Cohomology tree probability distribution: Difference between revisions

From Groupprops
(Created page with "{{wikilocal}} ==Definition== Let <math>p</math> be a prime number and <math>n</math> be a positive integer. The '''cohomology tree probability distribution''' is a probability distribution on the set of isomorphism classes of groups of order <math>p^n</math> defined inductively as follows. ===Base case of inductive definition: definition for prime order (<math>n = 1</math>)=== There is only one group of order <math>p</math>, namely the cyclic group (see group o...")
 
 
(6 intermediate revisions by the same user not shown)
Line 5: Line 5:
Let <math>p</math> be a [[prime number]] and <math>n</math> be a positive integer. The '''cohomology tree probability distribution''' is a probability distribution on the set of isomorphism classes of groups of order <math>p^n</math> defined inductively as follows.
Let <math>p</math> be a [[prime number]] and <math>n</math> be a positive integer. The '''cohomology tree probability distribution''' is a probability distribution on the set of isomorphism classes of groups of order <math>p^n</math> defined inductively as follows.


===Base case of inductive definition: definition for prime order (<math>n = 1</math>)===
===Base case of inductive definition: definition for prime order (n = 1)===


There is only one group of order <math>p</math>, namely the cyclic group (see [[group of prime order]]). Therefore there is only one allowed probability distribution, that assigns a weight of 1 to this unique group.
There is only one group of order <math>p</math>, namely the cyclic group (see [[group of prime order]]). Therefore there is only one allowed probability distribution, that assigns a weight of 1 to this unique group.


===Induction step: probability distribution for groups of order <math>p^n</math> based on probability distribution for groups of order <math>p^{n-1}</math> for <math>n > 1</math>===
===Induction step: probability distribution for groups of order p^n based on probability distribution for groups of order p^{n - 1} for n > 1===


Conceptually, the induction step is essentially a convolution product with the convolution process being the use of the second cohomology group for the trivial action of groups of order <math>p^{n-1}</math> on the group of order <math>p</math>, to define group extensions. Let's go over this more specifically.
Conceptually, the induction step is essentially a convolution product with the convolution process being the use of the second cohomology group for the trivial action of groups of order <math>p^{n-1}</math> on the group of order <math>p</math>, to define group extensions. Let's go over this more specifically.
Line 15: Line 15:
Denote by <math>C</math> the cyclic group of order <math>p</math>.
Denote by <math>C</math> the cyclic group of order <math>p</math>.


For any group <math>G</math> of order <math>p^{n-1}</math>, the elements of the [[second cohomology group for trivial group action]] <math>H^2(G, C)</math> correspond to extensions with central subgroup <math>C</math> and quotient group <math>G</math>. Each of these extensions is therefore a group of order <math>p^n</math>. For each element of <math>H^2(G, C)</math>, give the isomorphism class (as a group of order <math>p^n</math>) of the corresponding group extension, a weight that equals the probability distribution weight of <math>G</math> divided by the size of <math>H^2(G, C)</math>.
For any group <math>G</math> of order <math>p^{n-1}</math>, the elements of the [[second cohomology group for trivial group action]] <math>H^2(G; C)</math> correspond to extensions with central subgroup <math>C</math> and quotient group <math>G</math>. Each of these extensions is therefore a group of order <math>p^n</math>. For each element of <math>H^2(G; C)</math>, give the isomorphism class (as a group of order <math>p^n</math>) of the corresponding group extension, a weight that equals the probability distribution weight of <math>G</math> divided by the size of <math>H^2(G; C)</math>.


