Theorem A. If ys(x) and y2(x) are two linearly independent solutions of
y’ + P(x)y’ + Oley = 0,
then the zeros of these functions are distinct and occur alternately—in the
sense that y,(x) vanishes exactly once between any two successive zeros of
yo(x), and conversely.
Proof. The argument rests primarily on the fact (see the lemmas in Section
15) that since y, and y, are linearly independent, their Wronskian
W(yis¥2) = y(e)y2e) — valxdy ile)
does not vanish, and therefore—since it is continuous—must have constant
sign. First, it is easy to see that y, and y, cannot have a common zero; for if
they do, then the Wronskian will vanish at that point, which is impossible.
We now assume that x, and x2 are successive zeros of y: and show that y,
vanishes between these points. The Wronskian clearly reduces to y\(x)y3(x)
at x, and x2, so both factors y,(x) and yi(x) are #0 at each of these points.
Furthermore, y3(x,) and y3(x,) must have opposite signs, because if y is
increasing at x, it must be decreasing at x,, and vice versa. Since the
The convexity arguments given above in connection with the
equation y" + y = 0 make it clear that in discussing the oscillation of
solutions it is convenient to deal with equations in which the first
derivative term is missing. We now show that any equation of the form
y+ Pix)y’ + Oa)y = 0 (uy)
u" + g(x)u =0 (12)
by a simple change of the dependent variable. It is customary to refer to
(11) as the standard form, and to (12) as the normal form, of a
homogeneous second order linear equation. To write (11) in normal
form, we put y(x) = u(x)v(x), so that y’ = uv’ + u'v and y” = uv" +
2u'v' + u"v. When these expressions are substituted in (11), we obtain
can be written as
uu" + (Qu! + Pu)u’ + (v" + Pu’ + Quyu = 0. (13)
On setting the coefficient of u’ equal to zero and solving, we find that
vse rae (14)
reduces (13) to the normal form (12) with
1 1
a(x) = QO) ~ 7 POY — 5 PG). (1s)
Since v(x) as given by (14) never vanishes, the above transformation of
(11) into (12) has no effect whatever on the zeros of solutions, and
therefore leaves unaltered the oscillation phenomena which are the
objects of our present interest.Theorem B. /f q(x) <0, and if u(x) is a nontrivial solution of u" +
q(x)u = 0, then u(x) has at most one zero.
Proof. Let x, be a zero of u(x), so that u(x») = 0. Since u(x) is nontrivial
(ie., is not identically zero), Theorem 14-A implies that u(x») # 0. For
the sake of concreteness, we now assume that u'(x,) > 0, so that u(x) is
positive over some interval to the right of xo. Since q(x) <0, u(x) =
—q(x)u(x) is a positive function on the same interval. This implies that the
slope u'(x) is an increasing function, so u(x) cannot have a zero to the right
of x», and in the same way it has none to the left of x9. A similar argument
holds when u’(x9) < 0, so u(x) has either no zeros at all or only one, and
the proof is complete.
Theorem C. Let u(x) be any nontrivial solution of u" + q(x)u = 0, where
q(x) > 0 for all x > 0. If
fae dx =, (16)
f
then u(x) has infinitely many zeros on the positive x-axis.
Proof. Assume the contrary, namely, that u(x) vanishes at most a finite
number of times for 0 1 exists with the
property that u(x) # 0 for all x = xo. We may clearly suppose, without any
loss of generality, that u(x) > 0 for all x = x9, since u(x) can be replaced
by its negative if necessary. Our purpose is to contradict the assumption by
showing that u'(x) is negative somewhere to the right of x;—for, by the
Theorem A. Let y(x) be a nontrivial solution of equation (1) on a closed
interval [a,b]. Then y(x) has at most a finite number of zeros in this interval.which is known as the Sturm comparison theorem, shows that this
behavior is typical in the sense that the solutions of (1) oscillate more
rapidly when q(x) is increased.
Theorem B. Let y(x) and z(x) be nontrivial solutions of
y+ q@)y =0
and
2" + r(x)z =0,
where q(x) and r(x) are positive functions such that q(x) > r(x). Then y(x)
vanishes at least once between any two successive zeros of z(x).
we consider Bessel’s equation
xy" + xy’ + (x? — p?)y = 0
in normal form
‘ 1 ~ 4p?
u"t+ (1 + Pu =0,
and compare this with uw” +u=0, then we at once have the next
theorem.
Theorem C. Let y,(x) be a nontrivial solution of Bessel’s equation on the
positive x-axis. If 0 < p < 1/2, then every interval of length x contains at
least one zero of yp(x); if p = 1/2, then the distance between successive
zeros of y,(x) is exactly x; and if p > 1/2, then every interval of length x
contains at most one zero of y,(x).
Theorem A. Let x, be an ordinary point of the differential equation
y” + P(x)y’ + Oe)y = 0, (11)
and let a, and a, be arbitrary constants. Then there exists a unique function
y(x) that is analytic at xo, is a solution of equation (11) in a certain
neighborhood of this point, and satisfies the initial conditions y(x,) = ay and
y'(%») = a. Furthermore, if the power series expansions of P(x) and Q(x)
are valid on an interval |x — xo|O, then the power series
expansion of this solution is also valid on the same interval.Theorem A. Assume that x = 0 is a regular singular point of the differential
equation (1) and that the power series expansions (3) of xP(x) and x*Q(x)
are valid on an interval |x| < _R with R > 0. Let the indicial equation (5)
have real roots m, and m, with m, = m,. Then equation (1) has at least one
solution
(6)
on the interval 0