Math 519 review—first order differential equations

In calculus (math 222 in Madison) we learn how to “solve” two kinds of differential equations of the form \begin{equation} \frac{dy}{dx} = f(x, y) \label{eq:mother-of-all-ode} \end{equation}

First Order Separable Equations

By definition a separable differential equation is a diffeq of the form \begin{equation} \label{eq:separable-2} y'(x) = F(x) G(y(x)),\quad\text{or}\quad \frac{d y}{d x} = F(x)G(y). \end{equation} Thus the function $f(x,y)$ on the right hand side in \eqref{eq:mother-of-all-ode} has the special form \[ f(x,y) = F(x)\, G(y). \] For example, the differential equation \[ \frac{d y}{d x} = \sin(x) \bigl(1+y^2\bigr) \] is separable, and one has $F(x) = \sin x$ and $G(y) = 1+y^2$. On the other hand, the differential equation \[ \frac{d y}{d x} = x+y \] is not separable.

Solution method for separable equations

To solve this equation divide by $G(y(x))$ to get \begin{equation} \label{eq:separated} \frac{1}{ G(y(x))} \frac{d y}{d x} = F(x). \end{equation} Next find a function $H(y)$ whose derivative with respect to $y$ is \begin{equation}\label{eq:separable-3} H'(y) = \frac{1}{G(y)} \quad\left(\text{solution: } H(y) = \int {\frac{dy}{G(y)}}.\right) \end{equation} Then the chain rule implies that the left hand side in (\ref{eq:separated}) can be written as \[ \frac{1}{ G(y(x))} \frac{d y}{d x} = H'(y(x)) \frac{d y}{d x} = \frac{d H(y(x))}{d x}. \] Thus \eqref{eq:separated} is equivalent with \[ \frac{d H(y(x))}{d x} = F(x). \] In words: $H(y(x))$ is an antiderivative of $F(x)$, which means we can find $H(y(x))$ by integrating $F(x)$: \begin{equation} \label{eq:separable-solution} H(y(x)) = \int F(x) dx +C. \end{equation} Once we have found the integral of $F(x)$ this gives us $y(x)$ in implicit form: the equation (\ref{eq:separable-solution}) gives us $y(x)$ as an implicit function of $x$. To get $y(x)$ itself we must solve the equation (\ref{eq:separable-solution}) for $y(x)$.

Determining the constant. The solution we get from the above procedure contains an arbitrary constant $C$. If the value of the solution is specified at some given $x_0$, i.e. if $y(x_0)$ is known then we can express $C$ in terms of $y(x_0)$ by using \eqref{eq:separable-solution}.

A snag

We have to divide by $G(y)$ which is problematic when $G(y)=0$. This has as consequence that in addition to the solutions we found with the above procedure, there are at least a few more solutions: the zeroes of $G(y)$ (see the example below). In addition to the zeroes of $G(y)$ there sometimes can be more solutions

Example

We solve \[ \frac{d z}{d t} = (1+z^2)\cos t. \] Separate variables and integrate \[ \int\frac{d z}{1+z^2} = \int\cos t\,d t, \] to get \[ \arctan z = \sin t +C. \] Finally solve for $z$ and we find the general solution \[ z(t) = \tan \bigl (\sin (t) +C\bigr). \]

Example: the snag in action

If we apply the method to $y'(x)=y$, we get \[ y(x) = e^{x+C}. \] No matter how we choose $C$ we never get the function $y(x)=0$, even though $y(x)=0$ satisfies the equation. This is because here $G(y)= y$, and $G(y)$ vanishes for $y=0$.

First Order Linear Equations

Differential equations of the form equation \begin{equation} \frac{d y}{d x}+a(x)y = k(x) \label{eq:inhomogeneous-linear} \end{equation} are called first order linear.

