de soluciones!.

Demostración: Anteriormente mostramos que

$$\dfrac{d}{dt} e^{\mathbf{A}t} = \mathbf{A} e^{\mathbf{A}t}$$

lo que prueba que $\mathbf{Y}(t) = e^{\mathbf{A} t}$ es solución del sistema $\mathbf{Y}^{\prime} = \mathbf{AY}$.

Supongamos que $e^{\mathbf{A} t}$ está compuesta por la matriz de vectores $\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}$. Si $t = 0$, se tiene que

$$e^{\mathbf{A} 0} = e^{\mathbf{0}} = \mathbf{I}$$

y además el determinante es distinto de cero.

$$|e^{\mathbf{A} 0}| = |\mathbf{I}| = 1 \neq 0 \label{19} \tag{19}$$

o bien,

$$|e^{\mathbf{A} 0}| = W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n})(0) \neq 0 \label{20} \tag{20}$$

Por el teorema anterior concluimos que $\mathbf{Y}(t) = e^{\mathbf{A} t}$ es una matriz fundamental de soluciones del sistema lineal.

$\square$

Veamos un resultado más antes de llegar a nuestro objetivo.

Demostración: Sean $\hat{\mathbf{Y}}(t)$ y $\hat{\mathbf{Z}}(t)$ matrices fundamentales del sistema $\mathbf{Y}^{\prime} = \mathbf{AY}$. Supongamos que

$$\hat{\mathbf{Y}}(t) = \begin{pmatrix} \mathbf{Y}_{1} & \mathbf{Y}_{2} & \cdots & \mathbf{Y}_{n} \end{pmatrix} \label{22} \tag{22}$$

y

$$\hat{\mathbf{Z}}(t) = \begin{pmatrix} \mathbf{Z}_{1} & \mathbf{Z}_{2} & \cdots & \mathbf{Z}_{n} \end{pmatrix} \label{23} \tag{23}$$

Donde la $i$-ésima columna de las matrices anteriores son

$$\mathbf{Y}_{i} = \begin{pmatrix}
y_{1i} \\ y_{2i} \\ \vdots \\ y_{ni}
\end{pmatrix} \hspace{1cm} y \hspace{1cm} \mathbf{Z}_{i} = \begin{pmatrix}
z_{1i} \\ z_{2i} \\ \vdots \\ z_{ni}
\end{pmatrix} \label{24} \tag{24}$$

Como ambas matrices son matrices fundamentales de soluciones, entonces cada $\mathbf{Y}_{i}$ y cada $\mathbf{Z}_{i}$ $i = 1, 2, 3, \cdots, n$ son linealmente independientes. Esto nos permite escribir cualquier columna de $\hat{\mathbf{Y}}(t)$ como combinación lineal de las columnas de $\hat{\mathbf{Z}}(t)$, es decir, existen constantes $c_{1i}, c_{2i}, \cdots, c_{ni}$, tales que

$$\mathbf{Y}_{i}(t) = c_{1i} \mathbf{Z}_{1}(t) + c_{2i} \mathbf{Z}_{2}(t) + \cdots + c_{ni} \mathbf{Z}_{n}(t) \label{25} \tag{25}$$

Donde el índice $i$ de las constantes indica el número de columna de la matriz $\hat{\mathbf{Y}}(t)$, es decir, si definimos el vector

$$\mathbf{c}_{i} = \begin{pmatrix}
c_{1i} \\ c_{2i} \\ \vdots \\ c_{ni}
\end{pmatrix} \label{26} \tag{26}$$

podemos escribir la combinación lineal (\ref{25}) como

$$\mathbf{Y}_{i}(t) = \begin{pmatrix} \mathbf{Z}_{1} & \mathbf{Z}_{2} & \cdots & \mathbf{Z}_{n} \end{pmatrix} \begin{pmatrix}
c_{1i} \\ c_{2i} \\ \vdots \\ c_{ni}
\end{pmatrix} = \begin{pmatrix}
c_{1i}z_{11} + c_{2i}z_{12} + \cdots + c_{ni}z_{1n} \\
c_{1i}z_{21} + c_{2i}z_{22} + \cdots + c_{ni}z_{2n} \\
\vdots \\
c_{1i}z_{n1} + c_{2i}z_{n2} + \cdots + c_{ni}z_{nn}
\end{pmatrix} \label{27} \tag{27}$$

Definamos la matriz constante $\mathbf{C}$ como

$$\mathbf{C} = \begin{pmatrix} \mathbf{c_{1}} & \mathbf{c_{2}} & \cdots & \mathbf{c_{n}} \end{pmatrix} = \begin{pmatrix}
c_{11} & c_{12} & \cdots & c_{1n} \\
c_{21} & c_{22} & \cdots & c_{2n} \\
\vdots & & & \vdots \\
c_{n1} & c_{n2} & \cdots & c_{nn}
\end{pmatrix} \label{28} \tag{28}$$

En forma matricial la operación $\hat{\mathbf{Z}}(t) \mathbf{C}$ corresponde al producto de las siguientes matrices de $n \times n$.

$$\hat{\mathbf{Z}}(t) \mathbf{C} = \begin{pmatrix}
z_{11} & z_{12} & \cdots & z_{1n} \\
z_{21} & z_{22} & \cdots & z_{2n} \\
\vdots & & & \vdots\\
z_{n1} & z_{n2} & \cdots & z_{nn}
\end{pmatrix} \begin{pmatrix}
c_{11} & c_{12} & \cdots & c_{1n} \\
c_{21} & c_{22} & \cdots & c_{2n} \\
\vdots & & & \vdots \\
c_{n1} & c_{n2} & \cdots & c_{nn}
\end{pmatrix} \label{29} \tag{29}$$

Observemos con cuidado que el resultado (\ref{27}) corresponde a la $i$-ésima columna de hacer el producto $\hat{\mathbf{Z}}(t) \mathbf{C}$. Por lo tanto, las $n$ ecuaciones (\ref{25}) son equivalentes al sistema

$$\hat{\mathbf{Y}}(t) = \hat{\mathbf{Z}}(t) \mathbf{C}$$

$\square$

Hemos llegado al resultado final. Dicho resultado involucra el concepto de matriz inversa, recordemos este concepto de álgebra lineal.

La ecuación anterior queda como

$$\mathbf{AA}^{-1} = \mathbf{A}^{-1} \mathbf{A} = \mathbf{I} \label{31} \tag{31}$$

Demostremos el resultado que nos muestra cómo obtener la matriz $e^{\mathbf{A}t}$ a partir de una matriz fundamental de soluciones.

