Brought to you by:
Paper

Rational solutions of a Riccati equation

and

© 2016 RAS(DoM) and LMS
, , Citation V. V. Sokolov and A. B. Shabat 2016 Russ. Math. Surv. 71 787 DOI 10.1070/RM9728

0036-0279/71/4/787

Export citation and abstract BibTeX RIS

We consider the problem of solutions of the equation

Equation (1)

that are rational in the variable $x$. It is easy to see that such a solution must have the form

Equation (2)

Proposition 1.  A function (2) solves (1) if and only if

Equation (3)

and

Equation (4)

The coefficients $a_i$ in (1) are defined up to the scaling $x\mapsto \mu x$, $f\mapsto \mu^{-1} f$. For definiteness, in what follows we will assume that $a_2=1$, so that $k_1^2=1$. It is easy to see that, as $x\to \infty$, a rational solution of (1) has an expansion

Equation (5)

where $d=4a_0-a_1^2$. The properties of formal solutions of the form (5) and questions related to their convergence were considered in [1].

Example 1.  Assume that the potential in the Riccati equation is an even function (so that $a_1=a_{-1}=0$) and a rational solution is an odd function. Then a solution of the form

exists if and only if the Riccati equation has the form

Equation (6)

and the polynomial $P$ solves the linear equation $xP''+2(x^2+\lambda)P'-4nxP=0$. From it we find that $a_k=C^k_n\displaystyle\prod_{i=k+1}^n\biggl(\lambda+i-\dfrac{1}{2}\biggr)$ and $a_n=1$, where $\lambda$ is an arbitrary parameter.

For each positive integer $m$ the equation

has the solution

Here $\lambda$ is an arbitrary parameter. If $\lambda=-(2n+2m+1)/2$, then (6) has two rational solutions, $f^{+}$ and $f^{-}$.

Theorem.  Assume that (1) with $a_2\ne 0$ has two rational solutions,

Equation (7)

Then

Corollary.  Equation (1) with $a_2\ne 0$ has at most two rational solutions.

Taking $a_2=1$, we can assume without loss of generality that $k_1=1$ and ${\overline k_1=-1}$. Then for fixed $m$ and $\overline m$ and for the $a_{-2}$, $k_1$, $\overline k_1$, $k_{-1}$, and $\overline k_{-1}$ defined above, the relations (3) and (4) take us to a system of $m+\overline m+4$ equations with respect to the $m+\overline m+4$ unknowns $k_0$, $a_1$, $a_0$, $a_{-1}$, $y_i$, and $\overline y_i$. This system can easily be solved for small $m$ and $\overline m$. In particular, the case when $m=\overline m=0$ is described in Example 1.

Example 2.  For $m=1$ and $\overline m=0$ we obtain the Riccati equation

which has the two rational solutions

It is clear that a rational solution $f$ of a Riccati equation defines the solution $\psi=\exp\displaystyle\int f\,dx$ of the corresponding linear Schrödinger equation.

An algorithm for finding elementary solutions of a second-order linear equation with rational coefficients was put forward in [2]. In our case this algorithm is based on the observation that (for fixed numerical values of the coefficients $a_i$ in (1)) the formulae (3) determine at most four possible values of $m$, each of which must be investigated separately. The two rational solutions $f^{+}$ and $f^{-}$ yield [3] the solution $\phi=(f^{+}-f^{-})^{-1}$ of a third-order linear equation which is well known in soliton theory.

The authors acknowledge useful discussions with V. G. Marikhin and are grateful to the participants of the seminar on mathematical physics at the Landau Institute for Theoretical Physics for their interest in this work.

Please wait… references are loading.
10.1070/RM9728