Memoirs on Differential Equations and Mathematical Physics

Volume 73, 2018, 123–129

V. V. Rogachev

ON THE EXISTENCE OF SOLUTIONS TO HIGHER-ORDER

REGULAR NONLINEAR EMDEN–FOWLER TYPE EQUATIONS
WITH GIVEN NUMBER OF ZEROS ON THE PRESCRIBED INTERVAL
Abstract. The existence of solutions with a given number of zeros to higher-order regular-nonlinear
Emden–Fowler type equations is proven.1

2010 Mathematics Subject Classification. 34C11, 34E10

Key words and phrases. Higher-order Emden–Fowler type differential equations, regular nonlin-
earity, boundary value problem.

ÒÄÆÉÖÌÄ. ÌÀÙÀËÉ ÒÉÂÉÓ ÄÌÃÄÍ-×ÀÖËÄÒÉÓ ÔÉÐÉÓ ÀÒÀßÒ×ÉÅÉ, ÒÄÂÖËÀÒÖËÉ ÃÉ×ÄÒÄÍÝÉÀËÖÒÉ

ÂÀÍÔÏËÄÁÄÁÉÓÈÅÉÓ ÌÔÊÉÝÃÄÁÀ ÉÓÄÈÉ ÀÌÏÍÀáÓÍÄÁÉÓ ÀÒÓÄÁÏÁÀ, ÒÏÌËÉÓ ÍÖËÄÁÉÓ ÒÀÏÃÄÍÏÁÀ
ÌÏÝÄÌÖË ÓÀÓÒÖË ÛÖÀËÄÃÛÉ ßÉÍÀÓßÀÒ ÃÀÓÀáÄËÄÁÖËÉ ÒÉÝáÅÉÓ ÔÏËÉÀ.

1 Reported on Conference “Differential Equation and Applications”, September 4–7, 2017, Brno
Existence of Solutions to Emden–Fowler Type Equations with Given Number of Zeros. . . 125

1 Introduction
Consider the equation
y (n) + p(t, y, y ′ , . . . , y (n−1) )|y|k sgn y = 0, (1.1)
where n ≥ 2, k ∈ (1, +∞), the function p(t, y1 , y2 , y3 , . . . , yn ) ∈ C(R n+1
) is Lipschitz continuous in
(y1 , y2 , y3 , . . . , yn ) and for some m, M > 0 satisfies the inequalities

0 < m ≤ p(t, y1 , y2 , . . . , yn ) ≤ M < +∞.

The problem of the existence of solutions to (1.1) with the given number of zeros on the prescribed
domain is investigated.
Asymptotic classification of solutions to (1.1) with n = 3, 4, k ∈ (1, +∞), p(t, y, y ′ , . . . , y (n−1) ) ≡
const and with n = 3, k ∈ (0, 1), p(t, y, y ′ , . . . , y (n−1) ) ≡ const is provided in [1, 3] by I. Astashova.
Later, the existence of quasiperiodic solutions to the regular (k ∈ (1, +∞)) higher-order Emden–Fowler
type equations has been proved in [2].
Using [1], the existence of solutions with the given number of zeros was proved for the case of third-
and fourth-order equations with the constant coefficient p and with k ∈ (0, 1) ∪ (1, +∞) (see [4]).
Later, the case of the higher-order differential equation (1.1) with the constant potential and regular
nonlinearity (k > 1) was considered in [5]. In [6], the existence of solutions with the given number of
zeros was proved for (1.1) with n = 3, k ∈ (1, +∞). In [7], the existence of such solutions was proved
for the equation with k ∈ (0, 1).
Now we generalize these results to the case of equation (1.1).

2 Main result
Theorem 2.1. For any real a and b satisfying −∞ < a < b < +∞ and any integer S ≥ 2, equation
(1.1) has a solution defined on the segment [a, b], vanishing at its end points a, b and having exactly
S zeros on [a, b].