Demostración: Sabemos que $e^{\mathbf{A} t}$ y $\hat{\mathbf{Y}}(t)$ son matrices fundamentales de soluciones de $\mathbf{Y}^{\prime} = \mathbf{AY}$, de acuerdo al teorema anterior ambas funciones se relacionan de la siguiente forma.

$$e^{\mathbf{A} t} = \hat{\mathbf{Y}}(t) \mathbf{C} \label{33} \tag{33}$$

para alguna matriz constante $\mathbf{C}$.

Tomemos $t = 0$, por un lado

$$e^{\mathbf{A} 0} = e^{\mathbf{0}} = \mathbf{I}$$

Por otro lado, de (\ref{33})

$$e^{\mathbf{A} 0} = \hat{\mathbf{Y}}(0) \mathbf{C}$$

De ambas ecuaciones tenemos que

$$\hat{\mathbf{Y}}(0) \mathbf{C} = \mathbf{I} \label{34} \tag{34}$$

Esta ecuación obliga que

$$\mathbf{C} = \hat{\mathbf{Y}}^{-1}(0) \label{35} \tag{35}$$

Sustituyendo en (\ref{33}) concluimos que

$$e^{\mathbf{A} t} = \hat{\mathbf{Y}}(t) \hat{\mathbf{Y}}^{-1}(0)$$

$\square$

Finalicemos con un ejemplo.

Ejemplo: Determinar la matriz $e^{\mathbf{A} t}$, donde

$$\mathbf{A} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix}$$

Solución: Imagina lo complicado que sería este problema si lo intentáramos resolver usando la definición (\ref{11}).

En la entrada anterior vimos que una matriz fundamental de soluciones del sistema lineal

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \mathbf{Y}$$

es

$$\hat{\mathbf{Y}}(t) = \begin{pmatrix}
1 & e^{2t} & 0 \\ -1 & e^{2t} & 0 \\ 0 & 0 & e^{3t}
\end{pmatrix}$$

Determinemos la matriz $e^{\mathbf{A} t}$ usando la expresión (\ref{32}).

Calcular la matriz inversa puede ser una tarea larga y debido a que no es el objetivo llevar a cabo estas operaciones se puede comprobar que la matriz inversa de $\hat{\mathbf{Y}}(t)$ es

$$\hat{\mathbf{Y}}^{-1}(t) = \begin{pmatrix}
\dfrac{1}{2} & -\dfrac{1}{2} & 0 \\ \dfrac{1}{2 e^{2t}} & \dfrac{1}{2 e^{2t}} & 0 \\ 0 & 0 & \dfrac{1}{e^{3t}}
\end{pmatrix}$$

Basta probar que $\hat{\mathbf{Y}}(t) \hat{\mathbf{Y}}^{-1}(t) = \mathbf{I}$. Para calcular la inversa se puede hacer uso de algún programa computacional. Si en la matriz inversa evaluamos en $t = 0$, se tiene

$$\hat{\mathbf{Y}}^{-1}(0) = \begin{pmatrix}
\dfrac{1}{2} & -\dfrac{1}{2} & 0 \\ \dfrac{1}{2} & \dfrac{1}{2} & 0 \\ 0 & 0 & 1
\end{pmatrix}$$

Por lo tanto,

$$e^{\mathbf{A} t} = \hat{\mathbf{Y}}(t) \hat{\mathbf{Y}}^{-1}(0) = \begin{pmatrix}
1 & e^{2t} & 0 \\ -1 & e^{2t} & 0 \\ 0 & 0 & e^{3t}
\end{pmatrix} \begin{pmatrix}
\dfrac{1}{2} & -\dfrac{1}{2} & 0 \\ \dfrac{1}{2} & \dfrac{1}{2} & 0 \\ 0 & 0 & 1
\end{pmatrix}$$

Haciendo la multiplicación de matrices correspondiente obtenemos finalmente que

$$e^{\mathbf{A} t} = \begin{pmatrix}
\dfrac{e^{2t} + 1}{2} & \dfrac{e^{2t} -1}{2} & 0 \\ \dfrac{e^{2t} -1}{2} & \dfrac{e^{2t} + 1}{2} & 0 \\ 0 & 0 & e^{3t}
\end{pmatrix}$$

$\square$

Existen otras formas de calcular la exponencial de una matriz, una de ellas es usando la transformada de Laplace y otra puede ser diagonalizando matrices. Si lo deseas puedes investigar sobre estos métodos en la literatura, por nuestra parte serán temas que no revisaremos ya que, más que obtener exponenciales de matrices, estamos interesados en obtener soluciones de sistemas de ecuaciones diferenciales, aunque como vimos, están estrechamente relacionados.

Tarea moral

Los siguientes ejercicios no forman parte de la evaluación del curso, pero servirán para entender mucho mejor los conceptos vistos en esta entrada, así como temas posteriores.

  1. Sea

    $\mathbf{A}(t) = \begin{pmatrix} \sin(2t) \\ e^{3t} \\ 8t -1 \end{pmatrix}$
  • Demostrar que la derivada de la matriz $\mathbf{A}$ es

    $\dfrac{d}{dt} \mathbf{A}(t) = \begin{pmatrix} 2 \cos(2t) \\ 3e^{3t} \\ 8 \end{pmatrix}$
  • Demostrar que la integral de $0$ a $t$ de la matriz $\mathbf{A}$ es

    $\int_{0}^{t} \mathbf{A}(s) ds = \begin{pmatrix} -\dfrac{1}{2} \cos(2t) + \dfrac{1}{2} \\ \dfrac{1}{3} e^{3t} -\dfrac{1}{3} \\ 4t^{2} -t \end{pmatrix}$
  1. Demostrar que

    $\int{\begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end{pmatrix} ds} = t \mathbf{I} + \mathbf{C}$

    Donde,

    $\mathbf{I} = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end{pmatrix} \hspace{1cm} y \hspace{1cm} \mathbf{C} = \begin{pmatrix} 0 & c & c \\ c & 0 & c \\ c & c & 0 \end{pmatrix}$

    con $c$ una constante.
  1. Obtener la matriz $e^{\mathbf{A}t}$ para los siguientes casos:
  • $\mathbf{A} = \begin{pmatrix} 1 & 2 \\ 2 & 1 \end{pmatrix}$
  • $\mathbf{A} = \begin{pmatrix} 2 & 1 \\ 1 & 2 \end{pmatrix}$
  • $\mathbf{A} = \begin{pmatrix} 1 & 2 \\ 2 & 4 \end{pmatrix}$
  1. Sea $\hat{\mathbf{Y}}(t)$ una matriz fundamental de soluciones del sistema $\mathbf{Y}^{\prime} = \mathbf{AY}$. Demostrar que $$e^{\mathbf{A} (t -t_{0})} = \hat{\mathbf{Y}}(t) \hat{\mathbf{Y}}^{-1}(t_{0})$$
  1. Una matriz fundamental del sistema

