1: Using the terminology of \cite{Ronveaux}, we consider three categories of
2: solutions to Heun's equation: {\sl local solutions}, {\sl Heun functions} and
3: {\sl Heun polynomials}.
4:
5: {\bf Local solutions} to Heun's equation are valid in the
6: neighbourhood of one singularity and are associated with one of the two
7: exponents there. Since there are four singularities, there are a total of eight
8: such solutions. {\bf Heun functions} are solutions valid in an region containing two
9: adjacent singularities. {\bf Heun polynomials} are solutions to Heun's equation
10: valid at three singularities. Despite the name, not all such solutions are
11: polynomials, but are of the simple form
12: %
13: \begin{equation}\label{eq:HP}
14: H\!p\,(z) = z^{\sigma_1}(z-1)^{\sigma_2}(z-a)^{\sigma_3}p_n(z),
15: \end{equation}
16: %
17: where $p_n(z)$ is a polynomial of degree $n$ and $\sigma_{1,2,3}$ is one of the
18: exponents associated with the singularity at $z=0,1,a$ respectively.
19:
20: In order to determine which type of solution we need to look for, we have to
21: determine the domain of relevance for our particular Heun equation given by
22: \eref{eq:Heun}. The parameter $A$ and the quantities that depend on it are
23: defined on the intervals
24: %
25: \[ A\in (1,3), \qquad a \in (0,-1), \qquad q\in (0,-3/2). \]
26: %
27: The limit of inifinite central pressure is given by $A\to 1$ and the Newtonian
28: limit by $A \to 3$. We exclude these limits in this analysis (as indicated by
29: the open intervals), but note that Chandrasekhar and Miller include an interesting
30: discussion of the limit of infinite central pressure in \cite{CM74} and believe
31: that consideration of such limits in the context of the confluent Heun equation
32: could lead to interesting results.
33:
34: The interior of the star, i.e.\ the region of validity of \eref{eq:Heun},
35: is given by the interval
36: %
37: \begin{equation}\label{z-interval}
38: B\in [1,A/3] \Longrightarrow z\in \left[0,\frac{3-A}{6}\right]=
39: \left[0,\frac{a+1}{3}\right],
40: \end{equation}
41: %
42: where $z=0$ represents the centre of the star and $z=(3-A)/6$ its surface.
43:
44: Since $a$ is always strictly negative whereas $z$ is always real and positive,
45: we are interested in a solution including the singularity at zero and extending along
46: the real axis toward the singularity at $z=1$ (though not reaching it). The function
47: $\tilde\omega$ must remain finite (and non-zero) at the point $z=0$, whence
48: we require at the very least the local solution there corresponding to the exponent
49: $\sigma_1=0$.
50:
51: The remainder of this section is devoted to a survey of attempts that
52: can be made to find a closed form solution to our Heun equation
53: satisfying these constraints. Although none of these attempts proved
54: successful, the following subsections are intended to stimulate new
55: ideas in this direction, explain where problems lie in what may seem
56: promising solution strategies and help the reader avoid choosing
57: fruitless paths of thought.
58:
59: \subsection{Heun Polynomials}
60:
61: The most tractable solution to Heun's equation is one
62: of the Heun polynomials, so that we begin by searching for such a solution even
63: though the validity at the two `additional' singularities is not necessary
64: for our purposes. As was mentioned in footnote~\ref{uniqueness}, the particular
65: form of the Heun equation given in \eref{eq:Heun} is not unique since various
66: transformations map a Heun equation onto a new Heun equation with other parameters
67: and exponents. Indeed there are 192 mappings of Heun's equation onto itself, corresponding to the 192 local solutions that can be provided for it. These are made up of 24 transformations of the independent variable, which map the four singularities onto themselves, and 8 elementary power transformations of the dependent variable. A chapter on these transformations is contained in \cite{Ronveaux} and a discussion of their group structure as well as a useful table containing all of them can be found in \cite{Maier04}. These transformations will not occupy us further in this paper, but the statements made and conclusions drawn are valid (modulo obvious modifications resulting from the transformation) for all mappings of \eref{eq:Heun} onto itself.
68:
69: The exponents $\sigma_i$ (see \eref{eq:HP}) of the four Heun polynomials
70: consistent with our physical problem as well as the parameters $\alpha$ and $\beta$
71: for the solutions are listed in Table~\ref{tab:HP}.
72: %
73: \begin{table}
74: \caption{The values of the exponents $\sigma_i$ of \eref{eq:HP} and the
75: parameters $\alpha$ and $\beta$ corresponding to the four classes of Heun
76: polymials that would yield permissible solutions to our physical problem. The
77: numbering of the classes follows the conventions in \cite{Ronveaux}.
