Abstract
We study the existence of exact solutions either for linear systems of ODEs or for scalar second order linear differential equations with Dirichlet boundary value conditions.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Solutions of some differential equations can be exactly found using explicit formulas what is a remarkable feature. A class of such differential equations is studied in [6]. In this paper, we present a wider class of ODEs that are solvable exactly, i.e. its solutions can be found by using formulas. Such formulas are presented in Sects. 2 and 3 of this paper. We emphasize that our study of such differential equations is motivated also by series of papers [1,2,3,4,5] which are important for understanding significant atmospheric flows like sea breezes and atmospheric undular bores, as discussed in these papers. We also present related results for systems of linear ODEs in Sect. 4. Related results are presented in [7, 8] but it appears that our achievements are not contained in those comprehensive books.
2 Second order scalar ODEs
Motivated by [1,2,3,4,5,6,7,8], we consider a linear second order differential equation with a Dirichlet boundary condition
where \(p\in C^2(I,\mathbb {R})\), \(q,h\in C(I,\mathbb {R})\) are functions and \('=\frac{d}{ds}\). For solving (1), we need to solve
We set
in (2) for \(f,r\in C^2(I,\mathbb {R})\) to get
Our goal to write (4) as the following equation with constant coefficients
for some \(\mu ,\nu \in \mathbb {R}\) and \(z=f(s)\). This means we have to solve
From the 1st equation of (6), we get
by assuming
By inserting (7) into the 2nd equation of (6), we obtain
Solving (9), we have
Then integrating (10), we derive
Plugging (11) into (7), we get
Finally, putting (7) and (11) into the 3rd equation of (6), we have
From (13), we get
For simplifying the above formulas, we take
for \(g(s)>0\) on \(s\in I\) to obtain
Since \(f'(s)=\displaystyle \left( \frac{g'}{g}\right) (s)\), so (8) reads
Summarizing we arrive at the following result.
Theorem 2.1
Let \(g\in C^1(I,\mathbb {R})\) satisfy (16). If U(z) is a solution of (5), then u(s) given by (3) is a solution of (2) with the corresponding functions of (15).
Of course, solutions of (5) are well known formulae. As a simple example, we take
Then (15) gives
We consider g(s) in (15) as a parameter, so we have a rich family of ODEs in Theorem 2.1.
3 The non homogeneous problem
In this application section, we will use the method of Green’s functions.
Theorem 3.1
Let \(g\in C^1(I,\mathbb {R})\) satisfy (16), \( h\in C(I,\mathbb {R}) \) and \( \mu ,\nu \in \mathbb {R} \) be given constants. Consider the non homogeneous problem (1) with the corresponding functions of (15) and let \( \lambda _{1,2} \) be the roots of the equation \( \lambda ^{2}+\mu \lambda +\nu =0 \). Then there holds:
-
1.
Assume either \( \mu ^{2}-4\nu \ge 0 \) or
$$\begin{aligned} g(1)\ne g(0)\textrm{e}^{\frac{k\pi }{\beta }} \text{ for } \text{ every } k\in \mathbb {N}. \end{aligned}$$(18)Then the solution of (1) is given by
$$\begin{aligned} u(t)=\frac{u_{1}(t)}{N}\int _{0}^{t}u_{0}h(s)\;\textrm{d}s+\frac{u_{0}(t)}{N}\int _{t}^{1}u_{1}h(s)\;\textrm{d}s. \end{aligned}$$(19)The functions \( u_{0,1} \) and \( N\in \mathbb {R} \) are specified in the following:
-
1.
if \( \mu ^{2}-4\nu >0 \) and \( \lambda _{1,2} \) are the two distinct real roots then
$$\begin{aligned} u_{j}(s)=r(g^{\lambda _{2}}(j)g^{\lambda _{1}}(s)-g^{\lambda _{1}}(j)g^{\lambda _{2}}(s)),\;j=0,1, \end{aligned}$$and \( N=(g^{\lambda _{2}}(0)g^{\lambda _{1}}(1)-g^{\lambda _{1}}(0)g^{\lambda _{2}}(1))(\lambda _{1}-\lambda _{2}) \) is nonzero,
-
2.
if \( \mu ^{2}-4\nu <0 \), the assumption (18) is valid, and \( \lambda _{1}=\alpha +\beta i \), \( \lambda _{2}=\alpha -\beta i \) for \( \beta \ne 0 \) then
$$\begin{aligned} u_{j}(s)=rg^{\alpha }(s)\sin \left( \beta \,\textrm{ln}\,\frac{g(s)}{g(j)}\right) ,\;j=0,1, \end{aligned}$$and \( N=\beta \sin \left( \beta \,\textrm{ln}\,\frac{g(1)}{g(0)}\right) \) is nonzero,
-
3.
if \( \mu ^{2}-4\nu =0 \) and \( \lambda =\lambda _{1}=\lambda _{2} \) is the multiple root then
$$\begin{aligned} u_{j}(s)=rg^{\lambda }(s)(\textrm{ln}\,g(j)-\textrm{ln}\,g(s)),\;j=0,1, \end{aligned}$$and \( N=\textrm{ln}\frac{g(1)}{g(0)} \) is nonzero.
-
1.
-
4.
Let \( \mu ^{2}-4\nu <0 \) and \( \lambda _{1}=\alpha +\beta i \), \( \lambda _{2}=\alpha -\beta i \) for \( \beta \ne 0 \). Assume that (18) is not true. Denote \( u_{0}(s)=rg^{\alpha }(s)\sin \left( \beta \,\textrm{ln}\,\frac{g(s)}{g(0)}\right) \) and \( u_{2}(s)=rg^{\alpha }(s)\cos \left( \beta \,\textrm{ln}\,\frac{g(s)}{g(0)}\right) \). Then for every \( h\in C(I,\mathbb {R}) \) such that
$$\begin{aligned} \int _{0}^{1}u_{0}h(s)\;\textrm{d}s=0, \end{aligned}$$(20)the function
$$\begin{aligned} \nonumber u(t)=- & {} \frac{u_{0}(t)}{\beta }\int _{t}^{1}u_{2}h(s)\;\textrm{d}s -\frac{u_{2}(t)}{\beta }\int _{0}^{t}u_{0}h(s)\;\textrm{d}s\\+ & {} \frac{u_{0}(t)}{\beta ||u_{0}||_{2}^{2}} \left( \int _{0}^{1}u_{2}h(s)\int _{0}^{s}u_{0}^{2}(\tau )\;\textrm{d}\tau \;\textrm{d}s +\int _{0}^{1}u_{0}h(s)\int _{s}^{1}u_{0}u_{2}(\tau )\;\textrm{d}\tau \;\textrm{d}s \right) \end{aligned}$$is the unique solution of the problem (1) such that \( \int _{0}^{1}u_{0} u(t)\;\textrm{d}t=0. \)
Proof
Let us prove the case 1. We solve the problem (1) for a given \(h\in C(I,\mathbb {R})\) where the functions r, p, q are given in (15). Our goal is to find a Green’s function for (1) of the form
The functions \( u_{0} \) and \( u_{1} \) in (21) are nontrivial solutions of the homogeneous equation (2) such that it holds \( u_{0}(0)=0 \) and \( u_{1}(1)=0 \). This assures that G satisfies the boundary conditions in (1). The functions A and B are unknown and can be determined using the basic properties of Green functions: the continuity of G at the diagonal of \( I\times I \) and the jump discontinuity of the derivative \( G^{\prime }_{t} \) at the same diagonal.
Let \( \mu ,\nu \in \mathbb {R} \) are given. The form of the general solution of (5) depends on the choice of these constants. Due to this, we divide the case 1 into three subcases.
Case 1a. Let \( \mu ^{2}-4\nu >0 \). In this case, it holds that \( U(z)=c\textrm{e}^{\lambda _{1}z}+d\textrm{e}^{\lambda _{2}z} \) for some constants \( c,d\in \mathbb {R} \). From (3) and our choice of f, we deduce that \( u(s)=r(s)(cg(s)^{\lambda _{1}}+dg(s)^{\lambda _{2}}) \) and the functions \( u_{0,1} \) are particular solutions that satisfy \( u_{j}(j)=0 \) for \( j=0,1 \). \( u_{0,1} \) are necessary for finding the Green’s function of the form (21). Due to the assumption (16), it holds
or equivalently,
This means that the functions \( u_{0} \) and \( u_{1} \) are linearly independent and the Green function defined by (21) exists. Based on the properties of the Green function, it holds \( Au_{0}(s)=Bu_{1}(s) \) and \( Bu^{\prime }_{1}(s)-Au^{\prime }_{0}(s)=\frac{1}{p(s)} \) for \( s\in I \). Thus A and B are the solutions of the system
Observe that \( pr^{2}g^{\prime }g^{\lambda _{1}+\lambda _{2}-1}(s)=1 \). By a straightforward calculation, one can prove that
is nonzero due to (22). We apply the Cramer’s rule we deduce that \( A=\frac{u_{1}}{N} \) and \( B=\frac{u_{0}}{N} \). The function
is the desired solution of the problem (1).