    $\mathbf{Y}^{\prime} = \begin{pmatrix} 1 & -1 \\ 1 & 3 \end{pmatrix} \mathbf{Y} = \mathbf{AY}$

    es

    $\hat{\mathbf{Y}}(t) = \begin{pmatrix} e^{2t} & te^{2t} \\ -e^{2t} & -(1 + t) e^{2t} \end{pmatrix}$
  • Demostrar que la matriz anterior en efecto es una matriz fundamental de soluciones del sistema.
  • Demostrar que la matriz $e^{\mathbf{A} t}$ está dada por

    $e^{\mathbf{A} t} = \begin{pmatrix} (1 -t) e^{2t} & -te^{2t} \\ te^{2t} & (1 + t) e^{2t} \end{pmatrix}$

Más adelante…

En estas tres primeras entradas de la unidad 3 establecimos la teoría básica que debemos conocer sobre los sistemas lineales de primer orden compuestos por $n$ ecuaciones diferenciales lineales de primer orden. En particular, esta entrada es de interés, pues más adelante la exponencial de una matriz volverá a aparecer cuando estudiemos métodos de resolución y cuando justifiquemos los teoremas de existencia y unicidad.

En las siguientes entradas comenzaremos a desarrollar los distintos métodos de resolución de estos sistemas lineales. En particular, en la siguiente entrada desarrollaremos el método de eliminación de variables, éste método en realidad es muy sencillo, útil y práctico en muchas ocasiones, aunque también es un método muy limitado.

Entradas relacionadas

Agradecimientos

Trabajo realizado con el apoyo del Programa UNAM-DGAPA-PAPIME PE104522 «Hacia una modalidad a distancia de la Licenciatura en Matemáticas de la FC-UNAM – Etapa 2»

Ecuaciones Diferenciales I: Soluciones a sistemas de ecuaciones diferenciales

Por Omar González Franco

Los errores y dificultades no resueltos en el pasado de las matemáticas
siempre han sido las oportunidades de su futuro.
– E. T. Bell

Introducción

En la entrada anterior vimos lo que es un sistema de ecuaciones diferenciales, en particular un sistema lineal de primer orden. Vimos también lo que es un problema de valores iniciales y establecimos la notación matricial.

Así mismo, vimos cómo es que una ecuación diferencial lineal de orden $n$ se puede transformar en un sistema lineal de primer orden, esto tiene bastante ventaja ya que, una vez que veamos cómo resolver sistemas de ecuaciones diferenciales, muchas veces será más sencillo resolver el sistema que resolver la ecuación de orden $n$ aplicando los métodos que ya conocemos.

En esta entrada estudiaremos las propiedades de las soluciones de los sistemas lineales de primer orden.

Cabe mencionar que mucho de lo que desarrollaremos en esta entrada es bastante similar a la teoría vista con las ecuaciones diferenciales de orden $n$, comenzando por la validez del principio de superposición.

A partir de ahora sólo usaremos la notación matricial y toda la teoría básica del álgebra lineal que éstas conllevan.

Soluciones de sistemas lineales de primer orden

Comencemos por estudiar el caso homogéneo. El sistema lineal de primer orden homogéneo es

$$\begin{pmatrix}
y_{1}^{\prime} \\ y_{2}^{\prime} \\ \vdots \\ y_{n}^{\prime}
\end{pmatrix} = \begin{pmatrix}
a_{11}(t) & a_{12}(t) & \cdots & a_{1n}(t) \\
a_{21}(t) & a_{22}(t) & \cdots & a_{2n}(t) \\
\vdots & & & \vdots \\
a_{n1}(t) & a_{n2}(t) & \cdots & a_{nn}(t)
\end{pmatrix} \begin{pmatrix}
y_{1} \\ y_{2} \\ \vdots \\ y_{n}
\end{pmatrix} \label{1} \tag{1}$$

O bien,

$$\mathbf{Y^{\prime}} = \mathbf{AY} \label{2} \tag{2}$$

En la entrada anterior definimos la solución de un sistema de ecuaciones diferenciales en el intervalo $\delta$ como el conjunto de $n$ funciones

$$S_{0} = \{y_{1}(t), y_{2}(t), \cdots, y_{n}(t)\} \label{3} \tag{3}$$

definidas en $\delta$ y diferenciables en el mismo intervalo, tales que satisfacen simultáneamente las $n$ ecuaciones diferenciables de un sistema lineal.

Las soluciones pueden ser escritas como el vector

$$\mathbf{Y} = \begin{pmatrix}
y_{1}(t) \\ y_{2}(t) \\ \vdots \\ y_{n}(t)
\end{pmatrix} \label{4} \tag{4}$$

cuyos elementos son funciones derivables que satisfacen un sistema lineal en el intervalo $\delta$.

En las siguientes definiciones y teoremas se supondrá que los coeficientes $a_{ij}(t)$, $i, j \in \{1, 2, 3, \cdots, n\}$ y ,para el caso no homogéneo, las funciones $g_{i}(t)$, son continuas en algún intervalo común $\delta$.

Comencemos por mostrar que el principio de superposición también es valido para sistemas lineales.

Demostración: Consideremos la combinación lineal

$$\mathbf{Y} = c_{1} \mathbf{Y}_{1} + c_{2} \mathbf{Y}_{2} + \cdots + c_{m} \mathbf{Y}_{m}$$

con

$$\mathbf{Y}_{i} = \begin{pmatrix}
y_{1i} \\ y_{2i} \\ \vdots \\ y_{ni}
\end{pmatrix}$$

para $i = 1, 2, \cdots, m$. La derivada de $\mathbf{Y}_{i}$ esta dada por

$$\mathbf{Y}_{i}^{\prime} = \begin{pmatrix}
y_{1i}^{\prime} \\ y_{2i}^{\prime} \\ \vdots \\ y_{ni}^{\prime}
\end{pmatrix}$$

