3. Exponential Generating Functions

3.1Species

在这一章节中, 我们会考虑指数生成函数. 其内蕴的哲学思想依旧是通过生成函数来解决组合问题, 不过这一次, 不同于普通生成函数, 我们所面对的问题往往会有一个置换群的群作用, 而我们会想要计数的对象则变成了等价类. 所以, 传统代数组合中指数生成函数的定义则是 , 而非单纯的 .

在这一份讲义中, 我们会考虑一种更为系统的研究, 也就是通过组合物种这一框架/语言来研究关于指数生成函数的问题.

对于范畴论较为熟悉的同学不难看出, 在下述定义中, 一个组合物种 其实就是一个函子 , 其中 的对象为有限集合, 态射为集合之间的双射 .

然而, 在实际使用中, 我们完全不会去使用任何的范畴论语言, 甚至下面给出的更为具体的定义也并不是最好的去思考组合物种的方法. 我们将会把组合物种考虑为一种把组合对象添加标签的方式. 举例来说, 如果我们考虑树这个组合对象, 那么树的组合物种则是, 把没有标签的树添加标签的过程.

举例来说, 你可以把树的组合物种想象成是一个 " 工厂 ", 然后里面有各种各样的 " 原材料 " (它们是没有标签的树). 然后, 我作为 " 客户 ", 给这个 " 工厂 " 一个 " 订单 "(一个大小为 的有限集合 ), 然后工厂里的工人就开始没日没夜地把大小为 的没有标签的树全部提出来, 开始给这些没有标签的树根据我给出的有限集合上标签.

Example 3.1.1 (Graphs). We note there is no set of all graphs, i.e. the class of all graphs form a proper class. Thus, we need to add definite constrains to make the proper class into a set.

In other word, given a finite set , we can form the set of graphs on given by. In particular, we see the following properties:

1.

If then .

2.

We have isomorphism: it is a bijection such that if and only if .

3.

If is a bijection and we have , then there is a unique graph , such that is an isomorphism of graphs.

4.

In other word, a bijection of sets induces a bijection .

Definition 3.1.2. A species is a rule that assigns:

1.

To every finite set , a (finite) set called the set of -structures on .

2.

To every bijection of finite sets, we get a bijection . This is called a transportation of -structures along .

such that:

1.

If then .

2.

If is the identity map, then the transportation should be the identity.

3.

If and are bijections, then .

Definition 3.1.3. Let be a species, and , then we say and are isomorphic if and only if there exists bijection such that .

Definition 3.1.4. Let be a species, assume is finite for every finite set . Then if , we must have and hence we define for any set with . Then we define the exponential generating function associated with to be

Example 3.1.5. We have the following examples of species:

Name/Notation Structures on Transport along EGF
Sets(), i.e.
Linear Orders()
Endofunctions()
Permutations
Cyclic Permutations()
Trees ()same as graph
Set Partitions() is set partitions of

We remark for the last one, are called the Bell numbers and is equal

Definition 3.1.6. Let be two species. A natural transformation is a rule which assigns to every finite set a function , such that if is a bijection of sets, then the following diagram commutes:

Example 3.1.7.

1.

We have a natural transformation from trees to graphs by inclusion (if we have a sub-species then inclusion is always a natural transformation).

2.

From directed graphs to graphs we have the forgetful natural transformation, i.e. we forget the direction of edges.

3.

From endofunctions to directed graphs we have a natural transformation as follows. Given , the edges are given by .

4.

From any species to sets we get a natural transformation by forgetful natural transformation.

5.

From graphs to graphs we have the graph complement as a natural transformation.

Lemma 3.1.8. If are isomorphic, then and are isomorphic.

Definition 3.1.9. Let be a natural transformation. If is a bijection for all finite set , then we say is a natural equivalence and we say and are naturally equivalent and write .

Example 3.1.10.

1.

The graph complement is an equivalence from graphs to graphs.

2.

Let be the species of graphs and be the species of pairs of equal graphs, i.e. . Then the map is a natural equivalence.

3.

We have an equivalence between the species of permutations and the sub-species of where the graphs has every vertex with in-degree and out-degree .

Definition 3.1.11. We say two species and are numerically equivalent if and write .

Remark 3.1.12. We note natural equivalence implies numerically equivalence but not vice versa.

Example 3.1.13. We see and are numerically equivalent but .

Proof. Suppose is a natural transformation. Any two -structure of same order are isomorphic. But this is not true for and hence cannot be natural equivalence by the lemma above.

Definition 3.1.14. A species is connected if and only if .