Case 1b. Let \( \mu ^{2}-4\nu <0 \) and (18) be true. The corresponding real-valued solution is \( U(z)=\textrm{e}^{\alpha z}(c\sin \beta z+d\cos \beta z) \) where \( c,d\in \mathbb {R} \). Hence \( u(s)=r(s)g^{\alpha }(s)(c\sin (\beta \,\textrm{ln}\,g(s))+d\cos (\beta \,\textrm{ln}\,g(s))) \) and the functions \( u_{0,1} \) are particular solutions such that \( u_{j}(j)=0 \) for \( j=0,1 \).
The assumption (18) implies that the functions \( u_{0} \) and \( u_{1} \) are linearly independent. Note that since g is increasing it suffices to take \( k\in \mathbb {N} \) instead of \( k\in \mathbb {Z} \). Similarly as in the case 1, we come to the system (23) and the determinant
is nonzero due to (18). Hence \( A=\frac{u_{1}}{N} \), \( B=\frac{u_{0}}{N} \) and the solution of the problem (1) can be expressed by (19).
Case 1c. - \( \mu ^{2}-4\nu =0 \). In this case, the equation \( \lambda ^{2}+\mu \lambda +\nu =0 \) has one multiple root \( \lambda \in \mathbb {R} \). The corresponding solution is \( U(z)=c\textrm{e}^{\lambda z}+dz\textrm{e}^{\lambda z}\) where \( c,d\in \mathbb {R} \). Hence \( u(s)=rg^{\lambda }(s)(c+d\,\textrm{ln}\,g(s)) \) and again, \( u_{j}(j)=0 \) for \( j=0,1 \).
Again, we come to the system (23) and the determinant
is nonzero. Hence \( A=\frac{u_{1}}{N} \), \( B=\frac{u_{0}}{N} \) and the solution of the problem (1) can be expressed by (19).
Now, we prove the case 2. In this case, zero is an eigenvalue of the operator
defined on domain \( \mathcal {D}(L)=\{u\in C^{2};\;u(0)=u(1)=0\} \). The space \( L^{2}(0,1) \) is equipped with the standard integral norm \( ||u||_{2}:=\sqrt{\int _{0}^{1}u^{2}(s)\;\textrm{d}s} \). The function \( u_{0} \) is an eigenfunction of L corresponding to the zero eigenvalue. It is known that there exists a solution of the non homogeneous problem (2) if and only if (20) is valid; see e.g. [9]. Such solution is not even unique.
If (18) is not true then the Green’s function of the form (21) does not exist. However, it is still possible to find a Green’s function in a different form and to obtain the solution u of the problem (2). Let \( h\in C(I,\mathbb {R}) \) be such that (20) holds.
Note that \( u_{2} \) is a solution of (2) and \( u_{0},u_{2} \) are linearly independent. We take the following Green’s function
Clearly, \( G(s,0)=0 \) for \( s\in I \). Since we require that \( \int _{0}^{1}u_{0} u(s)\;\textrm{d}s=0 \) we find the functions A, B, C so that
This implies that \( u(t)=\int _{0}^{1}G(s,t)h(s)\;\textrm{d}s \) is the desired solution. Indeed, it holds
Using (25) and (24), we come to the equation
This along with the equations derived from properties of Green functions, namely the continuity of G and the jump discontinuity of the derivative \( G^{\prime }_{t} \), lead to the system
A simple computation shows that \( u_{0}u^{\prime }_{2}-u_{2}u^{\prime }_{0}=-\beta r^{2}g^{\prime }g^{2\alpha -1} \) and
is obviously nonzero. Thus
and the corresponding solution
satisfies the condition \( u(0)=0 \) and \( \int _{0}^{1}u_{0} u(s)\;\textrm{d}s=0 \). Moreover, due to (20), there holds \( u(1)=0 \). Rearranging the terms in the last expression, we obtain
and the assertion follows. \(\square \)
4 First order systems of ODEs
In this section, we find exact fundamental matrices of certain linear ODEs. Let M(n) be a linear space of \(n\times n\) matrices, and \(I\subset \mathbb {R}\) be an open interval.