Entonces la derivada de la combinación lineal es

\begin{align*}
\mathbf{Y}^{\prime} &= \begin{pmatrix}
c_{1}y_{11}^{\prime} + c_{2}y_{12}^{\prime} + \cdots + c_{m}y_{1m}^{\prime} \\
c_{1}y_{21}^{\prime} + c_{2}y_{22}^{\prime} + \cdots + c_{m}y_{2m}^{\prime} \\
\vdots \\
c_{1}y_{n1}^{\prime} + c_{2}y_{n2}^{\prime} + \cdots + c_{m}y_{nm}^{\prime}
\end{pmatrix} \\
&= c_{1} \begin{pmatrix}
y_{11}^{\prime} \\ y_{21}^{\prime} \\ \vdots \\ y_{n1}^{\prime}
\end{pmatrix} + c_{2} \begin{pmatrix}
y_{12}^{\prime} \\ y_{22}^{\prime} \\ \vdots \\ y_{n2}^{\prime}
\end{pmatrix} + \cdots + c_{m} \begin{pmatrix}
y_{1m}^{\prime} \\ y_{2m}^{\prime} \\ \vdots \\ y_{nm}^{\prime}
\end{pmatrix} \\
&= c_{1} \mathbf{Y}_{1}^{\prime} + c_{2} \mathbf{Y}_{2}^{\prime} + \cdots + c_{m} \mathbf{Y}_{m}^{\prime}
\end{align*}

Como cada $\mathbf{Y}_{i}$, $i = 1, 2, \cdots, m$, es solución del sistema homogéneo (\ref{2}) en $\delta$, entonces

$$\mathbf{Y}_{i}^{\prime} = \mathbf{A} \mathbf{Y}_{i}$$

así

\begin{align*}
\mathbf{Y}^{\prime} &= c_{1} (\mathbf{AY}_{1}) + c_{2} (\mathbf{AY}_{2}) + \cdots + c_{m} (\mathbf{AY}_{m}) \\
&= \mathbf{A}(c_{1} \mathbf{Y}_{1} + c_{2} \mathbf{Y}_{2} + \cdots + c_{m} \mathbf{Y}_{m}) \\
&= \mathbf{AY}
\end{align*}

En donde se ha hecho uso de la propiedad distributiva de la matriz $\mathbf{A}$ y de la hipótesis (\ref{5}). Por lo tanto, la combinación lineal

$$\mathbf{Y} = c_{1} \mathbf{Y}_{1} + c_{2} \mathbf{Y}_{2} + \cdots + c_{m} \mathbf{Y}_{m}$$

también es solución y los es en el mismo intervalo común $\delta$ ya que esta compuesta de soluciones definidas en dicho intervalo.

$\square$

Intenta hacer la demostración.

Realicemos un ejemplo.

Ejemplo: Probar que la combinación lineal

$$\mathbf{Y} = c_{1} \mathbf{Y}_{1} + c_{2} \mathbf{Y}_{2} + c_{3} \mathbf{Y}_{3} = c_{1}
\begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix} + c_{2} \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix} + c_{3} \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix}$$

es solución del sistema lineal

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \mathbf{Y}$$

Solución: Probemos que cada uno de los vectores de la combinación lineal es solución y usemos el principio de superposición.

Los vectores son

$$\mathbf{Y}_{1} = \begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix}, \hspace{1cm} \mathbf{Y}_{2} = \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix}, \hspace{1cm} \mathbf{Y}_{3} = \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix}$$

Por un lado, derivemos estos vectores.

$$\mathbf{Y}^{\prime}_{1} = \begin{pmatrix}
0 \\ 0 \\ 0
\end{pmatrix}, \hspace{1cm} \mathbf{Y}^{\prime}_{2} = \begin{pmatrix}
2e^{2t} \\ 2e^{2t} \\ 0
\end{pmatrix}, \hspace{1cm} \mathbf{Y}^{\prime}_{3} = \begin{pmatrix}
0 \\ 0 \\ 3e^{3t}
\end{pmatrix}$$

Por otro lado, sustituyamos cada uno de los vectores en el sistema lineal y usemos los resultados anteriores.

$$\mathbf{AY}_{1} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix} = \begin{pmatrix}
1 -1 \\ 1 -1 \\ 0
\end{pmatrix} = \begin{pmatrix}
0 \\ 0 \\ 0
\end{pmatrix} = \mathbf{Y}^{\prime}_{1}$$

$$\mathbf{AY}_{2} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix} = \begin{pmatrix}
e^{2t} + e^{2t} \\ e^{2t} + e^{2t} \\ 0
\end{pmatrix} = \begin{pmatrix}
2e^{2t} \\ 2e^{2t} \\ 0
\end{pmatrix} = \mathbf{Y}^{\prime}_{2}$$

y

$$\mathbf{AY}_{3} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix} = \begin{pmatrix}
0 \\ 0 \\ 3e^{3t}
\end{pmatrix} = \mathbf{Y}^{\prime}_{3}$$

De esta manera queda mostrado que los tres vectores son solución, ya que satisfacen el sistema. Por el principio de superposición concluimos que la combinación lineal

$$\mathbf{Y} = c_{1} \mathbf{Y}_{1} + c_{2} \mathbf{Y}_{2} + c_{3} \mathbf{Y}_{3} = c_{1}
\begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix} + c_{2} \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix} + c_{3} \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix}$$

también es solución del sistema lineal.

$\square$

El principio de superposición nos indica que un sistema lineal puede tener más de una solución, sin embargo, similar al caso de ecuaciones diferenciales de orden $n$, buscamos soluciones que sean linealmente independientes entre sí. A continuación definimos la dependencia e independencia lineal de las soluciones en este contexto.

En la unidad anterior definimos una herramienta muy útil que, además de ayudarnos a resolver ecuaciones diferenciales de orden superior en algunos métodos, nos ayuda a determinar si un conjunto de soluciones es linealmente independiente, dicha herramienta es el Wronskiano, la definición en el caso de los sistemas lineales de primer orden, es la siguiente.

Se puede demostrar que si el Wronskiano es distinto de cero, entonces las soluciones son linealmente independientes, igual que antes, esto es conocido como el criterio para soluciones linealmente independientes. Para demostrar este hecho es conveniente recordar algunos resultados de álgebra que podremos usar en la demostración.

