Linear first order differential equation

Let

$$y'(x)+a(x)y(x)=f(x)$$

where a and f are for the moment continuous function on an real closed interval.
We use a well known property of the definite Riemann integral of a continuous function, namely
let

$$F(x)=\int_{x_0}^xa(t)dt$$

then

$$F'(x)=a(x)$$

So, for

$$G(x)=e^{\int_{x_0}^xa(t)dt}$$

we have

$$G'(x)=a(x)e^{\int_{x_0}^xa(t)dt}$$

We may multiply our equation by G(x) and write

$$y'(x)e^{\int_{x_0}^xa(t)dt}+a(x)e^{\int_{x_0}^xa(t)dt}y(x)=f(x)e^{\int_{x_0}^xa(t)dt}$$

or

$$y'(x)G(x)+G'(x)y(x)=f(x)e^{\int_{x_0}^xa(t)dt}$$

which can be written

$$(y(x)G(x))'=f(x)e^{\int_{x_0}^xa(t)dt}$$

Now we want to take a definite integral in both sides , so we change the name of variables.

$$(y(t)G(t))'=f(t)e^{\int_{x_0}^ta(\tau)d\tau}$$

Integrating on $ [x_0,x]  $ we have

$$\int_{x_0}^x(y(t)G(t))'dt=\int_{x_0}^xf(t)e^{\int_{x_0}^ta(\tau)d\tau}dt$$
$$y(x)G(x)-y(x_0)G(x_0)=\int_{x_0}^xf(t)e^{\int_{x_0}^ta(\tau)d\tau}dt$$

But $ G(x_0)=1 $ so

$$y(x)G(x)=y(x_0)+\int_{x_0}^xf(t)e^{\int_{x_0}^ta(\tau)d\tau}dt$$

or

$$y(x)e^{\int_{x_0}^xa(\tau)dt}=<br />
y(x_0)+\int_{x_0}^xf(t)e^{\int_{x_0}^ta(\tau)d\tau}dt$$

or

$$y(x)=e^{-\int_{x_0}^xa(\tau)d\tau}y(x_0)+e^{-\int_{x_0}^xa(\tau)d\tau}\int_{x_0}^xf(t)e^{\int_{x_0}^ta(\tau)d\tau}dt=$$
$$=e^{-\int_{x_0}^xa(\tau)d\tau}y(x_0)+\int_{x_0}^xf(t)e^{-\int_{x_0}^xa(\tau)d\tau}e^{\int_{x_0}^ta(\tau)d\tau}dt=e^{-\int_{x_0}^xa(\tau)d\tau}y(x_0)+\int_{x_0}^xf(t)e^{-\int_{t}^xa(\tau)d\tau}dt$$

We have finally

$$y(x)=e^{-\int_{x_0}^xa(\tau)d\tau}y(x_0)+\int_{x_0}^xf(t)e^{-\int_{t}^xa(\tau)d\tau}dt$$

This formula shows thet our solution is a sum of two functions.
There is a very important interpretation for each of these two functions.
First we note that if f(x)==0 we have just the solution of the homogeneous equation

$$y_0(x)=e^{-\int_{x_0}^xa(\tau)d\tau}y(x_0)$$

Let

$$y_p(x)=\int_{x_0}^xf(t)e^{-\int_{t}^xa(\tau)d\tau}dt$$

For $ y(x_0)=0 $ our $ y(x)=y_p(x) $ is a special solution of the non-homogeneous equation which take the value zero in $ x_0 $
It is not worthless to verify that from well known Lagrange formula
for

$$y_p(x)=\int_{x_0}^xf(t)e^{-\int_{t}^xa(\tau)d\tau}dt$$

we have

$$y'_p(x)=f(x)+\int_{x_0}^x-a(x)f(t)e^{-\int_{t}^xa(\tau)d\tau}dt=-a(x)y_p(x)+f(x)$$

so

$$y'_p(x)+a(x)y_p(x)=f(x)$$
No votes yet

Back to top