Digital Library of Mathematical Functions
About the Project
NIST
1 Algebraic and Analytic MethodsAreas

§1.12 Continued Fractions

Contents

§1.12(i) Notation

The notation used throughout the DLMF for the continued fraction

1.12.1\cfracstyle{d}b_{0}+\cfrac{a_{1}}{b_{1}+\cfrac{a_{2}}{b_{2}+\raisebox{-6.0pt}{%
$\ddots$}}}

is

1.12.2b_{0}+\cfrac{a_{1}}{b_{1}+\cfrac{a_{2}}{b_{2}+}}\cdots.

§1.12(ii) Convergents

1.12.3C=b_{0}+\cfrac{a_{1}}{b_{1}+\cfrac{a_{2}}{b_{2}+\cdots}},a_{n}\not=0,
1.12.4C_{n}=b_{0}+\cfrac{a_{1}}{b_{1}+\cfrac{a_{2}}{b_{2}+\cdots\cfrac{a_{n}}{b_{n}}%
}}=\frac{A_{n}}{B_{n}}.

C_{n} is called the nth approximant or convergent to C. A_{n} and B_{n} are called the nth (canonical) numerator and denominator respectively.

Recurrence Relations

1.12.5
A_{k}=b_{k}A_{{k-1}}+a_{k}A_{{k-2}},
B_{k}=b_{k}B_{{k-1}}+a_{k}B_{{k-2}}, k=1,2,3,\dots,
1.12.6
A_{{-1}}=1,
A_{0}=b_{0},
B_{{-1}}=0,
B_{0}=1.

Equivalence

Two continued fractions are equivalent if they have the same convergents.

b_{0}+\displaystyle{\cfrac{a_{1}}{b_{1}+\cfrac{a_{2}}{b_{2}+\cdots}}} is equivalent to b^{{\prime}}_{0}+\displaystyle{\cfrac{a^{{\prime}}_{1}}{b^{{\prime}}_{1}+%
\cfrac{a^{{\prime}}_{2}}{b^{{\prime}}_{2}+\cdots}}} if there is a sequence \{d_{n}\}^{\infty}_{{n=0}}, d_{0}=1,
d_{n}\neq 0, such that

1.12.15a^{{\prime}}_{n}=d_{n}d_{{n-1}}a_{n},n=1,2,3,\dots,

and

1.12.16b^{{\prime}}_{n}=d_{n}b_{n},n=0,1,2,\dots.

Formally,

1.12.17b_{0}+\cfrac{a_{1}}{b_{1}+\cfrac{a_{2}}{b_{2}+\cfrac{a_{3}}{b_{3}+\cdots}}}={b%
_{0}+\cfrac{a_{1}/b_{1}}{1+\cfrac{a_{2}/(b_{1}b_{2})}{1+\cfrac{a_{3}/(b_{2}b_{%
3})}{1+\cdots\cfrac{a_{n}/(b_{{n-1}}b_{n})}{1+\cdots}}}}}={b_{0}+\cfrac{1}{(%
\ifrac{1}{a_{1}})b_{1}+\cfrac{1}{(\ifrac{a_{1}}{a_{2}})b_{2}+\cfrac{1}{(\ifrac%
{a_{2}}{(a_{1}a_{3})})b_{3}+\cfrac{1}{(\ifrac{a_{1}a_{3}}{(a_{2}a_{4})})b_{4}+%
\cdots}}}}}.

Series

when p_{k}\not=0, k=1,2,3,\dots.

1.12.19\sum^{n}_{{k=0}}c_{k}x^{k}=c_{0}+\cfrac{c_{1}x}{1-\cfrac{(\ifrac{c_{2}}{c_{1}}%
)x}{1+(\ifrac{c_{2}}{c_{1}})x-\cfrac{(\ifrac{c_{3}}{c_{2}})x}{1+(\ifrac{c_{3}}%
{c_{2}})x-\cdots\cfrac{(\ifrac{c_{n}}{c_{{n-1}}})x}{1+(\ifrac{c_{n}}{c_{{n-1}}%
})x}}}},n=0,1,2,\dots,

when c_{k}\not=0, k=1,2,3,\dots.