Recordemos que un sistema lineal de $n$ ecuaciones con $n$ incógnitas es un conjunto de ecuaciones

$$\begin{matrix}
b_{11}u_{1} + b_{12}u_{2} + \cdots + b_{1n}u_{n} = d_{1} \\
b_{21}u_{1} + b_{22}u_{2} + \cdots + b_{2n}u_{n} = d_{2}\\
\vdots\\
b_{n1}u_{1} + b_{n2}u_{2} + \cdots + b_{nn}u_{n} = d_{n}
\end{matrix} \label{9} \tag{9}$$

Con $b_{i, j}$ y $d_{i}$, $i, j \in \{1,2, 3, \cdots, n\}$ números reales dados y $u_{i}$, $i = 1, 2, \cdots, n$ las incógnitas. Usando la notación matricial podemos escribir el sistema (\ref{9}) como

$$\mathbf{BU} = \mathbf{D} \label{10} \tag{10}$$

con

$$\mathbf{B} = \begin{pmatrix}
b_{11} & b_{12} & \cdots & b_{1n} \\
b_{21} & b_{22} & \cdots & b_{2n} \\
\vdots & & & \vdots \\
b_{n1} & b_{n2} & \cdots & b_{nn}
\end{pmatrix}, \hspace{1cm} \mathbf{U} = \begin{pmatrix}
u_{1} \\ u_{2} \\ \vdots \\ u_{n}
\end{pmatrix}, \hspace{1cm} \mathbf{D} = \begin{pmatrix}
d_{1} \\ d_{2} \\ \vdots \\ d_{n}
\end{pmatrix}$$

Los resultados que nos interesan son los siguientes.

Si $\mathbf{D} = \mathbf{0}$, el sistema (\ref{10}) también recibe el nombre de sistema homogéneo.

Con estos resultados podemos demostrar el criterio para soluciones linealmente independientes que se enuncia a continuación.

Demostración:

$\Rightarrow$) Por demostrar: $W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}) \neq 0$.

Sea $t_{0} \in \delta$ en el que $W(t_{0}) = 0$, en donde $W(t_{0})$ denota al Wronskiano con cada vector solución evaluado en el punto $t_{0}$.

$$W(t_{0}) = W(\mathbf{Y}_{1}(t_{0}), \mathbf{Y}_{2}(t_{0}), \cdots, \mathbf{Y}_{n}(t_{0})) $$

En una combinación de ambos teoremas de los resultados de álgebra podemos deducir que existen constantes $c_{1}, c_{2}, \cdots, c_{n}$, no todos cero, tal que

$$\mathbf{Y}(t_{0}) = c_{1} \mathbf{Y}_{1}(t_{0}) + c_{2} \mathbf{Y}_{2}(t_{0}) + \cdots + c_{n} \mathbf{Y}_{n}(t_{0}) = 0 \label{11} \tag{11}$$

Lo que tenemos es un sistema lineal de $n$ ecuaciones homogéneo con $n$ incógnitas (sistema lineal en el contexto algebraico (\ref{10}) con $\mathbf{D} = \mathbf{0}$, no sistema lineal de ecuaciones diferenciales), dichas incógnitas son las constantes $c_{i}$, $i = 1, 2, \cdots, n$. La relación (\ref{11}) se cumple debido a que si el Wronskiano es igual a cero, entonces es posible que el sistema no tenga solución trivial y mucho menos una solución única, esto lo deducimos de los teoremas de álgebra que establecimos.

Por otro lado, sabemos por hipótesis que los vectores $\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}$ son solución del sistema homogéneo (\ref{2}) en el intervalo $\delta$, por el principio de superposición sabemos también que la combinación lineal

$$\mathbf{Y}(t) = c_{1} \mathbf{Y}_{1} + c_{2} \mathbf{Y}_{2} + \cdots + c_{n} \mathbf{Y}_{n}$$

es solución de (\ref{2}) en $\delta$. Del resultado (\ref{11}) y de la unicidad de la solución se deduce que $\mathbf{Y}(t) = 0$ para algún punto $t = t_{0} \in \delta$, es decir,

$$c_{1} \mathbf{Y}_{1} + c_{2} \mathbf{Y}_{2} + \cdots + c_{n} \mathbf{Y}_{n} = 0$$

Pero por hipótesis los vectores $\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}$ son linealmente independientes en $\delta$, lo que implica que

$$c_{1} = c_{2} = \cdots = c_{n} = 0$$

lo cual es una contradicción con lo que establecimos en (\ref{11}). Por lo tanto, el Wronskiano tiene que ser distinto de cero, es decir

$$W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}) \neq 0$$

$\Leftarrow$) Por demostrar: $S$ es linealmente independiente.

Este caso también lo demostraremos por contradicción. Supongamos que los vectores solución $\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}$ son linealmente dependientes en $\delta$, esto implica que existen constantes $c_{1}, c_{2}, \cdots, c_{n}$ no todos cero, tal que

$$c_{1} \mathbf{Y}_{1} + c_{2} \mathbf{Y}_{2} + \cdots + c_{n} \mathbf{Y}_{n} = 0$$

Este sistema lo podemos escribir en la forma (\ref{9}) como

$$\begin{matrix}
c_{1}y_{11} + c_{2}y_{12} + \cdots + c_{n}y_{1n} = 0 \\
c_{1}y_{21} + c_{2}y_{22} + \cdots + c_{n}y_{2n} = 0 \\
\vdots\\
c_{1}y_{n1} + c_{2}y_{n2} + \cdots + c_{n}y_{nn} = 0
\end{matrix}$$

En donde las funciones $y_{ij}$, $i, j \in \{1, 2, 3, \cdots, n\}$ son los coeficientes y las constantes $c_{i}$, $i = 1, 2, \cdots, n$ son las incógnitas. Debido a que las $c_{i}$ no son todas cero implica que el sistema no tiene solución trivial y por el segundo teorema de los resultados de álgebra concluimos que

$$W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}) = 0$$

Pero, por hipótesis

$$W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}) \neq 0$$

lo cual es una contradicción y todo nace de considerar a $S$ como un conjunto linealmente dependiente. Por lo tanto, el conjunto de soluciones

$$S = \{\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}\}$$

es linealmente independiente en $\delta$.

$\square$

Un resultado interesante se enuncia a continuación.

Este resultado nos garantiza que si $W \neq 0$ para algún punto $t_{0} \in \delta$, entonces $W \neq 0$ para toda $t \in \delta$ y por el criterio anterior las soluciones serán linealmente independientes en ese intervalo.

El conjunto de soluciones linealmente independientes del sistema lineal (\ref{2}) recibe un nombre especial.

El siguiente teorema nos garantiza la existencia de este conjunto.

El conjunto fundamental de soluciones está constituido por vectores que son linealmente independientes entre sí, con estos vectores es posible formar una matriz cuyas columnas están formadas con las entradas de dichos vectores, esta matriz tiene un nombre especial.

