- Research
- Open access
- Published:
Boundary value conditions for linear differential equations with power degenerations
Boundary Value Problems volume 2020, Article number: 110 (2020)
Abstract
On the interval \([0,1]\) we consider the nth order linear differential equation, the coefficient of the highest derivative of which is equivalent to the power function \(t^{\mu }\) when \(t\rightarrow 0\). The main aim of the paper is to pose “generalized” Cauchy conditions for the given equation at the point of singularity \(t=0\), which would be correct for any \(\mu >0\).
1 Introduction
Let us consider the following nth order linear differential equation:
where the coefficients \(a_{i}(\cdot )\), \(i=0,1,\ldots,n\), and the right-hand side \(f(\cdot )\) are continuous functions on \([0,1]\). Moreover, \(a_{n}(t)>0\) when \(t\in (0,1]\) and \(a_{n}(t)\) is equivalent to \(t^{\mu }\) when \(t\rightarrow 0\), i.e., equation (1) has a singularity at the point \(t=0\) of order μ. It is known that if \(0<\mu <1\), then all solutions of equation (1) belong to \(C^{n}[0,1]\). Hence, in this case we can pose the same boundary conditions for (1) as for a nonsingular equation. In particular, at the point of singularity the following Cauchy conditions can be posed:
In the case \(\mu \geq 1\), there are, in general, no finite limits \(\lim_{t\rightarrow 0} y^{(i)}(t)\) for all \(i=0,1,\ldots,n-1\). Therefore, the Cauchy conditions have no meaning. The main aim of the paper is to pose “generalized” Cauchy conditions for (1) at \(t=0\), which would be correct for any \(\mu >0\).
The problem will be solved in the following way. Let \(\gamma _{i}<1\), \(i=0,1,\ldots,n-1\), be an arbitrary set of n real numbers and \(\gamma _{n}=1\). Suppose that the set of numbers \(\overline{\alpha }=(\alpha _{0},\alpha _{1},\ldots,\alpha _{n})\) is such that
and \(\alpha _{0}\) is calculated from the equality
That gives
Using this set of numbers α̅ for \(y(t)\in C^{n}(0,1]\), we construct the following operations:
We call this differential operation (or operator) \(D_{\overline{\alpha }}^{i}\)the multiweighted derivative of the functionyof orderi,\(i=0,1,\ldots,n\). At the point of singularity \(t=0\) of equation (1) we pose the boundary conditions
where each \(D_{\overline{\alpha }}^{i}y(0)=0\), \(i=0,1,\ldots,n-1\), is understood in the sense of the existence of the finite limit \(\lim_{t\rightarrow 0}D_{\overline{\alpha }}^{i}y(t)=D_{ \overline{\alpha }}^{i}y(0)\). Conditions (4) are the required “generalized” Cauchy conditions, and in the paper we prove that problem (1) and (4) with (3) has a unique solution.
Let us note that the operator \(D_{\overline{\alpha }}^{n}\) gives the basis for a space \(W_{p,\overline{\alpha }}^{n}=W_{p,\overline{\alpha }}^{n}(I)\) of functions \(y:I\rightarrow \mathbb{R}\) with the finite semi-norm
where \(1< p<\infty \), \(I=(0,1)\) or \(I=(1,\infty )\).
