Introduction
'Tauberian' refers to a class of theorems with many applications. The full scope of Tauberian theorems is too large to cover here.
We prove here only a particular Tauberian theorem due originally to Hardy, Littlewood and Karamata, which is useful in analytic combinatorics.
Theorem
Theorem from Feller.[1]
If

where the sequence
is monotone (always non-decreasing or always non-increasing),
and
then

Proof
Proof from Feller.[2]
- We express our theorem in terms of a measure
with a density function
such that
.
- The Laplace transform of
is
[3]
- By the power series expansion of
, as
,
.
- Therefore,
as
.
as
.[4]
is the Laplace transform of
. To express this in terms of the measure
, we integrate the latter to get
.
- We use the continuity theorem to show that this implies
, or (setting
)
.
- We prove that this implies
.
- Therefore,
.
Measure and probability distributions
A measure
assigns a positive number to a set. It is a generalisation of concepts such as length, area or volume.
In our case, our measure is the area under the curve of the step function that takes the values of our coefficients,
. Therefore,
. If
is an integer,
. We define
to be the area under the curve confined to the interval
. If the interval
is contained in the interval
for some
and
is the length of
, then
.
A measure can be converted to a probability distribution
with density
, assigning a probability to how likely a random variable will take on a value less than or equal to
. We do this because probability distributions have two properties which we make use of in our proof.
We define
to be the probability that a random variable will take on a value in the set
.
Property 1: Expectation or mean
The expectation or mean of a probability distribution
with density
is calculated as
.
We also define
.
- Theorem 1
- Let
be a sequence of probability distributions with expectations
, if
then
for
.[5]
- Proof
- Assume
. Let
be a continuous function such that
for all
. Let A be an interval such that
, and therefore, for the complement
,
for
sufficiently large.
- Because
is continuous it is possible to partition
into intervals
so small that
oscillates by less than
.
- We can then estimate
in
by a step function
which assumes constant values within each
and such that
for all
. We define
for
, so that
for
.
- Therefore,
.
- For sufficiently large
, 
- Because
then for sufficiently large
,
.
- Putting it all together,

- which implies
.[6]
Property 2: Convergence
- Lemma 1
- Let
be an arbitrary sequence of points. Every sequence of numerical functions contains a subsequence that converges for all
.[7]
Row 4,
, (in blue) is contained in the previous 3 rows, contains all the subsequent rows and converges at
. All
(in green) are contained in
and therefore converge for
. This argument can be repeated for all
.
 |
 |
 |
 |
 |
|
 |
 |
 |
 |
 |
|
 |
 |
 |
 |
 |
|
 |
 |
 |
 |
 |
|
 |
 |
 |
 |
 |
|
 |
 |
 |
 |
 |
|
- Proof
- For a sequence of functions
we can find a subsequence
that converges at
. Out of the subsequence
we find a subsequence
that converges at
. Continue in this way to find sequences
that converge for the point
, but not necessarily any of the previous points.
- Construct a diagonal sequence
. This sequence converges for all
because all but the first
terms are contained in
. This is true for all
.[8]
- Theorem 2
- Every sequence
of probability distributions possesses a subsequence
that converges to a probability distribution
.[9]
- Proof
- By lemma 1, we can find a subsequence
that converges for all points
of a dense sequence. Denote the limit the sequence converges for each
by
. For
that are not one of
,
is the greatest lower bound of all
for
.
is increasing between 0 and 1. If we define
to be equal to
at all points of continuity and
if
is a point of discontinuity. For a point of continuity
, we can find two points such that
,
and
.
are monotone, therefore
. The limit of
differs from
by no more than
, so
as
for all points of continuity.[10]
- Theorem 3
- If
then the limit of every subsequence equals
.[11]
- Proof
- By the definition of limits,
for
. Therefore, for any subsequence
,
when
.[12]
The Laplace Transform can be seen as the continuous analogue of the power series, where
becomes
.
is replaced by
because it is easier to integrate.
We define the Laplace transform of a probability distribution
as
.
If
has the density
then we can also define the Laplace tranform of
as
.[13]
If our density
is zero for
, we can also see that the Laplace transform is equivalent to the expectation
.[14]
- Theorem 4
- Distinct probability distributions have distinct Laplace transforms.[15]
Continuity theorem
- Theorem 5
- Let
be a sequence of probability distributions with Laplace transforms
, then
implies
.[16]
- Proof
- Because the Laplace transforms
are equivalent to expectations, we can use theorem 1 with
to prove that
implies
.
- By theorem 2, if
is a subsequence that converges to
, then the Laplace transforms of
converge to the Laplace transform
of
.
- By assumption in the theorem, the Laplace transforms
converge to
, then by theorem 3 all its subsequences also converge to
so that
.
- Because Laplace transforms are unique,
. But this will be true for every subsequence so by theorem 3
.[17]
This proof can be extended more generally to measure by defining our probability distribution in terms of the measure
,
.[18]
Asymptotics of the density function
- Theorem 6
- If
has a monotone density
and
then
as
.[19]
- Proof
- For
,
.
- As
the right side tends to
. By theorem 2, there exists a sequence
such that

- Therefore

- which implies
. This limit is independent of the chosen sequence
therefore is true for any sequence. For 
as
.[20]
Dense
A subset
of a set
is called dense if the closure of
is equivalent to
, i.e.
. This means that if
then
is either in the subset
or is on the boundary of that subset. If it is on the boundary then we can select elements of
which are arbitrarily close to
.
Notes
- ↑ Feller 1971, pp. 447.
- ↑ Feller 1971, pp. 445-447.
- ↑ Evertse 2024, pp. 152.
- ↑ Mimica 2015, pp. 19.
- ↑ Feller 1971, pp. 249.
- ↑ Feller 1971, pp. 249-250.
- ↑ Feller 1971, pp. 267.
- ↑ Feller 1971, pp. 267.
- ↑ Feller 1971, pp. 267.
- ↑ Feller 1971, pp. 267-268.
- ↑ Feller 1971, pp. 267.
- ↑ Feller 1971, pp. 267-268.
- ↑ Feller 1971, pp. 431-432.
- ↑ Feller 1971, pp. 430.
- ↑ Feller 1971, pp. 430.
- ↑ Feller 1971, pp. 431.
- ↑ Feller 1971, pp. 431.
- ↑ Feller 1971, pp. 433.
- ↑ Feller 1971, pp. 446.
- ↑ Feller 1971, pp. 446.
References