Fractional Transformations

Define

1.12.20C_{n}(w)=b_{0}+\cfrac{a_{1}}{b_{1}+\cfrac{a_{2}}{b_{2}+\cdots\frac{a_{n}}{b_{n%
}+w}}}.

Then

§1.12(iii) Existence of Convergents

A sequence \{C_{n}\} in the extended complex plane, \Complex\cup\{\infty\}, can be a sequence of convergents of the continued fraction (1.12.3) iff

1.12.22
C_{0}\not=\infty,
C_{n}\not=C_{{n-1}},n=1,2,3,\dots.

§1.12(iv) Contraction and Extension

A contraction of a continued fraction C is a continued fraction C^{{\prime}} whose convergents \{C^{{\prime}}_{n}\} form a subsequence of the convergents \{C_{n}\} of C. Conversely, C is called an extension of C^{{\prime}}. If C^{{\prime}}_{n}=C_{{2n}}, n=0,1,2,\dots, then C^{{\prime}} is called the even part of C. The even part of C exists iff b_{{2k}}\not=0, k=1,2,\dots, and up to equivalence is given by

1.12.23b_{0}+\cfrac{a_{1}b_{2}}{a_{2}+b_{1}b_{2}-\cfrac{a_{2}a_{3}b_{4}}{a_{3}b_{4}+b%
_{2}(a_{4}+b_{3}b_{4})-\cfrac{a_{4}a_{5}b_{2}b_{6}}{a_{5}b_{6}+b_{4}(a_{6}+b_{%
5}b_{6})-\cfrac{a_{6}a_{7}b_{4}b_{8}}{a_{7}b_{8}+b_{6}(a_{8}+b_{7}b_{8})-%
\cdots}}}}.

If C^{{\prime}}_{n}=C_{{2n+1}}, n=0,1,2,\dots, then C^{{\prime}} is called the odd part of C. The odd part of C exists iff b_{{2k+1}}\not=0, k=0,1,2,\dots, and up to equivalence is given by

1.12.24\frac{a_{1}+b_{0}b_{1}}{b_{1}}-\cfrac{a_{1}a_{2}b_{3}/b_{1}}{a_{2}b_{3}+b_{1}(%
a_{3}+b_{2}b_{3})-\cfrac{a_{3}a_{4}b_{1}b_{5}}{a_{4}b_{5}+b_{3}(a_{5}+b_{4}b_{%
5})-\cfrac{a_{5}a_{6}b_{3}b_{7}}{a_{6}b_{7}+b_{5}(a_{7}+b_{6}b_{7})-\cdots}}}.

§1.12(v) Convergence

A continued fraction converges if the convergents C_{n} tend to a finite limit as n\to\infty.

Pringsheim’s Theorem

The continued fraction \displaystyle{\cfrac{a_{1}}{b_{1}+\cfrac{a_{2}}{b_{2}+\cdots}}} converges when

1.12.25|b_{n}|\geq|a_{n}|+1,n=1,2,3,\dots.

With these conditions the convergents C_{n} satisfy |C_{n}|<1 and C_{n}\to C with |C|\leq 1.

Van Vleck’s Theorem

Let the elements of the continued fraction \displaystyle{\cfrac{1}{b_{1}+\cfrac{1}{b_{2}+\cdots}}} satisfy

1.12.26-\tfrac{1}{2}\pi+\delta<\mathop{\mathrm{ph}\/}\nolimits b_{n}<\tfrac{1}{2}\pi-\delta,n=1,2,3,\dots,

where \delta is an arbitrary small positive constant. Then the convergents C_{n} satisfy

1.12.27-\tfrac{1}{2}\pi+\delta<\mathop{\mathrm{ph}\/}\nolimits C_{n}<\tfrac{1}{2}\pi-\delta,n=1,2,3,\dots,

and the even and odd parts of the continued fraction converge to finite values. The continued fraction converges iff, in addition,

1.12.28\sum^{\infty}_{{n=1}}|b_{n}|=\infty.

In this case |\mathop{\mathrm{ph}\/}\nolimits C|\leq\tfrac{1}{2}\pi.

§1.12(vi) Applications

For analytical and numerical applications of continued fractions to special functions see §3.10.