The idea to study function spaces with the purpose to apply them to different problems concerning differential equations appeared in works by Sobolev in the thirties. From this time the theory of Sobolev spaces has been developed to be a very powerful instrument for solving boundary value problems of differential equations. Moreover, such concept as a “weight function” was introduced to take care of different problems connected to singularities. Correspondingly, Kudryavtsev presented a fairly complete theory of one-dimensional Sobolev spaces with power weights (see, e.g., [8–15] and the references given there). He considered a space \(L_{p,\gamma }^{n}=L_{p,\gamma }^{n}(I)\) of functions \(y:I\rightarrow \mathbb{R}\), which on the interval I have nth order derivative with the finite semi-norm
For the interval \(I=(0,1)\) it was shown that if \(\gamma >n-\frac{1}{p}\), then the function y, in general, does not have finite limit value when \(t\rightarrow 0\). Thus, the introduction of an additional function as “weight” does not always solve singularity problems. This argumentation served as a motivation for constructing the operator \(D_{\overline{\alpha }}^{n}\) and defining the space \(W_{p,\overline{\alpha }}^{n}\) with \(n+1\) weights. The theory of the function space \(W_{p,\overline{\alpha }}^{n}\) generalizes the previous theory of the function space \(L_{p,\gamma }^{n}\). In the series of works (see, e.g., [1–6]) there were considered similar problems for the \(W_{p,\overline{\alpha }}^{n}\) as those considered by Kudryavtsev for the space \(L_{p,\gamma }^{n}\). The results for the space \(W_{p,\overline{\alpha }}^{n}\) cover such singularities that cannot be handled by one weight, but can be handled with many weights, and they give the basis for the main result of the present paper. More precisely, in the paper [5] it was shown that the properties of the space \(W_{p,\overline{\alpha }}^{n}\) are dependent on the values \(\gamma _{i}\), \(i=0,1,\ldots,n-1\), in accordance with which we have the following three cases of the degeneration of the weight functions \(t^{\alpha _{i}}\), \(i=1,2,\ldots,n\):
- 1.
\(\gamma _{\max }=\max_{0\leq i\leq n-1}\gamma _{i}<1- \frac{1}{p}\) (weak degeneration);
- 2.
\(\gamma _{\min }=\min_{0\leq i\leq n-1}\gamma _{i}>1- \frac{1}{p}\) (strong degeneration);
- 3.
\(\gamma _{\min }<1-\frac{1}{p}< \gamma _{\max }\) (mixed degeneration).
Then in the paper [6] it was proved that, for any function \(y\in W_{p,\overline{\alpha }}^{n}\), the condition \(\gamma _{\max }<1-\frac{1}{p}\) (weak degeneration) is necessary and sufficient for the existence of limit values \(\lim_{t\rightarrow 0}D_{\overline{\alpha }}^{i}y(t)=D_{ \overline{\alpha }}^{i}y(0)\) for all \(0\leq i\leq n-1\). If we resolve the equalities in (3) with respect to \(\alpha _{i}\), \(i=1,2,\ldots,n\), we get the equalities in (2). Moreover, by the assumption the values \(\gamma _{i}\), \(i=0,1,\ldots,n-1\), satisfy the condition \(\gamma _{i}<1\). Thus, we have a weak degeneration of the weight functions \(t^{\alpha _{i}}\), \(i=1,2,\ldots,n\), which guarantees the existence of characteristics at the singular point \(t=0\).
The paper is organized as follows: In Sect. 2 we collect all the required notations, definitions, and statements; in Sect. 3 we state and prove our main result concerning the existence of a unique solution of problem (1) and (4).
2 Preliminaries
Let us introduce the following family of functions \(K_{k}(x,t)\), \(k=0,1,\ldots,n\), assuming that \(K_{n}(x,t)\equiv 1\), \(K_{n-1}(x,t)=\int _{x}^{t} s^{-\alpha _{n-1}}\,ds\), \(K_{n-2}(x,t)=\int _{x}^{t} y^{-\alpha _{n-2}} \int _{x}^{y}s^{- \alpha _{n-1}}\,ds \,dy\) and, in general,
for \(t>x\). Moreover, we assume that \(K_{k}(x,t)=0\), \(k=0,1,\ldots,n-1\), for \(t\le x\).
Let \(n-1\geq k\geq 0\). If in the integrals of \(K_{k}(x,t)\) we successively change the variables \(t_{n-1}=x\tau _{n-1}\), \(t_{n-2}=x\tau _{n-2}\), …, \(t_{k+1}=x\tau _{k+1}\), we get
and
i.e., the function \(x^{-\alpha _{n}}K_{k}(x,t)\) has a singularity at zero of order \(x^{-\gamma _{k}}\) for \(0<\gamma _{k}<1\).
Let \(n-1\geq k\geq 0\), \(i=k,k+1,\ldots,n-1\). Changing the order of integration and using (3), for \(t>0\), we get
where \(d_{k}=\frac{1}{\prod_{i=k}^{n-1}(1-\gamma _{i})}\).