Now, sum up these weights as <math>G</math> varies over all isomorphism classes of groups of order <math>p^n</math>, to get a probability distribution for isomorphism classes of groups of order<math>p^n</math>.
Now, sum up these weights as <math>G</math> varies over all isomorphism classes of groups of order <math>p^n</math>, to get a probability distribution for isomorphism classes of groups of order<math>p^n</math>.
===Notes===
* [[Prime power order implies not centerless]] can be used to show that every group of order <math>p^n</math> has a positive weight. Specifically, the center is nontrivial, so it has a subgroup of prime order, and the whole group can be obtained as an extension with normal subgroup as this group of prime order, and quotient of order <math>p^{n-1}</math>.
* The term "cohomology tree" refers to the idea that this process of generating extensions can be depicted using a tree, whose root is the group of order <math>p</math>, the next layer is the groups of order <math>p^2</math>, and so on. Each layer of the tree is groups of order <math>p^n</math> for some <math>n</math>. Each node of the tree can be assigned a numeric value which is its weight in its probability distribution; the sum of the numeric values in each layer is 1. The value for a given node is the sum of the values of all its children.
==Worked example for groups of prime-square order==
Let's work out the cohomology tree probability distribution for [[groups of prime-square order]], i.e., groups of order <math>p^2</math> where <math>p</math> is a prime number.
Denote by <math>C</math> the cyclic group of order <math>p</math>.
Since there's only one group of order <math>p</math>, namely <math>C</math>, the cohomology tree probability distribution for order <math>p^2</math> is just based on the distribution of isomorphism classes of groups corresponding to the [[second cohomology group for trivial group action]] <math>H^2(C, C)</math>.
This cohomology group is worked out at [[second cohomology group for trivial group action of group of prime order on group of prime order]]. The group <math>H^2(C; C)</math> has order <math>p</math>, with the identity element corresponding to the extension that is the elementary abelian group of order <math>p^2</math>, and the remaining <math>p - 1</math> non-identity elements corresponding to the extension that is the cyclic group of order <math>p^2</math>.
The cohomology tree probability distribution therefore works out to the following:
* Cyclic group of order <math>p^2</math>: This appears <math>p - 1</math> out of <math>p</math> times, so it gets weight <math>(p - 1)/p</math> or equivalently <math>1 - 1/p</math>.
* Elementary abelian group of order <math>p^2</math>: This appears 1 out of <math>p</math> times, so it gets weight <math>1/p</math>.
==Sketch of worked example for groups of order 8==
Let's work out the cohomology tree probability distribution for groups of order 8.
For groups of order 4, we have, per the above distribution for groups of prime-square order, that [[cyclic group:Z4]] and [[Klein four-group]] (the elementary abelian group of order 4) both have weight 1/2.
To get the probability distribution for groups of order 8, we need to look at two cohomology groups.
===Second cohomology group for trivial group action of cyclic group of order 4 on cyclic group of order 2===
This is covered in [[second cohomology group for trivial group action of Z4 on Z2]]. The cyclic group of order 4 itself has weight 1/2 in the cohomology tree probability distribution on groups of order 4, so the weights need to be multiplied by 1/2.
The result is that the second cohomology group is cyclic of order two, and produces two isomorphism classes:
* [[Direct product of Z4 and Z2]] occurs once, as the identity element, so it gets weight 1/2 times 1/2 = 1/4.
* [[Cyclic group:Z8]] occurs once, as the non-identity element, so it gets weight 1/2 times 1/2 = 1/4.
===Second cohomology group for trivial group action of Klein four-group on cyclic group of order 2===
This is covered in [[second cohomology group for trivial group action of V4 on Z2]]. The Klein four-group itself has weight 1/2 in the cohomology tree probability distribution on groups of order 4, so the weights need to be multiplied by 1/2.
The result is that the second cohomology group is cyclic of order 8, and produces four isomorphism classes:
* [[Elementary abelian group:E8]] occurs once, as the identity element, so it gets weight 1/2 times 1/8 = 1/16.
* [[Direct product of Z4 and Z2]] occurs 3 times, so it gets weight 1/2 times 3/8 = 3/16.
* [[Dihedral group:D8]] occurs 3 times, so it gets weight 1/2 times 3/8 = 3/16.
* [[Quaternion group]] occurs once, so it gest weight 1/2 times 1/8 = 1/16.
===Summing up===
* [[Cyclic group:Z8]] is in only the first list, with a total weight of 1/4.
* [[Direct product of Z4 and Z2]] is in both lists, with weights of 1/4 and 3/16, totaling to 7/16.
* [[Dihedral group:D8]] is in only the second list, with a total weight of 3/16.
* [[Quaternion group]] is in only the second list, with a total weight of 1/16.
* [[Elementary abelian group:E8]] is in only the second list, with a total weight of 1/16.
We can verify that these weights add up to 1, confirming that we have obtained a probability distribution on the groups of order 8.