Un hecho interesante es que el determinante de la matriz fundamental de soluciones corresponde al Wronskiano.

$$W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}) = |\mathbf{M}(t)| \label{13} \tag{13}$$

Realicemos un ejemplo, para ello consideremos el sistema lineal del ejemplo anterior.

Ejemplo: Mostrar que las soluciones

$$\mathbf{Y}_{1} = \begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix}, \hspace{1cm} \mathbf{Y}_{2} = \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix}, \hspace{1cm} \mathbf{Y}_{3} = \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix}$$

del sistema lineal

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \mathbf{Y}$$

son linealmente independientes.

Solución: En el ejemplo anterior ya comprobamos que efectivamente son solución del sistema lineal dado. Para determinar si son linealmente independientes veamos si el Wronskiano es distinto de cero.

$$W(\mathbf{Y}_{1}, \mathbf{Y}_{2}, \mathbf{Y}_{3}) = \begin{vmatrix}
1 & e^{2t} & 0 \\ -1 & e^{2t} & 0 \\ 0 & 0 & e^{3t}
\end{vmatrix} = e^{5t} + 0 + 0 -0 -0 -(-e^{5t}) = 2e^{5t} \neq 0$$

Como $W \neq 0$, $\forall$ $t \in \mathbb{R}$, entonces los vectores dados son linealmente independientes y por lo tanto forman un conjunto fundamental de soluciones en $\mathbb{R}$.

$$S = \left\{ \begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix}, \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix}, \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix} \right\}$$

La matriz fundamental de soluciones es

$$\mathbf{M}(t) = \begin{pmatrix}
1 & e^{2t} & 0 \\ -1 & e^{2t} & 0 \\ 0 & 0 & e^{3t}
\end{pmatrix}$$

$\square$

Un buen ejercicio sería mostrar que un conjunto de soluciones del sistema lineal homogéneo (\ref{2}) forma un espacio vectorial, es relativamente sencillo probar cada una de las propiedades o axiomas que definen a un espacio vectorial. El resultado a demostrar de tarea moral es el siguiente.

Soluciones generales a sistemas lineales

Ahora que conocemos algunas propiedades de las soluciones de sistemas lineales, es momento de conocer la forma general de las soluciones de los sistemas lineales tanto homogéneos como no homogéneos.

Comencemos por enunciar el teorema que establece la forma de la solución general de un sistema lineal homogéneo (\ref{2}).

Demostración: Sea $\mathbf{Y}(t)$ una solución arbitraria del sistema lineal homogéneo en el intervalo $\delta$, sea $t_{0} \in \delta$ y supongamos que

$$\mathbf{Y}(t_{0}) = \begin{pmatrix}
b_{1} \\ b_{2} \\ \vdots \\ b_{n}
\end{pmatrix} = \mathbf{Y}_{0}$$

Es decir, la función $\mathbf{Y}(t)$ satisface el problema de valores iniciales $\mathbf{Y}^{\prime} = \mathbf{AY}; \mathbf{Y}(t_{0}) = \mathbf{Y}_{0}$.

Por otro lado, por el principio de superposición sabemos que la combinación lineal

$$\hat{\mathbf{Y}}(t) = c_{1} \mathbf{Y}_{1}(t) + c_{2} \mathbf{Y}_{2}(t) + \cdots + c_{n} \mathbf{Y}_{n}(t)$$

también es solución del sistema lineal $\mathbf{Y}^{\prime} = \mathbf{AY}$. Donde $c_{i}$, $i = 1, 2, \cdots, n$ son constantes arbitrarias y las $\mathbf{Y}_{i}$, $i = 1, 2, \cdots, n$ son las soluciones del conjunto fundamental de soluciones del sistema lineal. Supongamos que

$$\hat{\mathbf{Y}}(t_{0}) = c_{1} \mathbf{Y}_{1}(t_{0}) + c_{2} \mathbf{Y}_{2}(t_{0}) + \cdots + c_{n} \mathbf{Y}_{n}(t_{0}) = \mathbf{Y}_{0}$$

Lo que tenemos es el siguiente sistema de $n$ ecuaciones.

$$\begin{matrix}
c_{1}y_{11}(t_{0}) + c_{2}y_{12}(t_{0}) + \cdots + c_{n}y_{1n}(t_{0}) = b_{1} \\
c_{1}y_{21}(t_{0}) + c_{2}y_{22}(t_{0}) + \cdots + c_{n}y_{2n}(t_{0}) = b_{2} \\
\vdots \\
c_{1}y_{n1}(t_{0}) + c_{2}y_{n2}(t_{0}) + \cdots + c_{n}y_{nn}(t_{0}) = b_{n}
\end{matrix}$$

En donde las incógnitas son las contantes $c_{i}$, $i = 1, 2, \cdots, n$. Como las funciones $y_{ij}$, $i,j \in \{1, 2, 3, \cdots, n \}$ pertenecen a vectores del conjunto de soluciones, entonces sabemos que $\mathbf{Y}_{1}, \mathbf{Y}_{2}, \cdots, \mathbf{Y}_{n}$ son linealmente independientes y por el criterio para soluciones linealmente independientes inferimos que $W(t_{0}) \neq 0$, donde

$$W(t_{0}) = W(\mathbf{Y}_{1}(t_{0}), \mathbf{Y}_{2}(t_{0}), \cdots, \mathbf{Y}_{n}(t_{0}))$$

De los resultados de álgebra deducimos que el sistema de $n$ ecuaciones tiene solución única, esto significa que existen constantes únicas $c_{1}, c_{2}, \cdots, c_{n}$, tal que

$$c_{1} \mathbf{Y}_{1}(t_{0}) + c_{2} \mathbf{Y}_{2}(t_{0}) + \cdots + c_{n} \mathbf{Y}_{n}(t_{0}) = \mathbf{Y}_{0}$$

Esto nos indica que

$$\hat{\mathbf{Y}}(t) = c_{1} \mathbf{Y}_{1}(t) + c_{2} \mathbf{Y}_{2}(t) + \cdots + c_{n} \mathbf{Y}_{n}(t)$$

es solución del problema de valores iniciales. Por el teorema de existencia y unicidad para sistemas lineales homogéneas concluimos que $\mathbf{Y}(t) = \hat{\mathbf{Y}}(t)$, es decir,

$$\mathbf{Y}(t) = c_{1} \mathbf{Y}_{1}(t) + c_{2} \mathbf{Y}_{2}(t) + \cdots + c_{n} \mathbf{Y}_{n}(t)$$

Como $\mathbf{Y}(t)$ es una solución arbitraria, entonces debe ser la solución general del sistema lineal homogéneo en $\delta$.