3 Preliminary results
The following statements are used to prove the main theorem.

Lemma 3.1 (Generalization of 7.1 from [1]). Let y(t) be a solution to (1.1). If for some t0 the
inequalities
y(t0 ) ≥ 0, y ′ (t0 ) > 0, y ′′ (t0 ) ≥ 0, . . . , y (n−1) (t0 ) ≥ 0
hold, then there is a local supremum of y at some point t′0 > t0 satisfying the inequalities

t′0 − t0 ≤ (µy ′ (t0 ))− k+n−1 ,

k−1

y(t′0 ) > (µy ′ (t0 )) k+n−1 ,

n

where µ > 0 is a constant depending only on n, k, m, M .

Lemma 3.2 (Generalization of 7.2 from [1]). Let y(t) be a solution to (1.1). If for some t′0 the
inequalities
y(t′0 ) > 0, y ′ (t′0 ) ≤ 0, . . . , y (n−1) (t′0 ) ≤ 0
hold, then y is equal to zero at some point t0 > t′0 satisfying the inequalities

t0 − t′0 ≤ (µy(t′0 ))−

k−1
n ,
y ′ (t0 ) < −(µy(t′0 ))
k+n−1
n ,

where µ > 0 is a constant depending only on n, k, m, M .

126 V. V. Rogachev

Lemma 3.3 (Generalization of 7.3 from [1]). Under the conditions of Lemmas 3.1, 3.2, for any t1 > t0
such that y(t0 ) = 0, y(t1 ) = 0, the inequality
|y ′ (t1 )| > Q|y ′ (t0 )|
holds, where Q > 1 is a constant depending only on k, m, M .
Lemma 3.4. Let D be a subset of Rn and D e be a subset of Rn+1 . Suppose that for any c ∈ D
e Consider a continuous function f (c, x) : D
there exists xc > 0 such that {c} × [0, xc ] ⊂ D. e → R and
introduce the following conditions:
• f (c, 0) = 0 for any c ∈ D,
• for every c ∈ D, there exists a point x1 (c) ∈ (0, xc ) such that f (c, x1 (c)) = 0 and f (c, x) ̸= 0
whenever x ∈ (0, x1 (c)),
df
• f (c, x) is differentiable in x, and dx (c, x1 (c)) ̸= 0 for all c ∈ D.
If these conditions hold, then x1 (c) : D → R is a continuous function.
Proof. By definition, x1 (c) describes the distance from 0 to the first zero of the function f (c, · ). The
existence of such a zero is stated in the second condition of the lemma. Therefore x1 (c) is actually a
function (its value is defined for every c ∈ D), but, perhaps, discontinuous. We intend to prove that
x1 (c) is a continuous function.
At every point (c, x1 (c)) ∈ D, e the function f (c, x) fulfills the conditions of the Implicit Function
Theorem. Therefore for any e c ∈ D there exist rectangular neighborhoods U ⊂ D of e c, V ⊂ R of x1 (e
c),
and a continuous function gec (c) : U → V such that for all (c, x) ∈ U × V the conditions f (c, x) = 0
and x = gec (c) are equivalent.
It is clear that x1 (e c), but we have to prove that x1 (c) ≡ gec (c) in some neighborhood of e
c) = gec (e c.
(We know that f (c, gec (c)) = 0, but the zeros of f (c, · ) provided by gec (c) may not be the zeros closest
to the point x = 0.)
We will prove this by contradiction. Suppose that in any punctured neighborhood of some point
c∗ ∈ D there exists a point c such that gc∗ (c) ̸= x1 (c). Then we have an infinite set {cα } such that for
every cα the inequality gc∗ (cα ) ̸= x1 (cα ) holds. We can extract from {cα } a sequence {cn } tending to
the point c∗ . The implicit function theorem for f (c, x) takes place in a neighborhood U × V of the
point (c∗ , x1 (c∗ )).
Now we look closely at the set {(cn , x1 (cn ))}. It is a sequence in D, e which cannot enter U × V ,
because otherwise the condition f (cn , x1 (cn )) = 0 inside U × V contradicts the very definition of {cn }.
At the same time, the points (cn , x1 (cn )) cannot be above the graph of gc∗ (c) and above U × V by
the definition of the function x1 (c).
So, the sequence {x1 (cn )} is bounded by zero from below and by inf V < x1 (c∗ ) from above.
Hence {x1 (cn )} has a limit inferior x∗ < x1 (c∗ ). We extract a subsequence {x1 (cni )} tending to
the above limit and then consider a sequence {(cni , x1 (cni ))}. The function f (c, x) is continuous,
f (cni , x1 (cni )) = 0, and (cni , x1 (cni )) → (c∗ , x∗ ) as i → ∞. Therefore, f (c∗ , x∗ ) = 0. But at the same
time we have x∗ < x1 (c∗ ), and this contradicts the conditions of the lemma. Therefore, the point c∗ ,
in fact, does not exist.
This means that for every point e c ∈ D the equality x1 (c) ≡ gec (c) is true in some neighborhood of e c.
Every function gec (c) is continuous near e c. Therefore, x1 (c) is continuous at every point c ∈ D.