78: \label{tab:HP}}
79: \begin{indented}
80: \item[]\begin{tabular}{cccccc}\br
81: Class & $\sigma_1$ & $\sigma_2$ & $\sigma_3$ & $\alpha$ & $\beta$ \\ \mr
82: %
83: I & 0 & 0 & 0 & $-n$ & $\gamma+\delta+\epsilon+n-1$ \\
84: III & 0 & $1-\delta$ & 0 & $\delta-n-1$ & $\gamma+\epsilon+n$ \\
85: V & 0 & 0 & $1-\epsilon$ & $\epsilon-n-1$ & $\gamma+\delta+n$ \\
86: VII & 0 & $1-\delta$ & $1-\epsilon$ & $\delta+\epsilon-n-2$ & $\gamma+n+1$ \\ \br
87: \end{tabular}
88: \end{indented}
89: \end{table}
90: %
91: A comparison with \eref{eq:Heun} shows us that only a Heun polynomial of
92: class~I is consistent with the given parameters since $n$, the degree of the
93: polynomial $p_n(z)$ in \eref{eq:HP}, can of course only be a non-negative
94: integer. The set of equations that must be solved in general in order to determine
95: $p_n(z)$ provides a polynomial equation of degree $n+1$ for the accessory parameter
96: $q$. In our case, the equation is simply $q=0$, which lies just outside the allowed
97: range for $q$ and thus there exists no Heun polynomial, which is a solution of the
98: physical problem being considered.
99:
100: \subsection{Heun Functions}
101:
102: Since we require a solution valid at $z=0$ and extending along the real axis
103: toward $z=1$, a Heun Function seems to be the appropriate solution. It is
104: related to the two-point connection problem treated in \cite{SS80} and
105: discussed further in \cite{SL00}. The search for such solutions generally
106: implies solving an eigenvalue problem and is valid only for a restricted set
107: of values for the accessory parameter $q$. In our case, $q$ is not independent
108: of the singularity $a$, and no such solutions exist.
109:
110: \subsection{Rational and other Transformations}
111:
112: The next step we take in looking for a solution of our Heun equation is to
113: consider a broader set of transformations. Since Heun's equations are closely
114: related to the hypergeometric equations, one can hope to find a transformation
115: relating the two equations. Kuiken \cite{Kuiken79} addressed the question as to
116: when a hypergeometric equation can be transformed into a Heun equation via a
117: rational substitution of the independent variable (excluding the trivial case
118: $\alpha \beta = q =0$). Her work was reexamined and completed by Maier
119: \cite{Maier02}. Using their results, we have found that no rational transformation of the independent
120: variable maps our Heun equation (with the specified parameter range) onto a
121: hypergeometric equation%
122: %
123: \footnote{Many of the computations in this paper made use of the computer algebra
124: program Maple\texttrademark. Maple is a trademark of Waterloo Maple Inc.}.
125: %
126: Even had such a transformation been found, it would only have been valid for
127: specific values (or one specific value) of the parameter $A$, because of the
128: nature of the necessary conditions for such a transformation to exist.
129:
130: Having ruled out rational transformations, we turn our attention briefly to
131: integral transformations derived by Carlitz and Valent and described in
132: \cite{Valent86}, which impose no restriction on the accessory parameter. Such
133: transformations have proved useful in finding previously unknown, closed form
134: solutions to Heun's equation. In our case, these methods only allow for marginal
135: progress. Consider first a transformation of the independent variable
136: \[ B' = \frac{B-a}{B(1-a)}.\]
137: Applying to this the transformation (9) of \cite{Valent86}, and denoting the new
138: quantities by a douple prime, one finds that
139: \[ \gamma''=\varepsilon''=\frac{1}{2} \qquad {\rm and} \qquad
140: \alpha''+\beta''=\delta''\]
141: and can apply the quadratic transformation discussed on pg.~59 of \cite{Erdelyi3}.
142: Here, however, one is not guaranteed that the solution will be regular at the centre
143: of the star.
144:
145: More modern ideas, such as relating the Heun equation to the Schr\"odinger equation
146: through the ``generalized associated Lam\'e" (GAL) potential seem
147: tailor made to the situation being considered here. Indeed, the analogue of (32)
148: in \cite{KS05} with $b=3/2$ can be applied to \eref{eq:Heun}, but the restriction
149: to the accessory parameter is $q=0$ as it was in the case of a Heun polynomial.
150:
151: We know in fact that no meromorphic solution to our Heun equation exists (for arbitrary $q$)
152: since the necessary condition $\epsilon=1/2-m$, $m \in \mathbb{Z}$ \cite{Valent05} is not met.
153: