Sturm's Separation and Comparison Theorems
Sturm's Separation and Comparison Theorems
y ′′ + q(x)y = 0, (7.1)
where q is a continuous function. It is true that any second order ODE is essentially of this form,
via a change of variable, we shall not discuss further on this point. Further details can be found in
Myint-U’s book [20]. The results that we are going to prove also hold for operators in self-adjoint
form.
Proof :
Proof of (i) is an exercise; assuming that the zero set of a solution has a limit point, one can
show that the solution is a trivial solution by virtue of satisfying an IVP with initial values zero.
Proof of (ii) is by contradiction. Assuming that φ2 does not have a zero on (x1 , x2 ), we conclude
that φ2 does not have a zero on [x1 , x2 ]. This is due to the fact that φ1 and φ2 is a fundamental
pair (equivalently, linearly independent).
Therefore we can define a function ψ on [x1 , x2 ] by
φ1 (x)
ψ(x) = . (7.3)
φ2 (x)
Then the function ψ has the properties:
(i) ψ is continuous on [x1 , x2 ].
(ii) ψ ′ exists on (x1 , x2 ).
(iii) ψ(x1 ) = ψ(x2 ) = 0.
59
60 7.2. Sturm’s comparison theorem
Therefore by Rolle’s theorem, there exists a c such that x1 < c < x2 such that ψ ′ (c) = 0.
This means wronskian is zero at c and this is not possible as φ1 and φ2 is a fundamental pair
(equivalently, linearly independent). Therefore, φ2 has at least one zero on (x1 , x2 ).
We claim that φ2 does not have more than one zero in (x1 , x2 ). If the claim were not true, then
there would exist at least two zeroes of φ2 in (x1 , x2 ). With out loss of generality, let x3 and
x4 be two consecutive zeroes of φ2 in (x1 , x2 ), such that x1 < x3 < x4 < x2 . Then by previous
arguments, φ1 will have a zero in (x3 , x4 ) and thus contradicting the fact that x1 and x2 are two
consecutive zeros of φ1 .
Corollary 7.2 If φ1 and φ2 are as in the above theorem, then zeros of φ1 separate and are
separated by those of φ2 .
However the above theorem does not address the existentce of zeros! The following examples
illustrate this point.
Example 7.3
y ′′ − y = 0. (7.4)
x −x
(i) φ1 (x) = e and φ2 (x) = e are linearly independent solutions. None of them have zeros!
(ii)ψ1 (x) = sinh x and ψ2 (x) = cosh x are linearly independent solutions. ψ1 has one zero, ψ2 has
no zero on R.
Theorem 7.4 If q(x) ≤ 0 on an interval I, then no non-trivial solution of y ′′ + qy = 0 can have
two zeros on I.
Proof :
Suppose if possible, that a non-trivial soluton y has at least two zeroes on I. WLOG let x1 < x2
be two consecutive zeroes of y, and that y > 0 on (x1 , x2 ).
This implies that
y ′ (x1 ) > 0, and y ′ (x2 ) < 0. (7.5)
The inequalities in (7.5) follow from continuity of y ′ . We indicate the proof of the first inequality,
for instance. We prove that y ′ (x1 ) ≤ 0 is not possible, for, if y ′ (x1 ) = 0 then since y(x1 ) = 0
already, it follows that y ≡ 0, a contradiction. On the other hand, if y ′ (x1 ) < 0 then y will be
negative for x > x1 and x near-by x1 , again contradicting the assumption that y > 0 on (x1 , x2 ).
Thus
y ′′ (x) = −q(x)y(x) ≥ 0 on (x1 , x2 ). (7.6)
The last inequality implies that y ′ is an increasing function on (x1 , x2 ). Hence
0 > y ′ (x2 ) ≥ y ′ (x1 ) > 0. (7.7)
This is a contradiction, and thus the non-trivial solution y can not have two zeroes.
Remark 7.5 Trivial solution is the only solution of y ′′ + qy = 0 with q(x) ≤ 0 on an interval I,
which has two zeroes.
Proof :
Let x1 and x2 with x1 < x2 be consecutive zeroes of φ1 . Assume WLOG φ1 > 0 on (x1 , x2 )
(if not, consider −φ1 which has these properties). Consequently, by arguments in the proof of
Theorem 7.4,
φ′1 (x1 ) > 0, and φ′1 (x2 ) < 0. (7.10)
Suppose that φ2 does not have a zero on (x1 , x2 ). WLOG let φ2 > 0 on (x1 , x2 ). Multiplying the
equation satisfied by φ1 , with φ2 , and vice versa, and then subtract the two equations we get
Note that LHS of (7.12) is non-positive. The RHS is strictly positive unless q1 (x) ≡ q2 (x) on
(x1 , x2 ). Therefore, if q1 (x) 6≡ q2 (x) on (x1 , x2 ), we arrive at a contradiction. This finishes the
proof of theorem.
Exercise 7.7 Can we prove that LHS of (7.12) is negative? If yes, we can weaken the hypothesis
of our theorem, thereby strengthening the result.
Definition 7.8 With q1 and q2 as in the statement of Theorem 7.6, Equation (7.9) is said to be a
Sturm majorant of equation (7.8) on I. Equation (7.8) is said to be a Sturm minorant of equation
(7.9) on I.
Remark 7.9 A non-trivial solution of the equation (py ′ )′ + qy = 0 has infinitely many zeros on
an interval I if and only if every (non-trivial) solution has infinitely many zeros on I. This follows
from Sturm’s comparison theorem.
Remark 7.11 An equation (py ′ )′ +qy = 0 is oscillatory on an interval I if and only if all solutions
of the equation have infinitely many zeros on I.
Exercise 7.12 (i) The equation y ′′ + y = 0 is oscillatory on R.
(ii) The equation y ′′ − y = 0 is non-oscillatory.
(iii) The equation y ′′ = 0 is non-oscillatory.
(iv) The equation y ′′ + xy = 0 is oscillatory on (1, ∞). Prove also that any of its non-trivial
solutions have a zero between nπ and (n + 1)π for any integer n.
For x > 0, making a change of variable y = √v , the equation (7.13) transforms into
x
1 − 4ν 2
v ′′ + 1 + v = 0 (x > 0). (7.14)
4x2
√
(to obtain the above equation, start differentiating the equation xy = v).
Case 1: 0 < ν < 1/2 In this case, compare (7.14) with
y ′′ + y = 0, (7.15)
which has a solution sin x with zeros at x = nπ, n ∈ N. Therefore a solution v of (7.14) has at
least one zero on each of the open intervals ((n − 1)π, nπ), n ∈ N.
Case 2: ν > 1/2 In this case, compare once again (7.14) with
y ′′ + y = 0, (7.16)
and conclude that between any two consecutive zeros, α and β of v, there exists at least one zero
of sin x. Thus, we have α < nπ < β for some n ∈ N.
Remark 7.13 (i) Sturm’s comparison theorem guarantees the existence of at least one zero.
(ii) The assumption q2 (x) ≥ q1 (x) can not be dropped. Consider the equations on x ≥ 0, y ′′ +y = 0
(q1 (x) = 1), and v ′′ − v = 0 (q2 (x) = −1); and let y and v be their non-trivial solutions
respectively. Between any two zeros of y, v does not admit a zero.
(iii) Consider the equations on x ∈ R, y ′′ + y = 0 (q1 (x) = 1), and v ′′ + 4v = 0 (q2 (x) = 4); and
let y = sin x and v = sin 2x are their non-trivial solutions respectively. But there is no zero of
y between two consecutive zeros of v.