Lemma 1
Let\(\gamma _{n-1}<1\). Then, for\(y\in C^{n}(0,1]\)satisfying the condition\(\sup_{0\leq t\leq 1} |D_{\overline{\alpha }}^{n}y(t)|< \infty \), there exists\(D_{\overline{\alpha }}^{n-1}y(0)\)and the estimate
holds.
Proof. Indeed, by assumption it follows that the function \(t^{-\alpha _{n}}D_{\overline{\alpha }}^{n}y(t)\) is absolutely summable on the interval \((0,1]\). Therefore, by Newton–Leibniz formula for \(0< t\leq 1\), we have
Since \(\alpha _{n}=\gamma _{n-1}\) and \(\gamma _{n-1}<1\), from (8) we get
The last gives (7). The proof of Lemma 1 is complete.
Lemma 2
Let\(n-1\geq k \geq 0\)and\(\gamma _{i}<1\)for\(i=k,k+1,\ldots,n-1\). Let a function\(y\in C^{n}(0,1]\)satisfy the conditions: \(\sup_{0\leq t\leq 1}|D_{\overline{\alpha }}^{n}y(t)|< \infty \)and\(D_{\overline{\alpha }}^{i}y(0)=0\), \(i=k+1,\ldots,n-1\). Then there exist\(D_{\overline{\alpha }}^{k}y(0)\)and
Proof. Since \(\sup_{0\leq t\leq 1}|D_{\overline{\alpha }}^{n}y(t)|< \infty \) and \(\gamma _{i}<1\), \(i=k,k+1,\ldots,n-1\), from (6) we have that the function \(K_{i}(s,t)s^{-\alpha _{n}} D_{\overline{\alpha }}^{n}y(s)\) is absolutely summable on \((0,t)\). Therefore, there exists
for all \(i=k,k+1,\ldots,n-1\).
Since \(D_{\overline{\alpha }}^{n-1}y(0)=0\), from (8) we have
The last, together with (11) for \(i=n-2\), gives
By the condition \(D_{\overline{\alpha }}^{n-2}y(0)=0\), hence
If we continue this process, using the fact that \(D_{\overline{\alpha }}^{i}y(0)=0\), \(i=k+1,\ldots,n-1\), and the finiteness of (11), we get
That gives
i.e., there exists \(D_{\overline{\alpha }}^{k}y(0)\) and (10) holds. From (6) and (12) we have
Thus, we get (9). The proof of Lemma 2 is complete.
Corollary 1
Let\(n-1\geq k \geq 0\). Suppose that the conditions of Lemma 2hold and\(D_{\overline{\alpha }}^{k}y(0)=0\). Then
and
for\(i=k,k+1,\ldots,n-1\).
Lemma 3
Let\(y:I\rightarrow \mathbb{R}\)be such that\(y\in C^{n}(0,1]\). Then
where\(b_{k,k}=1\), \(k=0,1,\ldots,n\), and coefficients\(b_{k,i}\), \(i=1,2,\ldots,k-1\), \(k=1,2,\ldots n\), and\(b_{k,0}\), \(k=1,2,\ldots n\), are defined by the recurrent formulas:
Proof. For \(k=0\), we have
where \(b_{0,0}=1\).
For \(k=1\), we have
That, using \(\alpha _{1}=\gamma _{0}-\gamma _{1}+1\), gives
where \(b_{1,0}=-b_{0,0}(\alpha _{0}+1-1)\) and \(b_{1,1}=1\).
Now, we assume that
is true. Then, using (2), we get
3 Main result
Theorem 1
Let (3) hold. Let the coefficients\(a_{i}(t)\), \(i=0,1,\ldots,n\), of equation (1) be continuous functions on\((0,1]\)and satisfy the conditions
where the constants\(c_{1}>0\)and\(c_{2}>0\)do not depend on\(t\in (0,1]\). Then, for any\(\mu >0\)and for any function\(f(\cdot )\)continuous on\([0,1]\), there exists a unique solution of problem (1) and (4), and the following estimate
holds, where\(c>0\)does not depend onf.
Proof. If we substitute (15) in (1), then we have
By introducing the notations \(\widetilde{a}_{k}(t)=\sum_{i=k}^{n}a_{i}(t) b_{i,k} t^{ \gamma _{k}- \gamma _{0}-\alpha _{0}-i}\), \(k=0,1,\ldots,n\), we get
From conditions (16) and (17) we have
By condition (3) we have that \(\gamma _{i}<1\), \(i=0,1,\ldots,n-1\), therefore (14) and (13) are valid for \(k=0\).