$\square$

Para concluir la entrada estudiemos el caso no homogéneo.

Sistemas no homogéneos

El sistema lineal de primer orden no homogéneo es

$$\begin{pmatrix}
y_{1}^{\prime}(t) \\ y_{2}^{\prime}(t) \\ \vdots \\ y_{n}^{\prime}(t)
\end{pmatrix} = \begin{pmatrix}
a_{11}(t) & a_{12}(t) & \cdots & a_{1n}(t) \\
a_{21}(t) & a_{22}(t) & \cdots & a_{2n}(t) \\
\vdots & & & \vdots \\
a_{n1}(t) & a_{n2}(t) & \cdots & a_{nn}(t)
\end{pmatrix} \begin{pmatrix}
y_{1}(t) \\ y_{2}(t) \\ \vdots \\ y_{n}(t)
\end{pmatrix} + \begin{pmatrix}
g_{1}(t) \\ g_{2}(t) \\ \vdots \\ g_{n}(t)
\end{pmatrix} \label{15} \tag{15}$$

O bien,

$$\mathbf{Y^{\prime}} = \mathbf{AY} + \mathbf{G} \label{16} \tag{16}$$

El vector de funciones que satisface el sistema (\ref{16}) es una solución y recibe un nombre.

A continuación se enuncia el teorema que nos muestra la forma general de la solución de un sistema lineal no homogéneo.

Demostración: Sea

$$\mathbf{Y}_{p}(t) = \begin{pmatrix}
y_{1p} \\ y_{2p} \\ \vdots \\ y_{np}
\end{pmatrix}$$

una solución particular de (\ref{16}) y sean $\mathbf{Y}_{1}(t), \mathbf{Y}_{2}(t), \cdots, \mathbf{Y}_{n}(t)$, $n$ soluciones linealmente independientes del sistema homogéneo asociado $\mathbf{Y^{\prime}} = \mathbf{AY}$.

Sea $\mathbf{Y}(t)$ una solución arbitraria del sistema no homogéneo, notemos lo siguiente.

\begin{align*}
(\mathbf{Y}(t) -\mathbf{Y}_{p}(t))^{\prime} &= \mathbf{Y}^{\prime}(t) -\mathbf{Y}_{p}^{\prime}(t) \\
&= (\mathbf{AY}(t) + \mathbf{G}) -(\mathbf{AY}_{p}(t) + \mathbf{G}) \\
&= \mathbf{A} (\mathbf{Y}(t) -\mathbf{Y}_{p}(t))
\end{align*}

Este resultado nos indica que $\mathbf{Y}(t) -\mathbf{Y}_{p}(t)$ es solución del sistema homogéneo, eso significa que se puede escribir como

$$\mathbf{Y}(t) -\mathbf{Y}_{p}(t) = c_{1}\mathbf{Y}_{1}(t) + c_{2}\mathbf{Y}_{2}(t) + \cdots + c_{n}\mathbf{Y}_{n}(t)$$

entonces, la solución $\mathbf{Y}$ tiene la forma

$$\mathbf{Y}(t) = c_{1}\mathbf{Y}_{1}(t) + c_{2}\mathbf{Y}_{2}(t) + \cdots + c_{n}\mathbf{Y}_{n}(t) + \mathbf{Y}_{p}(t) \label{19} \tag{19}$$

La solución $\mathbf{Y}(t)$, al ser cualquier solución del sistema lineal no homogéneo, podemos deducir que la solución general debe tener la forma (\ref{19}), por lo que concluimos que $\mathbf{Y}(t)$ se trata de la solución general de (\ref{16}).

Considerando la hipótesis (\ref{17}) concluimos que la solución general del sistema lineal no homogéneo es

$$\mathbf{Y}(t) = \mathbf{Y}_{c}(t) + \mathbf{Y}_{p}(t)$$

$\square$

Cuando estamos trabajando con un sistema lineal no homogéneo, la solución general del sistema lineal homogéneo asociado (\ref{17}) recibe un nombre particular.

Concluyamos con un ejemplo.

Ejemplo: Probar que el vector

$$\mathbf{Y}_{p} = \begin{pmatrix}
-\dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} \\ -e^{t} + \dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} \\ \dfrac{1}{2}t^{2}e^{3t}
\end{pmatrix}$$

es una solución particular del siguiente sistema lineal no homogéneo.

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \mathbf{Y} + \begin{pmatrix}
e^{t} \\ e^{2t} \\ te^{3t}
\end{pmatrix}$$

Solución: Por un lado, derivemos el vector dado.

$$\mathbf{Y}^{\prime}_{p} = \begin{pmatrix}
-\dfrac{1}{2}e^{2t} + \dfrac{1}{2}e^{2t} + te^{2t} \\ -e^{t} + \dfrac{1}{2}e^{2t} + \dfrac{1}{2}e^{2t} + te^{2t} \\ te^{3t} + \dfrac{3}{2}t^{2}e^{3t}
\end{pmatrix} = \begin{pmatrix}
te^{2t} \\ -e^{t} + e^{2t} + te^{2t} \\ te^{3t} + \dfrac{3}{2}t^{2}e^{3t}
\end{pmatrix}$$

Por otro lado, sustituyamos directamente en el sistema al vector dado.

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \begin{pmatrix}
-\dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} \\ -e^{t} + \dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} \\ \dfrac{1}{2}t^{2}e^{3t}
\end{pmatrix} + \begin{pmatrix}
e^{t} \\ e^{2t} \\ te^{3t}
\end{pmatrix}$$

Operando obtenemos lo siguiente.

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
-\dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} -e^{t} + \dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} + e^{t} \\ -\dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} -e^{t} + \dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t}+e^{2t} \\ \dfrac{3}{2}t^{2}e^{3t} + te^{3t}
\end{pmatrix} = \begin{pmatrix}
te^{2t} \\ -e^{t} + e^{2t} + te^{2t} \\ te^{3t} + \dfrac{3}{2}t^{2}e^{3t}
\end{pmatrix}$$

Los resultados obtenidos son los mismos, por lo tanto el vector $\mathbf{Y}_{p}$ es solución del sistema.