Latest revision as of 03:51, 7 December 2024

BEWARE! This term is nonstandard and is being used locally within the wiki. [SHOW MORE]

Definition

Let be a prime number and be a positive integer. The cohomology tree probability distribution is a probability distribution on the set of isomorphism classes of groups of order defined inductively as follows.

Base case of inductive definition: definition for prime order (n = 1)

There is only one group of order , namely the cyclic group (see group of prime order). Therefore there is only one allowed probability distribution, that assigns a weight of 1 to this unique group.

Induction step: probability distribution for groups of order p^n based on probability distribution for groups of order p^{n - 1} for n > 1

Conceptually, the induction step is essentially a convolution product with the convolution process being the use of the second cohomology group for the trivial action of groups of order on the group of order , to define group extensions. Let's go over this more specifically.

Denote by the cyclic group of order .

For any group of order , the elements of the second cohomology group for trivial group action correspond to extensions with central subgroup and quotient group . Each of these extensions is therefore a group of order . For each element of , give the isomorphism class (as a group of order ) of the corresponding group extension, a weight that equals the probability distribution weight of divided by the size of .

Now, sum up these weights as varies over all isomorphism classes of groups of order , to get a probability distribution for isomorphism classes of groups of order.

Notes

  • Prime power order implies not centerless can be used to show that every group of order has a positive weight. Specifically, the center is nontrivial, so it has a subgroup of prime order, and the whole group can be obtained as an extension with normal subgroup as this group of prime order, and quotient of order .
  • The term "cohomology tree" refers to the idea that this process of generating extensions can be depicted using a tree, whose root is the group of order , the next layer is the groups of order , and so on. Each layer of the tree is groups of order for some . Each node of the tree can be assigned a numeric value which is its weight in its probability distribution; the sum of the numeric values in each layer is 1. The value for a given node is the sum of the values of all its children.

Worked example for groups of prime-square order

Let's work out the cohomology tree probability distribution for groups of prime-square order, i.e., groups of order where is a prime number.

Denote by the cyclic group of order .

Since there's only one group of order , namely , the cohomology tree probability distribution for order is just based on the distribution of isomorphism classes of groups corresponding to the second cohomology group for trivial group action .

This cohomology group is worked out at second cohomology group for trivial group action of group of prime order on group of prime order. The group has order , with the identity element corresponding to the extension that is the elementary abelian group of order , and the remaining non-identity elements corresponding to the extension that is the cyclic group of order .

The cohomology tree probability distribution therefore works out to the following:

  • Cyclic group of order : This appears out of times, so it gets weight or equivalently .
  • Elementary abelian group of order : This appears 1 out of times, so it gets weight .

Sketch of worked example for groups of order 8

Let's work out the cohomology tree probability distribution for groups of order 8.

For groups of order 4, we have, per the above distribution for groups of prime-square order, that cyclic group:Z4 and Klein four-group (the elementary abelian group of order 4) both have weight 1/2.

To get the probability distribution for groups of order 8, we need to look at two cohomology groups.

Second cohomology group for trivial group action of cyclic group of order 4 on cyclic group of order 2

This is covered in second cohomology group for trivial group action of Z4 on Z2. The cyclic group of order 4 itself has weight 1/2 in the cohomology tree probability distribution on groups of order 4, so the weights need to be multiplied by 1/2.

The result is that the second cohomology group is cyclic of order two, and produces two isomorphism classes:

Second cohomology group for trivial group action of Klein four-group on cyclic group of order 2

This is covered in second cohomology group for trivial group action of V4 on Z2. The Klein four-group itself has weight 1/2 in the cohomology tree probability distribution on groups of order 4, so the weights need to be multiplied by 1/2.

The result is that the second cohomology group is cyclic of order 8, and produces four isomorphism classes:

Summing up

We can verify that these weights add up to 1, confirming that we have obtained a probability distribution on the groups of order 8.