Lemma 4.1
If \(G: I\rightarrow M(n)\) has a derivative \(G'(z_0)\) at \(z_0\in I\) then also \(H: I\rightarrow M(n)\) given as \(H(z)=e^{G(z)}\) has a derivative at \(z_0\). If in addition
then
Proof
The mapping \(e: M(n)\rightarrow M(n)\) given by \(e^K=\sum _{i=0}^\infty \frac{K^i}{i!}\) is \(C^\infty \)-smooth and a straightforward computation of the derivative \(De^KS\) of \(e^K\) at \(S\in M(n)\) leads to
whenever \(KS=SK\). Since \(H(z)=e^{G(z)}\), the chain rule gives
The proof is completed. \(\square \)
Assume \(0\in I\). If \(A\in L^1_{loc}(I,M(n))\) then it is well-known that \(G(z)=\int _0^zA(s)ds\) has a derivative for almost each (f.a.e.) \(z\in I\) with \(G'(z)=A(z)\). Then applying Lemma 4.2, we obtain
Lemma 4.2
If \(A\in L^1_{loc}(I,M(n))\) then
has a derivative \(X'(z)\) f.a.e. \(z\in I\). If in addition
then
Proof
(29) implies
This verifies (26) and proof is finished. \(\square \)
Lemma 4.2 shows that (28) is the fundamental solution of (30). If in addition, A(z) is T-periodic in Lemma 4.2, i.e., \(A(z+T)=A(z)\) for any \(z\in I=\mathbb {R}\), then (28) has a form
with \(P(z+T)=P(z)\) for any \(z\in \mathbb {R}\), which is the Floquet’s theorem.
Assumption (29) seems to be very special, but there is the following family of such A(z):
for \(1\le k\le n^2\), \(a_i\in L^1_{loc}(I,\mathbb {R})\), \(A_i\in M(n)\) which are commutative
Indeed, we have
so (29) is verified. Then (28) has a form
If in addition, it holds
then we derive
We note that any \(B\in M(n)\), which is not a multiple of I, generates such a 3-parametric family by
for any \(C\in Com(B)\setminus [I,B]\), where
Formula (32) can be extended to
under assumption (33) along with
As a special case is
under assumption
that is
Another family is presented in the quaternion field \(\mathbb {H}\) given by
for \(p_i,q_i\in L^\infty (I,\mathbb {C})\) satisfying
Remark 4.3
The above results can be directly extended either to Banach spaces or to Banach algebras.
Data availability statement
No data.
References
Constantin, A., Johnson, R.S.: Atmospheric undular bores. Math. Ann. 388, 4011–4036 (2024)
Constantin, A., Johnson, R.S.: On the propagation of nonlinear waves in the atmosphere. Proc. A. 478(2260), 20210895 (2022)
Constantin, A., Johnson, R.S.: On the propagation of waves in the atmosphere. Proc. A. 477(2250), 20200424 (2021)
Johnson, R.S.: On the mathematical fluid dynamics of atmospheric gravity (buoyancy) waves. Monatsh. Math. 201, 1125–1147 (2023)
Johnson, R.S.: An Introduction to the Mathematical Fluid Dynamics of Oceanic and Atmospheric Flows, ESI Lectures in Mathematics and Physics. EMS Press, Berlin (2023)
Marynets, K.: Sturm-Liouville boundary value problem for a sea-breeze flow. J. Math. Fluid Mech. 25, 6 (2023)
Neuman, F.: Global Properties of Linear Ordinary Differential Equations, Mathematics and its Applications 52. Springer, Dordrecht (2013)
Polyanin, Andrei D., Zaitsev, Valentin F.: Handbook of Exact Solutions for Ordinary Differential Equations, 2nd edn. Chapman & Hall/CRC, Boca Raton, FL (2003)
Roach, G.F.: Green’s Functions. Cambridge University Press, London (2013)
Acknowledgements
We are grateful to the reviewer for valuable comments which improved our paper.
Funding
Open access funding provided by The Ministry of Education, Science, Research and Sport of the Slovak Republic in cooperation with Centre for Scientific and Technical Information of the Slovak Republic.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Communicated by Adrian Constantin.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work is partially supported by the National Natural Science Foundation of China (12371163), and the Slovak Grant Agency VEGA Nos. 1/0084/23 and 2/0062/24.
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
Fečkan, M., Pačuta, J. & Wang, J. Exact solvability of certain linear ODEs. Monatsh Math (2024). https://doi.org/10.1007/s00605-024-01992-w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00605-024-01992-w