Suppose that $I=(-\infty,\infty)$ and $p,r$ are periodic of period $\zeta$. Define $P$ on $I$ as; $$ P(x)= \int_{0}^{x} p(t) dt , \forall x\in I $$ Then the following statements are true
- A solution $\phi$ of the equation $y'+p(x)y=r(x)$ on $I$ is periodic of period $\zeta$ $\leftrightarrow$ $\phi(0)=\phi(\zeta)$.
- The equation $y'+py=r$ has exactly one periodic solution of period $\zeta \leftrightarrow e^{P(\zeta)} \neq 1$.
- Every solution of the equation $y'+py=r$ is periodic of period $\zeta \leftrightarrow e^{P(\zeta)} =1$ AND $$ \int_{0}^{\zeta} e^{P(t)} r(t) dt=0 $$
I have already proved statement 1. I let $\phi$ be a solution of the equation $y'+p(x)y=r$. Then I defined another function $\psi$ ...showed that $\psi$ is a solution of it also and after some computation solved that $\phi$ is periodic of period $\zeta$ using the fact that $\chi(x)= ce^{-P(x)}$.
Can anyone assist me in proving parts 2 and 3? Or atleast point me in the right direction. I'm really confused as to how to even begin these parts
For the sake of completeness, I'll start by tackling item (1) even though our OP Jason Moore expresses some satisfaction with his own solution.
1.) Given that $\phi(x)$ is periodic of period $\zeta$, we have by definition that
$\phi(x + \zeta) = \phi(x), \; \forall x \in I = (-\infty, \infty); \tag 1$
thus if we take $x = 0$ we find that
$\phi(\zeta) = \phi(0). \tag{2}$
Next, we note that, if $\phi(x)$ satisfies
$y'(x) + p(x)y(x) = r(x), \tag 3$
that is,
$\phi'(x) + p(x)\phi(x) = r(x), \tag 4$
then we have
$\phi'(x + \zeta) + p(x + \zeta)\phi(x + \zeta) = r(x + \zeta), \tag 5$
which, since $p(x)$ and $r(x)$ are periodic of period $\zeta$, that is, since
$p(x + \zeta) = p(x), \; r(x + \zeta) = r(x), \tag 6$
implies that
$\phi'(x + \zeta) + p(x)\phi(x + \zeta) = r(x); \tag 7$
we see that $\phi(x)$ and $\phi(x + \zeta)$ each satisfy (3)-(4), with initial conditions $\phi(0)$ and $\phi(0 + \zeta) = \phi(\zeta)$ respectively; since we assume that $\phi(\zeta) = \phi(0)$, $\phi(x)$ and $\phi(x + \zeta)$ both satisfy the same differential equation (3)-(4) with identical initial conditions $\phi(0) = \phi(\zeta)$, and thus by uniqueness of solutions we must have
$\phi(x + \zeta) = \phi(x),\; \forall x \in I; \tag 8$
the periodicity of $\phi(x)$ given $\phi(\zeta) = \phi(0)$ is thus established.
2.) With
$P(x) = \displaystyle \int_0^x p(t)\; dt, \tag 9$
the solution $\phi(x)$ to (3)-(4) with initial condition $\phi(0)$ may be expressed in the form
$\phi(x) = e^{-P(x)}(\phi(0) + \displaystyle \int_0^x e^{P(t)}r(t) \; dt), \tag{10}$
which may easily be verified using direct differentiation; from (10),
$\phi'(x) = -P'(x)e^{-P(x)}(\phi(0) + \displaystyle \int_0^x e^{P(t)}r(t) \; dt) + e^{-P(x)}(\phi(0) + \displaystyle \int_0^x e^{P(t)}r(t) \; dt)'$ $= -P'(x)e^{P(x)}(\phi(0) + \displaystyle \int_0^x e^{P(t)}r(t) \; dt) + e^{-P(x)} e^{P(x)}r(x) = -p(x) \phi(x) + r(x), \tag{11}$
since, from (9),
$P'(x) = p(x); \tag{12}$
(11) is easily seen to be equivalent to (3)-(4); we also see that (10) implies
$\phi(0) = e^{-P(0)}(\phi(0) + \displaystyle \int_0^0 e^{P(t)}r(t) \; dt) = e^{P(0)}\phi(0) = \phi(0), \tag{13}$
since, again from (9),
$P(0) = \displaystyle \int_0^0 p(t) \; dt = 0; \tag{14}$
thus (10) is consistent with the initial conditions on (3)-(4) as well.