Let \(z(t)=\widetilde{a}_{n}(t) D_{\overline{\alpha }}^{n}y(t)\), then from (4), (13), and (19) we obtain
Now, we prove that integral equation (22) has a unique solution continuous on \([0,1]\), for which the estimate
holds, where \(\overline{c}>0\) does not depend on f.
Despite the fact that integral equation (22) has the form of a Volterra equation, (5) implies that the kernel of the integral operator in (22) is unbounded if, for some \(0\le k\le n-1\), we have \(0<\gamma _{k}<1\). Therefore, we cannot apply the regular theory of Volterra integral equations in the space of continuous functions (see, e.g., [16]). Hence, let us first solve equation (22) on some interval \([0,\delta ]\), \(0<\delta <1\), using the method of contraction mapping (see, e.g., [7]).
By (21) there exists \(1>\delta >0\) such that
In \(C[0,\delta ]\) we consider the integral operator
Due to (6), (21), and (24) we have
Therefore, K is a contraction operator in \(C[0,\delta ]\). Applying the method of contraction mapping to integral equation (22) ([7], pp. 88–89), we have that equation (22) has a unique solution \(\bar{z}_{0}\in C[0,\delta ]\); in addition, \(\max_{0\leq t\leq \delta }|\bar{z}_{0}(t)|\leq \bar{c}_{0} \max_{0\leq t\leq 1}|f(t)|\). The successive approximations \(z_{0}\), \(z_{1}\), … , \(z_{n}\), … to this solution are of the form
where any function from \(C[0,\delta ]\) can be chosen as the first approximation \(z_{0}(t)\).
Thus, we have found the solution \(\bar{z}_{0}\in C[0,\delta ]\) of equation (22) on the interval \([0,\delta ]\). From equation (22) we have
Let us now solve equation (22) on the interval \([\delta ,1]\) with the condition
On \([\delta ,1]\) we present equation (22) in the form
Since \(z=\bar{z}_{0}\) on the interval \([0,\delta ]\), we have
It means that the kernel of the integral operator K is continuous on the bounded set \(\delta \le s\le t\le 1\). Hence, equation (22) is a regular Volterra integral equation on the interval \([\delta ,1]\). Therefore, it has a unique continuous solution \(\bar{z}_{1}\) on the interval \([\delta ,1]\) (see, e.g., [16]), for which the estimate \(\max_{\delta \leq t\leq 1}|\bar{z}_{1}(t)|\leq \bar{c}_{1} \max_{0\leq t\leq 1}|f(t)|\) holds.
From (26) we have
The last gives that \(\bar{z}_{0}(\delta )=\bar{z}_{1}(\delta )\), i.e., (25) holds. Hence, the function
belongs to \(C[0,1]\) and it is a unique solution of (22). From the estimates for \(\bar{z}_{0}\) and \(\bar{z}_{1}\), we get (23) with \(\bar{c}=\max \{\bar{c}_{0},\bar{c}_{1}\}\).
Thus, problem (1) and (4) is reduced to the problem
with condition (3), where \(z\in C[0,1]\) is a unique solution of (22).
Since \(z\in C[0,1]\), the function \(\widetilde{a}_{n}^{-1}(t)= (a_{n}(t)t^{-\mu })^{-1}\) is continuous on \((0,1]\) and by (20) we have that \(\frac{1}{c_{2}}\leq \widetilde{a}_{n}^{-1}(t)\leq \frac{1}{c_{1}}\), \(t\in (0,1]\). Then from (27) we get \(D_{\overline{\alpha }}^{n}y(t)\in C(0,1]\) and \(\sup_{0\leq t \leq 1}|D_{\overline{\alpha }}^{n}y(t)|< \infty \). Moreover, by condition (3) we have \(\gamma _{i}<1\), \(i=0,1,\ldots,n-1\). Then, on the basis of Lemmas 1 and 2 and from (13) and (14), we get
and the estimate
From (29) the uniqueness of problem (27) follows.
