Analytic Combinatorics/Meromorphic Functions

From testwiki
Revision as of 10:09, 9 March 2025 by imported>Dom walden
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Introduction

This article explains how to estimate the coefficients of meromorphic generating functions which have only one dominant singularity.

Theorem

Theorem due to Sedgewick[1].

(1)mmf(a)amg(m)(a)(1a)nnm1

Proof

Proof due to Sedgewick[3] and Wilf[4].

h(z)=hm(za)m++h1za+h0+h1(za)+
hm(za)m++h1za
(1)mhmam(n+m1n)(1a)n
limza(za)mh(z)=m!f(a)g(m)(a)
  • (n+m1n)nm1(m1)! as n (Proof)
  • Therefore, putting it all together:
[zn]h(z)(1)mhmam(n+m1n)(1a)n(1)mmf(a)amg(m)(a)(1a)nnm1 as n.

Asymptotic equality

We will make use of the asymptotic equality

f(z)g(z) as zζ

which means

limzζf(z)g(z)=1

This allows us to use g(z) as an estimate of f(z) as z gets closer to ζ.

For example, we often present results of the form

ang(n) as n

which means, for large n, g(n) becomes a good estimate of an.

Meromorphic function

The above theorem only applies to a class of generating functions called meromorphic functions. This includes all rational functions (the ratio of two polynomials) such as 1(1z)2 and z1zz2.

A meromorphic function is the ratio of two analytic functions. An analytic function is a function whose complex derivative exists[7].

One property of meromorphic functions is that they can be represented as Laurent series expansions, a fact we will use in the proof.

It is possible to estimate the coefficients of functions which are not meromorphic (e.g. ln(z) or ez). These will be covered in future chapters.

Laurent series

When we want a series expansion of a function f(z) around a singularity c, we cannot use the Taylor series expansion. Instead, we use the Laurent series expansion[8]:

+a2(zc)2+a1zc+a0+a1(zc)+a2(zc)2+

Where an=12πiγf(z)(zc)n+1dz and γ is a contour in the annular region in which f(z) is analytic, illustrated below.

Pole

A pole is a type of singularity.

A singularity of h(z) is a value of z for which h(z)=[9]

If limza(za)mh(z)=L0 and L is defined then a is called a pole of h(z) of order m[10].

We will make use of this fact when we calculate hm.

For example, 1(12z)2 has the singularity 12 because 1(1212)2=1(11)2=102=10= and 12 is a pole of order 2 because limz12(z12)21(12z)2=limz12(z12)21(2)2(z12)2=limz121(2)2=14.

Closest to the origin

We are treating h(z) as a complex function where the input z is a complex number.

A complex number has two parts, a real part (Re) and an imaginary part (Im). Therefore, if we want to represent a complex number we do so in a two-dimensional graph.

If we want to compare the "size" of two complex numbers, we compare how far they are away from the origin in the two-dimensional plane (i.e. the length of the blue arrow in the above image). This is called the modulus, denoted |z|.

Principle part

Proof due to Wilf[11].

The principle part of a Laurent series expansion are the terms with a negative exponent, i.e.

hm(za)m++h1za

We will denote the principle part of h(z) at a by PP(h,a).

If a is the pole closest to the origin then the radius of convergence R=|a| and as a consequence of the Cauchy-Hadamard theorem[12]:

[zn]h(z)(1|a|+ϵ)n (for some ϵ>0 and for sufficiently large n)

h(z)PP(z) no longer has a pole at a because (hm(za)m++h1za+h0+h1(za)+)(hm(za)m++h1za)=h0+h1(za)+.

If the second closest pole to the origin of h(z) is a then a is the largest pole of h(z)PP(z) and, by the above theorem, the coefficients of h(z)PP(h,a)(1|a|+ϵ)n (for sufficiently large n).

Therefore, the coefficients of PP(h,a) are at most different from the coefficient of h(z) by (1|a|+ϵ)n (for sufficiently large n).

Note that if a is the only pole, the difference is at most ϵn (for sufficiently large n).

If |a|>|a| then as n gets large (1|a|)n will be much smaller than (1|a|)n and, therefore, PP(h,a) is a good enough approximation of h(z).

However, if |a|=|a| then the behaviour of the coefficients is more complicated.

Biggest coefficient

Compare:

[zn]hm(za)m=hmam(n+m1n)(1a)nhmam(1a)nnm1[13]

with:

[zn]h(m1)(za)m1=h(m1)am1(n+m2n)(1a)nh(m1)am1(1a)nnm2

The nth coefficient of the former is only different to the latter by O(nm2an).

Computation of coefficient of first term

hm(za)m=(1)mhmam(1za)m by factoring out (1a)m.

(1)mhmam(1za)m=(1)mhmamn0(n+m1n)(za)n by the binomial theorem for negative exponents[14].

Computation of h_-m

h(z)=hm(za)m+h(m1)(za)m1++h1za+h0+h1(za)+.

Therefore, limza(za)mh(z)=limzahm+h(m1)(za)++h1(za)m1+h0(za)m+h1(za)m+1+=hm.

To compute limza(za)mh(z)=limza(za)mf(z)g(z), because the numerator and denominator are both 0 at a, we need to use L'Hôpital's rule[15]:

limza(za)mf(z)g(z)=limza((za)mf(z))g(z)

Indeed, if a is a root of order m>1 of g(z) and (za)mf(z), it is also a root of g(z) and ((za)mf(z)) and therefore limza((za)mf(z))g(z) is also indeterminate. Therefore, we need to apply L'Hôpital's rule m times:

limza((za)mf(z))(m)g(m)(z)=limza((za)mf(z)+m(za)m1f(z))(m1)g(m)(z)=limza((za)mf(z)+2m(za)m1f(z)+m(m1)(za)m2f(z))(m2)g(m)(z)==m!f(a)g(m)(a)

Proof of binomial asymptotics

(n+m1n)=(n+m1)!n!(m1)!=(n+1)(n+2)(n+m1)(m1)!nm1(m1)! as n.

Notes

Template:Reflist

References

Template:BookCat

  1. Sedgewick, pp. 59.
  2. Sedgewick, (errata), pp. 8.
  3. Sedgewick, pp. 59-60.
  4. Wilf 2006, pp. 185-186.
  5. See Stroud 2003, pp. 919-923, Lang 1999, pp. 161-163, Orloff, pp. 10-13, w:Laurent_series, v:Complex_Analysis_in_plain_view#Laurent_Series_and_the_z-Transform_Example_Note.
  6. Wilf 2006, pp. 185-186.
  7. Flajolet and Sedgewick 2009, pp. 231.
  8. Stroud 2003, pp. 919-920.
  9. This is a bit of an over-simplification. For further information, see Stroud 2003, pp. 863-867, 915 and w:Mathematical_singularity.
  10. Stroud 2003, pp. 915.
  11. Wilf 2006, pp. 52, 185-186.
  12. Wilf 2006, pp. 50-52.
  13. See #Computation of coefficient of first term and #Proof of binomial asymptotics.
  14. Biggs 2002, pp. 364-366.
  15. Stroud 2001, pp. 792, v:Calculus/Limits#L'Hôpital's_Rule, w:L'Hôpital's_rule.