The Integrating Factor

Linear equations can always be solved by multiplying both sides of the equation with a specially chosen function $m(x)$ called the integrating factor. It is defined by \begin{equation} m (x) = e^{A (x)},\text{ where } A (x) = \int a(x)\,d x. \label{eq:integrating-factor-defined} \end{equation} If we multiply the equation \eqref{eq:inhomogeneous-linear} with the integrating factor $m(x)$ we get \[ m(x)\frac{d y}{d x}+a(x)m(x)y = m(x)k(x). \] By the chain rule the integrating factor satisfies \[ \frac{d m(x)}{d x} = \frac{d\, e^{A(x)}} {d x} = \underbrace{A'(x)}_{=a(x)} ~\underbrace{e^{A(x)}_{}}_{=m(x)} = a(x)m(x). \] Therefore one has \begin{align*} \frac{d m(x)y}{d x} &= m(x)\frac{d y}{d x}+a(x)m(x)y \\ &= m(x)\Bigl\{\frac{d y}{d x}+a(x)y \Bigr\}\\ &= m(x)k(x). \end{align*} Integrating and then dividing by the integrating factor gives the solution \[ y=\frac1{m(x)}\left(\int m(x)k(x)\,d x+C\right). \] In this derivation we have to divide by $m(x)$, but since $m(x)=e^{A (x)}$ and since exponentials never vanish we know that $m (x)\neq0$, so we can always divide by $m (x)$.

An example

Find the general solution to the differential equation \[ \frac{d y} {d x} = y + x. \] Then find the solution that satisfies \begin{equation} y(2)=0. \label{eq:example-linear-initial-condition} \end{equation} Solution. We first write the equation in the standard linear form \begin{equation} \label{eq:diffeq-linear-example} \frac{d y} {d x} - y = x, \end{equation} and then multiply with the integrating factor $m(x)$. We could of course memorize the formulas \eqref{eq:integrating-factor-defined} that lead to the integrating factor, but a safer approach is to remember the following procedure, which will always give us the integrating factor. Assuming that $m(x)$ is as yet unknown we multiply the differential equation \eqref{eq:diffeq-linear-example} with $m$, \begin{equation} m(x)\frac{d y} {d x} - m(x) y = m(x)x. \label{eq:example-multiplied-with-m} \end{equation} If $m(x)$ is such that \begin{equation} -m(x) = \frac{d m(x)} {d x}, \label{eq:integrating-factor-condition} \end{equation} then equation \eqref{eq:example-multiplied-with-m} implies \[ m(x)\frac{d y} {d x} + \frac{d m(x)} {d x} y = m(x)x. \] The expression on the left is exactly what comes out of the product rule -- this is the point of multiplying with $m(x)$ and then insisting on \eqref{eq:integrating-factor-condition}. So, if $m(x)$ satisfies \eqref{eq:integrating-factor-condition}, then the differential equation for $y$ is equivalent with \[ \frac{d m(x) y} {d x} = m(x) x. \] We can integrate this equation, \[ m(x) y = \int m(x) x \;d x, \] and thus find the solution \begin{equation} y(x) = \frac{1} {m(x)} \int m(x) x \;d x. \label{eq:example-linear-almost-solved} \end{equation} All we have to do is find the integrating factor $m$. This factor can be any function that satisfies \eqref{eq:integrating-factor-condition}. Equation \eqref{eq:integrating-factor-condition} is a differential equation for $m$, but it is separable, and we can easily solve it: \[ \frac{d m} {d x} = -m \iff \frac{d m} {m} = -d x \iff \ln|m| = -x +C. \] Since \textit{we only need one integrating factor} $m$ we are not interested in finding all solutions of \eqref{eq:integrating-factor-condition}, and therefore we can choose the constant $C$. The simplest choice is $C=0$, which leads to \[ \ln |m| = -x \iff |m| = e^{-x} \iff m = \pm e^{-x}. \] Again, we only need one integrating factor, so we may choose the $\pm$ sign: the simplest choice for $m$ here is \[ m(x) = e^{-x}. \] With this choice of integrating factor we can now complete the calculation that led to \eqref{eq:example-linear-almost-solved}. The solution to the differential equation is \begin{align*} y(x) &= \frac{1} {m(x)} \int m(x) x \;d x \\ &= \frac{1} {e^{-x}} \int e^{-x}x \;d x \\ &= e^{x} \Bigl\{-e^{-x}x - e^{-x} + C\Bigr\} \\ &= -x-1+Ce^x. \end{align*} This is the general solution. To find the solution that satisfies not just the differential equation, but also the “initial condition” \eqref{eq:example-linear-initial-condition}, i.e. $y(2)=0$, we compute $y(2)$ for the general solution, \[ y(2) = -2-1+Ce^2 = -3 + Ce^2. \] The requirement $y(2) = 0$ then tells us that $C=3e^{-2}$. The solution of the differential equation that satisfies the prescribed initial condition is therefore \[ y(x) = -x-1+3e^{x-2}. \]