3.1 Proof of the main result

Proof of Theorem 2.1. Consider a maximally extended solution y(t) to (1.1) with initial data y (i) (a) =
yi , i ∈ 0, n − 1.
It follows from Lemmas 3.1–3.3 that if the inequalities
y(t0 ) ≥ 0, y ′ (t0 ) > 0, y ′′ (t0 ) ≥ 0, . . . , y (n−1) (t0 ) ≥ 0
hold at some point t0 , then there exists a point t1 > t0 such that
y(t1 ) = 0, y ′ (t1 ) < 0, y ′′ (t1 ) ≤ 0, . . . , y (n−1) (t1 ) ≤ 0
Existence of Solutions to Emden–Fowler Type Equations with Given Number of Zeros. . . 127

and
t1 − t0 ≤ (µ′ y ′ (t0 ))− k+2 ,
k−1

where µ′ > 0 and Q > 1 are constants depending only on k, m, and M .

The analogous statement takes place if

y(t0 ) ≤ 0, y ′ (t0 ) < 0, y ′′ (t0 ) ≤ 0, . . . , y (n−1) (t0 ) ≤ 0.

Hence, if y0 = 0 and yi > 0 for i ∈ 1, n − 1, then y(t) is an oscillating solution, i.e., it has an
infinite sequence of zeros {a, t1 , t2 , . . .}. In the sequel, y0 = 0 and yi > 0 for i ∈ 1, n − 1.
We denote the distance between zeros by Li = ti − ti−1 . The distance from a to the (S − 1)st zero
is a function
∑
S−1
L(y1 , y2 , . . . , yn−1 ) = Lj (y1 , y2 , . . . , yn−1 ),
j=1

and its value depends on the initial data of the solution y(t).
If L(y1 , y2 , . . . , yn−1 ) = b − a, then the solution y(t) has exactly S zeros on [a, b]. To prove the
theorem we have to prove that for any b and a the last equation has at least one solution.
First, notice that L is a continuous function. If we rewrite (1.1) as a system of first-order ODEs,
that system will satisfy the conditions of the continuous dependence on initial data theorem [8, § 7,
Theorem 6]. By Y (t, a, y0 , y1 , y2 , . . . , yn−1 ) we denote a maximally extended solution to (1.1) with
initial data y (i) (a) = yi , i ∈ 0, n − 1. Therefore, Y (t, a, y0 , y1 , y2 , . . . , yn−1 ) and n of its derivatives in
t are continuous functions on their domains.
Are the conditions of Lemma 3.4 fulfilled? Put
{ }
D = (y1 , y2 , . . . , yn−1 ) | yi > 0 ⊂ Rn−1 .

For every such (y1 , y2 , . . . , yn−1 ) we have already proved the existence of the first zero t1 , which
satisfies y ′ (t1 ) ̸= 0. Further, there exists the second zero t2 , and for D e ⊂ Rn we take the area above
D × {0} and under the graph of t2 (y1 , y2 , . . . , yn−1 ). Obviously, Y (a, a, y0 , y1 , y2 , . . . , yn−1 ) = 0, and
e (Here a is fixed and y0 is equal to zero.)
Y (t, a, y0 , y1 , y2 , . . . , yn−1 ) is defined on D.
The conditions of Lemma 3.4 are fulfilled, hence t1 (y1 , y2 , . . . , yn−1 ), or L1 is a continuous function
on D. It is possible to prove by using Lemma 3.4 that all Li , and therefore L are continuous. For L2 ,
for example, notice that y(t1 (y1 , y2 , . . . , yn−1 )), y ′ (t1 (y1 , y2 , . . . , yn−1 )), . . . , y (n−1) (t1 (y1 , y2 , . . . , yn−1 ))
are also continuous, because they are compositions of continuous functions Y (i) (·, a, y0 , y1 , y2 , . . . , yn−1 )
and t1 (y1 , . . . , yn−1 ).
Now we are to find an upper estimate of L. It is already proved that