En los ejemplos anteriores de esta entrada probamos que el conjunto fundamental de soluciones del sistema lineal homogéneo asociado

$$\mathbf{Y}^{\prime} = \begin{pmatrix}
1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 3
\end{pmatrix} \mathbf{Y}$$

esta constituido por los vectores linealmente independientes

$$\mathbf{Y}_{1} = \begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix}, \hspace{1cm} \mathbf{Y}_{2} = \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix}, \hspace{1cm} \mathbf{Y}_{3} = \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix}$$

de manera que la función complementaria es

$$\mathbf{Y}_{c} = c_{1}
\begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix} + c_{2} \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix} + c_{3} \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix}$$

Como la solución general es

$$\mathbf{Y} = \mathbf{Y}_{c} + \mathbf{Y}_{p}$$

Entonces la solución general del sistema lineal no homogéneo es

$$\mathbf{Y}(t) = c_{1} \begin{pmatrix}
1 \\ -1 \\ 0
\end{pmatrix} + c_{2} \begin{pmatrix}
e^{2t} \\ e^{2t} \\ 0
\end{pmatrix} + c_{3} \begin{pmatrix}
0 \\ 0 \\ e^{3t}
\end{pmatrix} + \begin{pmatrix}
-\dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} \\ -e^{t} + \dfrac{1}{4}e^{2t} + \dfrac{1}{2}te^{2t} \\ \dfrac{1}{2}t^{2}e^{3t}
\end{pmatrix}$$

$\square$

Tarea moral

Los siguientes ejercicios no forman parte de la evaluación del curso, pero servirán para entender mucho mejor los conceptos vistos en esta entrada, así como temas posteriores.

  1. Los siguientes vectores son soluciones de un sistema lineal homogéneo $\mathbf{Y}^{\prime} = \mathbf{AY}$. Determinar si forman un conjunto fundamental de soluciones en $\mathbb{R}$.
  • $\mathbf{Y}_{1} = \begin{pmatrix}
    1 \\ -1
    \end{pmatrix} e^{t}, \hspace{1cm} \mathbf{Y}_{2} = \begin{pmatrix}
    2 \\ 6
    \end{pmatrix}e^{t} + \begin{pmatrix}
    8 \\ -8
    \end{pmatrix}te^{t}$
  • $\mathbf{Y}_{1} = \begin{pmatrix}
    1 \\ 6 \\ -13
    \end{pmatrix},\hspace{1cm} \mathbf{Y}_{2} = \begin{pmatrix}
    1 \\ -2 \\ -1
    \end{pmatrix}e^{-4t}, \hspace{1cm} \mathbf{Y}_{3}= \begin{pmatrix}
    2 \\ 3 \\ -2
    \end{pmatrix}e^{3t}$
  1. Probar que el vector $\mathbf{Y}_{p}$ es una solución particular del sistema lineal dado.
  • $\mathbf{Y}^{\prime} = \begin{pmatrix}
    2 & 1 \\ 3 & 4
    \end{pmatrix} \mathbf{Y} -\begin{pmatrix}
    1 \\ 7
    \end{pmatrix}e^{t}, \hspace{1cm} \mathbf{Y}_{p} = \begin{pmatrix}
    1 \\ 1
    \end{pmatrix}e^{t} + \begin{pmatrix}
    1 \\ -1
    \end{pmatrix}te^{t}$
  • $\mathbf{Y}^{\prime} = \begin{pmatrix}
    1 & 2 & 3 \\
    -4 & 2 & 0 \\
    -6 & 1 & 0
    \end{pmatrix} \mathbf{Y} + \begin{pmatrix}
    -1 \\ 4 \\ 3
    \end{pmatrix} \sin(3t), \hspace{1cm} \mathbf{Y}_{p} = \begin{pmatrix}
    \sin(3t) \\ 0 \\ \cos (3t)
    \end{pmatrix}$
  1. Mostrar que la solución general de

    $\mathbf{Y}^{\prime} = \begin{pmatrix}
    0 & 6 & 0 \\ 1 & 0 & 1 \\ 1 & 1 & 0
    \end{pmatrix} \mathbf{Y}$

    en el intervalo $(-\infty, \infty)$ es

    $\mathbf{Y} = c_{1} \begin{pmatrix}
    6 \\ -1 \\ -5
    \end{pmatrix}e^{-t} + c_{2} \begin{pmatrix}
    -3 \\ 1 \\ 1
    \end{pmatrix}e^{-2t} + c_{3} \begin{pmatrix}
    2 \\ 1 \\ 1
    \end{pmatrix}e^{3t}$
  1. Mostrar que la solución general de

    $\mathbf{Y}^{\prime} = \begin{pmatrix}
    -1 & -1 \\ -1 & 1
    \end{pmatrix} \mathbf{Y} + \begin{pmatrix}
    1 \\ 1
    \end{pmatrix}t^{2} + \begin{pmatrix}
    4 \\ -6
    \end{pmatrix}t + \begin{pmatrix}
    -1 \\ 5
    \end{pmatrix}$

    en el intervalo $(-\infty, \infty)$ es

    $\mathbf{Y} = c_{1} \begin{pmatrix}
    1 \\ -1 -\sqrt{2}
    \end{pmatrix}e^{\sqrt{2t}} + c_{2} \begin{pmatrix}
    1 \\ -1 + \sqrt{2}
    \end{pmatrix}e^{-\sqrt{2t}} + \begin{pmatrix}
    1 \\ 0 \end{pmatrix}t^{2} + \begin{pmatrix}
    -2 \\ 4
    \end{pmatrix}t + \begin{pmatrix}
    1 \\ 0
    \end{pmatrix}$
  1. Demostrar que el conjunto de soluciones del sistema lineal homogéneo $\mathbf{Y}^{\prime} = \mathbf{AY}$ forma un espacio vectorial con la suma y el producto por escalares usuales de matrices.

Más adelante…

Ahora que conocemos lo que son los sistemas lineales de ecuaciones diferenciales y las propiedades de sus soluciones estamos casi listos para comenzar a desarrollar los distintos métodos de resolución, sin embargo, antes de ello es necesario definir una herramienta matemática que será de suma utilidad en el desarrollo posterior de esta unidad. Dicha herramienta es la exponencial de una matriz.

En la siguiente entrada definiremos lo que significa $e^{\mathbf{A} t}$, donde $\mathbf{A}$ es una matriz de $n \times n$ con componentes constantes y veremos como se relaciona con un sistema lineal $\mathbf{Y}^{\prime} = \mathbf{AY}$. Así mismo, profundizaremos en el concepto de matriz fundamental de soluciones.

Entradas relacionadas

Agradecimientos

Trabajo realizado con el apoyo del Programa UNAM-DGAPA-PAPIME PE104522 «Hacia una modalidad a distancia de la Licenciatura en Matemáticas de la FC-UNAM – Etapa 2»