The formula (10) is in fact exceedingly well-known and its derivation my be found in many textbooks and at many web sites on the internet.
Now suppose
$e^{P(\zeta)} \ne 1; \tag{15}$
then
$e^{-P(\zeta)} \ne 1 \tag{16}$
as well, and if $x = \zeta$, (10) reads
$\phi(\zeta) = e^{-P(\zeta)}(\phi(0) + \displaystyle \int_0^\zeta e^{P(t)}r(t) \; dt), \tag{17}$
which may be re-written as
$e^{P(\zeta)} \phi(\zeta) - \phi(0) = \displaystyle \int_0^\zeta e^{P(t)} r(t) \; dt; \tag{18}$
we have seen in item (1) that $\phi(x)$ is a periodic solution with period $\zeta$ to (3)-(4) if and only if $\phi(\zeta) = \phi(0)$; if we choose $\phi(0)$, based upon (18), so that
$e^{P(\zeta)} \phi(0) - \phi(0) = (e^{P(\zeta)} - 1) \phi(0) = \displaystyle \int_0^\zeta e^{P(t)} r(t) \; dt, \tag{19}$
which is always possible by virtue of (15), then clearly (17) implies
$\phi(\zeta) = e^{-P(\zeta)}(\phi(0) + \displaystyle \int_0^\zeta e^{P(t)}r(t) \; dt) = \phi(0), \tag {20}$
and we may infer from item (1) that the trajectory initialized at $\phi(0)$ is periodic of period $\zeta$. Furthermore, it is now clear from (19)-(20) that $\phi(0)$ is the only possible initial condition such that $\phi(\zeta) = \phi(0)$.
Thus we see that $e^{P(\zeta)} \ne 1$ implies the existence of a unique periodic solution to (2)-(3).
Now suppose there is precisely one periodic trajectory which has initial value $\phi(0)$, and that $e^{P(\zeta)} = 1$. Then we also have $e^{-P(\zeta)} = 1$, and for this solution (20) must hold, whence
$\phi(\zeta) = \phi(0) + \displaystyle \int_0^\zeta e^{P(t)}r(t) \; dt = \phi(0), \tag {21}$
which in turn implies
$\displaystyle \int_0^\zeta e^{P(t)}r(t) \; dt = 0; \tag{22}$
but, via (17) we find
$\phi(\zeta) = \phi(0), \tag{23}$
no matter what value $\phi(0)$ may take. This shows that every solution must be periodic, which contradicts our assumption of a single periodic orbit. Therefore we cannot have $e^{P(\zeta)} = 1$ if the periodic solution is unique; we conclude that $e^{P(\zeta)} \ne 1$ in this case.
Our demonstration of item (2) is thus complete.
3.) We have already seen in item (2) that the combined requirements
$e^{P(\zeta)} = 1,\; \displaystyle \int_0^\zeta e^{P(t)} r(t) \; dt = 0 \tag{24}$
are sufficient to imply every solution to (3)-(4) is periodic; so, supposing every $\phi(x)$ obeying (3)-(4) also satisfies
$\phi(\zeta) = \phi(0),\tag{25}$
then (20), and hence (19), holds for every value of $\phi(0)$; but as we have seen under item (2), $e^{P(\zeta)} \ne 1$ implies that (18) is valid for precisely one value of $\phi(0)$; we thus conclude that
$e^{P(\zeta)} = 1, \tag{26}$
but then (19) immediatly yields
$\displaystyle \int_0^\zeta e^{P(t)} r(t) \; dt = 0, \tag{27}$
and we are done.