L1 ≤ (µ′ y1 )− k+n−1 .
k−1

It follows from Lemma 3.3 that

|y ′ (ti )| ≥ Qi |y ′ (a)|.
Consider Li . Since − k+n−1
k−1
< 0, we have

Li ≤ (µ′ Qi−1 y1 )− k+n−1 = (Q− k+n−1 )i−1 (µ′ y1 )− k+n−1 .

k−1 k−1 k−1

e = Q− k+n−1
Put Q
k−1
. Since Q > 1, − k+n−1
k−1 e < 1, the upper estimates of Li
< 0, and therefore 0 < Q
form a decreasing geometric progression. Therefore,

eS
1−Q
(µ′ y ′ (a))− k+n−1 = c1 y ′ (a)− k+n−1 ,
k−1 k−1
L = L1 + L2 + · · · + LS−1 ≤
1−Qe
L < c1 y ′ (a)− k+n−1 ,
k−1
(3.1)

where c1 is a constant depending on n, k, m, M , and S.

128 V. V. Rogachev

To get a lower estimate of L it is sufficient to make a lower estimation of L1 . Consider a point

t′0 ∈ [a, t1 ] such that y ′ (t′0 ) = 0. On the segment [t′0 , t1 ], the derivatives y ′ , y ′′ are non-positive.
Therefore,
∫t1
′ ′ ′ ′
Qy (a) < |y (t1 )| = |y (t1 )| − |y (t′0 )| = |y ′′ (ξ)| dξ < |t1 − t′0 | max
′
|y ′′ |.
[t0 ,t1 ]
t′0

We must get an upper estimate of max

′
|y ′′ |. Notice the behaviour of the derivatives of y(t) as t
[t0 ,t1 ]
goes from a to t1 . On the segment [a, t1 ], the inequality y(t) > 0 holds. First, near a, every derivative,
except y (n) , is positive. It follows that y (n−1) is decreasing and after some point the inequality
y (n−1) < 0 holds, when y (n) is still negative. Hence, now y (n−2) starts to decrease, and we can repeat
the same steps, until the solution y intersects the 0 − t-axis, i.e., when we move t from a to t1 , the
derivatives change their signs in order and higher-order derivatives change sign before low-order ones.
Therefore, on [t′0 , t1 ], the second derivative of the solution y is negative, because on the segment [t′0 , t1 ]
the first derivative y ′ (t) < 0.
Denote |y|k sgn y by |y|k± . All initial data are positive, hence

(t − a)2 (t − a)n−3
0 > y ′′ (t) = y2 + y3 (t − a) + y4 + · · · + yn−1
2! (n − 3)!
∫t ∫t
− · · · p(t, y, . . . , y (n−1) )|y|k± (d t)n−2
a a
∫t ∫t
>− ··· p(t, y, . . . , y (n−1) )|y|k± (d t)n−2 > −M |t − a|n−2 max |y|k ,
[a,t1 ]
a a

whence
max |y ′′ | < M |t1 − a|n−2 max |y|k .
[t′0 ,t1 ] [a,t1 ]

Now we get an upper estimation of max |y|k . The inequality y(t) > 0 holds on [a, t1 ], whence
[a,t1 ]