Remark 3.1.15 (Labelled & Unlabelled Diagrams). Given a species , we can think -structures on as diagrams with “widgets” that are labelled by elements of .

For example, we can think as graphs with each vertex has in-degree and out-degree : We could also have some examples of species that are not “graph-like” but still represented by diagrams. For instance, consider set partitions, and we have the set partition correspond to the following tableau (can’t draw tableau here...)

With this point of view of species, the transportations of -structures are always relabelling of our diagrammatic objects. Then the isomorphism types in this diagrammatic point of view are just unlabelled drawings and now we consider the widgets as “label receivers”. Also, the “order” of species in this context is just the number of label receivers.

We will write for the set of isomorphism types of species . Then, we can just think species as a set of isomorphism types plus a labelling operation.

For example, consider the species of trees, . Then we have some of the isomorphism types as

To form , we see we have two isomorphism types with nodes, and we just need to label them with and this gives us the set , where we are going to list some examples:

3.2Species Operations

Remark 3.2.1. In this section, will always be species.

Definition 3.2.2 (Sum and Difference). We define as follows. The structures are given byas the disjoint union of and . This construction extends to infinite sums. In terms of isomorphism types, we have .

Now assume , i.e. for all . Then we define .

Definition 3.2.3 (Filtering by Order). We define .

The definitions are given as follows

Example 3.2.4. For example, we have is the non-empty set operation. We also define to be the species of singletons and denoted by . Finally we have .

Definition 3.2.5 (Products). We define .

The first one is Cartesian product with a set. Here is a set, usually finite, and we define

The second one is species product. The species product is defined as follows

We note the following:

1.

To put an structure on , we do the following:

1.

partition into .

2.

put an structure on .

3.

put a structure on .

2.

By default, .

3.

is the multiplicative identity of this multiplication.

4.

We note .

The third one is called superposition or Hadamard product. It is defined as

Definition 3.2.6 (Sequence). We define the sequence bywhere . We note:

1.

If is connected and is finite for all , then is connected and is fintie.

2.

Definition 3.2.7 (Rooting and Derivatives). We define and .

The rooting (aka marking or pointing) is defined by

The derivative (aka mark and delete) is defined bywhere is an element not in .

We note:

1.

is more or less and the difference is how we interprete them.

2.

Definition 3.2.8 (Composition). We define (or ). We assume is connected.

Then we defineWe note:

1.

To put an -structure on , we do the following:

1.

Form a set partition of .

2.

Put an -structure on , which is a set with elements.

3.

Put a -structure on each for .

4.

We note is not fixed here.

2.

Unlabelled -structure constructed as follows:

1.

Start with unlabelled -structure

2.

Replace each label receiver in by -structure .

3.

Label receiver of composite object are label receivers of .

Example 3.2.9.

1.

is the species of forest.

2.

is the forest with 2 components. We note as

3.

.

4.

.

5.

.

Theorem 3.2.10. Suppose we have be two species with and , then:

1.

Sum and Difference:

2.

Filters

3.

Products

4.

Sequence:

5.

Derivatives and Rooting

6.

Composition:

Proof. All of the proofs are elementary counting.

For example, we consider . We see

3.3Examples

Remark 3.3.1. We have some common notations. If then we write . We note we have and hence this operation is not linear.

Example 3.3.2 (Connected Graphs). Let be graphs and be connected graphs. We know and therefore we havewhere we know . Hence we see

Example 3.3.3 (Permutations with even number of cycles). The species we want to consider is . We see and and hence we seeThus the number of permutations of with an even number of cycles is

Example 3.3.4 (Surjective Functions). How many surjective functions ?

Method 1: Fix and let be the species and . Then we havewhere we are going to think as surjective functions onto subset and is just the set . Thus we seeand henceand the answer is .

Method 2: Fix and let be the species given byThen we seeas products, where given we think the products as .

Thus we seeand hence the answer is .

Example 3.3.5 (Trees). The species of trees have no natural decomposition. On the other hand, species of rooted trees does. In particular, we havewhere is the root and is the set of rooted tree branches. Thus we getand apply LIFT we getand henceHowever, note each tree has ways to root the trees, and therefore we seeand we conclude the number of trees of nodes is .

Example 3.3.6 (Trivalent trees). Let be the species of trees in which every vertex has degree or . We will use the same trick and first look at the corresponding rooted structure.

However, note given a trivalent tree and if we break up, we do not get a set of trivalent trees and hence we we need to find something else in the recursive definition of .

What we want to use is , the rooted trees in which every vertex has updegree or . Then we getBy the above two relations, we get

ThereforeNow use LIFT we seeThus we seeand hence