Analytic Combinatorics/Darboux Method

From testwiki
Jump to navigation Jump to search

Introduction

Darboux's method is one way of estimating the coefficients of generating functions involving roots.

It is easier than [[../Singularity_Analysis|Singularity Analysis]], but it applies to a smaller set of functions.

Big O

We will make use of the "Big O" notation.

f(z)=O(g(z)) as zζ

which means there exists positive real numbers M,δ such that if 0<|zζ|<δ:

|f(z)|Mg(z)

Alternatively, we can say that

f(z)=O(g(n))

for positive integer n, meaning there exists a positive real number M and positive integer N such that:

|f(n)|Mg(n) for nN

Theorem

Theorem due to Wilf[1].

If we have a function (1z)βf(z) where β{0,1,2,} where f(z) has a radius of convergence greater than 1 and an expansion near 1 of j0fj(1z)j, then:

[zn](1z)βf(z)=j=0mfjnβj1Γ(βj)+O(nmβ2)

Example

The theorem is a bit abstract, so I will show an example of how you might use it before going into the proof.

Taking an example from Wilf[2]:

ez/2z2/41z=(1z)12ez/2z2/4

ez/2z2/4 is a complete function, so its radius of convergence is greater than 1.

Near 1 it can be expanded using the Taylor series:

ez/2z2/4=e3/4+e3/4(1z)+

Therefore, for m=0:

[zn]ez/2z2/41z=e3/4n12Γ(12)+O(n32)

Or, if we want more precision we can set m=1:

[zn]ez/2z2/41z=e3/4n12Γ(12)+e3/4n32Γ(12)+O(n52)

and so on.

Proof

Proof due to Wilf[3].

We have:

(1z)βf(z)=j0fj(1z)β+j

and:

j0fj(1z)β+j=j=0mfj(1z)β+j+j>mfj(1z)β+j

By factoring out (1z)β+m+1 from the last sum:

j>mfj(1z)β+j=(1z)β+m+1g(z)

Therefore:

(1z)βf(z)=j=0mfj(1z)β+j+(1z)β+m+1g(z)

We have to prove that:

  1. [zn]j=0mfj(1z)β+j=j=0mfjnβj1Γ(βj)
  2. [zn](1z)β+m+1g(z)=O(nmβ2)

Proof of 1

By applying #Lemma 1:

[zn]j=0mfj(1z)β+j=j=0mfj[zn](1z)β+jj=0mfjnβj1Γ(βj)

Proof of 2

[zn](1z)β+m+1=O(nmβ2) (by #Lemma 1)
[zn]g(z)=O(θn)(0<θ<1) (because, by assumption in the theorem, the radius of convergence R is greater than 1 and Cauchy's inequality tells us that [zn]g(z)=O(1Rn) and 1Rn<1)
|0kn2hkgnk|{max0kn2|hk|}{0kn2Aθnk}=max{B,Bnmβ2}Cθn2Dθn2=O(θn2) (for A,B,C,D constants and assuming that mβ2<0).
|n2knhkgnk|{maxn2kn|hk|}{n2knAθnk}Bnmβ2=O(nmβ2)

because {n2knAθnk}Cmaxn2knθnk=C because θ2θ1θ0=1.

Putting it all together:

[zn](1z)β+m+1g(z)=O(θn2)+O(nmβ2)=O(nmβ2)

because Cθn2=o(nmβ2)[4] because θ<1[5].

Lemma 1

[zn](1z)βnβ1Γ(β)

Proof:

[zn](1z)β=(βn)(1)n=(nβ1n)=Γ(nβ)Γ(β)Γ(n+1)=1Γ(β)(nβ)(β+1)nβ1Γ(β)[6]

where x(n) is the rising factorial.

Szegő's theorem

We can apply a similar theorem to functions with multiple singularities. From Wilf[7] and Szegő[8].

If f(z) is analytic in |z|<1, has a finite number of singularities eiϕ1,eiϕ2,,eiϕr on the unit circle |z|=1 and in the neighbourhood of each singularity has the expansion

f(z)=v0cv(k)(1zeiϕk)αk+vβk(k=1,2,,r and βk>0)

Then we have the asymptotic series

[zn]f(z)=v0k=1rcv(k)(αk+vβkn)(eiϕk)n

Notes

Template:Reflist

References

Template:BookCat

  1. Wilf 2006, pp. 194.
  2. Wilf 2006, pp. 195.
  3. Wilf 2006, pp. 193-195.
  4. w:Big_O_notation#Little-o_notation.
  5. w:Big_O_notation#Orders_of_common_functions.
  6. w:Falling_and_rising_factorials#Properties.
  7. Wilf 2006, pp. 196.
  8. Szegő 1975, pp. 207-208.