∫t ∫t
(t − a)2 (t − a)n−1
y(t) = y1 (t − a) + y2 + · · · + yn−1 − ··· p(ξ, y, . . . , y (n−1) )|y|k± (dξ)n
2! (n − 1)!
a a
(t − a)2 (t − a)n−1
< y1 (t − a) + y2 + · · · + yn−1 .
2! (n − 1)!
Therefore,
( (t1 − a)n−1 )k
max |y(t)|k < y1 (t1 − a) + · · · + yn−1 .
[a,t1 ] (n − 1)!
Combining both estimates, we get
( (t1 − a)n−1 )k
Qy1 < M |t1 − t′0 ||t1 − a|n−2 y1 (t1 − a) + · · · + yn−1 .
(n − 1)!
By definition, t1 − a = L1 and |t1 − t′0 | < L1 , hence
( Ln−1 )k
Qy1 < M Ln−1 y1 L1 + · · · + yn−1 1
.
1
(n − 1)!
Suppose y1 = y2 = · · · = yn−1 and y1 is a variable. In this case,
( Ln−1 )k
M Ln−1 L1 + · · · + 1
> Qy11−k .
1
(n − 1)!
Existence of Solutions to Emden–Fowler Type Equations with Given Number of Zeros. . . 129

In the left-hand side of the inequality we have the function of L1 which is defined for every L1 > 0,
is equal to zero when L1 = 0, and is monotonically increasing. The value of the right-hand side may be
arbitrarily large as y1 is arbitrarily small. Hence, for any λ > 0, we can choose initial data providing
L > λ.
But, due to (3.1), for any λ > 0 we can choose initial data providing 0 < L < λ. There-
fore, the value of L(y1 , y2 , . . . , yn−1 ) may be arbitrarily large, arbitrarily small, and, at the same
time, L(y1 , y2 , . . . , yn−1 ) is proven to be continuous. Thus, we conclude that the range of values of
L(y1 , y2 , . . . , yn−1 ) is (0, +∞). Therefore, the equation

L(y1 , y2 , . . . , yn−1 ) = b − a

can be resolved for any b > a. This proves the theorem.

Acknowledgment
The author is grateful to Prof. I. V. Astashova for her support.

References
[1] I. V. Astashova, Qualitative properties of solutions to quasilinear ordinary differential equations.
(Russian) In: Astashova I. V. (Ed.) Qualitative Properties of Solutions to Differential Equations
and Related Topics of Spectral Analysis, pp. 22–290, UNITY-DANA, Moscow, 2012.
[2] I. Astashova, On quasi-periodic solutions to a higher-order Emden-Fowler type differential equa-
tion. Bound. Value Probl. 2014, 2014:174, 8 pp.
[3] I. V. Astashova, On asymptotic classification of solutions to nonlinear regular and singular third-
and fourth-order differential equations with power nonlinearity. In: Differential and difference
equations with applications, 191–203, Springer Proc. Math. Stat., 164, Springer, [Cham], 2016.
[4] I. V. Astashova and V. V. Rogachev, On the number of zeros of oscillating solutions of the third-
and fourth-order equations with power nonlinearities. (Russian) NelīnīǐnīKoliv. 17 (2014), no. 1,
16–31; translation in J. Math. Sci. (N.Y.) 205 (2015), no. 6, 733–748.
[5] V. V. Rogachev, On existence of solutions with given number of zeros to high order Emden–Fowler
type equation. Abstracts of Conference on Differential and Difference Equations and Applications,
41–42, Jasna, Slovak Republic, 2014.
[6] V. V. Rogachev, On the existence of solutions with prescribed number of zeros to regular nonlinear
Emden–Fowler type third-order equation with variable coefficient. (Russian) Vestnik SamGU
6(128) (2015), 117–123.
[7] V. V. Rogachev, On existence of solutions to higher-order singular nonlinear Emden-Fowler type
equation with given number of zeros on prescribed interval. Funct. Differ. Equ. 23 (2016), no.
3-4, 141–151.
[8] A. F. Filippov, Introduction to the Theory of Differential Equations. (Russian) 3rd ed.,
Ser. MSU classical textbooks, URSS, Moscow, 2010; https://books.google.ru/books?id=
f2xkkgAACAAJ.

(Received 27.09.2017)

Author’s address:

Lomonosov Moscow State University, 1 Leninskiye Gory, Moscow, Russia.

E-mail: [email protected]