Since \(\mu =\gamma _{0}+\alpha _{0}+n-1\), from (15) we have
The last, together with (30), gives the estimate
where \(\tilde{c}>0\) depends on c̄, \(c_{1}\), \(d_{i}\) and \(b_{k,i}\), \(k,i=0,1,\ldots,n\), and does not depend on f. The last implies the validity of (18). The proof of Theorem 1 is complete.
Remark 1
Under the conditions of Theorem 1 problem (1) and (4) is solvable for any set \(\alpha _{0}\), \(\alpha _{1}\), …, \(\alpha _{n}\) satisfying condition (3).
References
Abdikalikova, Z., Baiarystanov, A., Oinarov, R.: Compactness of embedding between spaces with multiweighted derivatives—the case \(p\leq q\). Math. Inequal. Appl. 14(4), 793–810 (2011)
Abdikalikova, Z., Kalybay, A.: Summability of a Tchebysheff system of functions. J. Funct. Spaces Appl. 8(1), 87–102 (2010). https://doi.org/10.1155/2010/405313
Abdikalikova, Z., Oinarov, R., Persson, L.-E.: Boundedness and compactness of the embedding between spaces with multiweighted derivatives when \(1\leq q < p <\infty \). Czechoslov. Math. J. 61 7 (2011). https://doi.org/10.1007/s10587-011-0014-1
Baideldinov, B.L.: On stabilization to the generalized polynomial of functions with an integrable weight derivative. Dokl. Math. 354(3), 295–297 (1997)
Baideldinov, B.L.: On equivalent norms in multiweighted spaces. Izv. Nats. Acad. Nauk Resp. Kaz. Ser. Fiz.-Mat. 3, 8–14 (1998) (in Russian)
Kalybay, A.: A generalized multiparameter weighted Nikol’skii–Lizorkin inequality. Dokl. Math. 68(1), 121–127 (2003)
Kolmogorov, A.N., Fomin, S.V.: Elements of the Theory of Functions and Functional Analysis. Nauka, Moskow (1989) (in Russian)
Kudryavtsev, L.D.: Direct and inverse embedding theorems. Applications to the solution of elliptic equations by variational method. Trudy Mat. Inst. Akad. Nauk SSSR 55, 3–182 (1959) (in Russian)
Kudryavtsev, L.D.: Imbedding theorems for functions defined on unbounded regions. Sov. Math. Dokl. 4, 530–532 (1963) English transl. in
Kudryavtsev, L.D.: An imbedding theorem for a class of functions defined in the whole space or in the half-space. I. Mat. Sb. 69(111)(4), 616–639 (1966) (in Russian)
Kudryavtsev, L.D.: Imbedding theorems for classes of functions defined on the entire space or on a half space. II. Mat. Sb. 70(112)(1), 3–35 (1966) (in Russian)
Kudryavtsev, L.D.: Equivalent norms in weighted spaces. Proc. Steklov Inst. Math. 170, 185–218 (1987)
Kudryavtsev, L.D.: On variational problems for quadratic weight functionals on infinite intervals. Proc. Steklov Inst. Math. 172, 223–235 (1987)
Kudryavtsev, L.D.: An analogue for an infinite interval of the Lizorkin–Nikol’skii inequality. Proc. Steklov Inst. Math. 173, 151–160 (1987)
Kudryavtsev, L.D.: On some inequalities for functions and their derivatives on an infinite interval. Proc. Steklov Inst. Math. 204, 143–147 (1994)
Smirnov, V.I.: A Course of Higher Mathematics. Nauka, Moskow (1974) (in Russian)
Acknowledgements
The author would like to thank Professor Ryskul Oinarov and the unknown referees for their generous suggestions and remarks, which have improved this paper.
Availability of data and materials
Not applicable.
Funding
The paper was written with financial support of the Ministry of Education and Science of the Republic of Kazakhstan, grant no. AP05130975 in the area “Scientific research in the field of natural sciences”.
Author information
Authors and Affiliations
Contributions
All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The author declares that she has no competing interests.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Kalybay, A. Boundary value conditions for linear differential equations with power degenerations. Bound Value Probl 2020, 110 (2020). https://doi.org/10.1186/s13661-020-01412-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13661-020-01412-6