Zulip Chat Archive

Stream: maths

Topic: Dihedral group element composition definitions


Ben (Nov 30 2022 at 21:29):

Getting a little confused by the dihedral group definition and what r and sr are... these *facts* here don't seem to line up with the ones on wikipedia (yah not a great source ik). Are the ones on wikipedia incorrect or referring to different things... image.png. In mathlib dihedral_group.r i * dihedral_group.sr j = dihedral_group.sr (j - i) whereas wikipedia states r^i * s^j=s^{i+j}?

Adam Topaz (Nov 30 2022 at 21:34):

These formulas depend on how you define s_j. If we take the "standard" presentation to be
D2n=r,s  rn=s2=(rs)2=1D_{2n} = \langle r,s \ | \ r^n = s^2 = (rs)^2 = 1 \rangle
and define rj=rjr_j = r^j and sj=rjss_j = r^j \cdot s, then we would have risj=si+jr_i \cdot s_j = s_{i+j}.

But if we define sj=srjs_j = s \cdot r^j then we would have risj=risrj=srji=sjir_i \cdot s_j = r_i \cdot s \cdot r^j = s \cdot r^{j-i} = s_{j-i}.

Adam Topaz (Nov 30 2022 at 21:37):

It's all just a question of what you mean by the "j-th" reflection and the "i-th" rotation. You can parameterize these things in different ways.

Ben (Nov 30 2022 at 21:44):

hmm never seen that chevron construction for the dihedrals before...
So does the mathlib def sr i mean a reflection then a rotation of i? Both wikipedia and mathlib say that s or sr is a "reflection".

Also what is the difference between rir_i and rir^i...?

Adam Topaz (Nov 30 2022 at 21:46):

What is your definition of D_{2n}?

Ben (Nov 30 2022 at 22:20):

the dihedral group... wait you can define the D2nD_{2n} in different ways...

Kevin Buzzard (Nov 30 2022 at 22:21):

There are loads of ways to define it, which is why your question is so hard to understand

Kevin Buzzard (Nov 30 2022 at 22:21):

Different definitions, especially using similar or the same notation to mean different things, can lead to confusion if you don't understand the material properly

Ben (Nov 30 2022 at 22:23):

I think I like the lean definition, where you can assume an element is either one of two variants.

Adam Topaz (Nov 30 2022 at 22:24):

Okay, so you want to define D2nD_{2n} as some type with 2n2 \cdot n elements, denoted
{r0,,rn1,s0,,sn1}\{r_0,\ldots,r_{n-1},s_0,\ldots,s_{n-1}\},
but now the question is how you want to define the product!

Junyan Xu (Nov 30 2022 at 22:27):

Ben said:

So does the mathlib def sr i mean a reflection then a rotation of i?

By usual convention, where multiplication means function composition, I think it means a rotation then a reflection: when you apply a*b, b is first applied, then a.

Both wikipedia and mathlib say that s or sr is a "reflection".

Yes, in a dihedral group, the product of a reflection with a rotation (to the left or to the right) is always a reflection.

Also what is the difference between rir_i and rir^i...?

I think there's no difference.

Ben (Nov 30 2022 at 22:36):

Yah I guess it's like how Peano numbers are only one way of defining the natural numbers... didn't realise there were different forms, especially when they use the the same words and similar terms...

Ben (Nov 30 2022 at 22:38):

So the element composition definitions on wikipedia are correct, they are just using a different definition (which they don't define, just call it a reflection)

Ben (Nov 30 2022 at 22:40):

What does the s and r in the sr constructor mean in the definition on mathlib... how does it match up geometrically on a ngon?

Kevin Buzzard (Nov 30 2022 at 22:42):

That depends on how you're defining the group law on the n-gon. If you do reflection aa and then reflection bb, have you just done abab or baba? Both answers are "valid" in the sense that both will give you groups, but the groups won't be identical (they will be isomorphic, but the isomorphism won't be the identity function)

Adam Topaz (Nov 30 2022 at 22:45):

If you have the regular nn-gon, centered in the plane, with a vertex on the yy-axis, you can define rr to be rotation counter-clockwise by 2π/n2 \cdot \pi /n radians (equiv. multiplication by exp(2πi/n)\exp(2 \cdot \pi \cdot i / n) in the complex plane. Also define ss to be reflection about the yy-axis.

With these conventions, the collection of all symmetries of the regular nn-gon are {r0,r1,,rn1,r0s,r1s,,rn1s}\{r^0,r^1,\ldots,r^{n-1},r^0s,r^1s,\ldots,r^{n-1}s\}. The multiplication rules follow from the formulas: rn=s2=(rs)2=1r^n = s^2 = (rs)^2 = 1 (left as an exercise).

Adam Topaz (Nov 30 2022 at 22:45):

In this case, you can define ri=rir_i = r^i and si=riss_i = r^i \cdot s to get risj=si+jr_i \cdot s_j = s_{i+j} (arithmetic being done modulo nn).

Kevin Buzzard (Nov 30 2022 at 22:46):

Adam the convention in mathlib is that sr i means srisr^i so doing rjsr^js is only adding to the confusion.

Adam Topaz (Nov 30 2022 at 22:47):

Alternatively, note that ris=srir^i s = s r^{-i} so you can equivalently view D2nD_{2n} as the set {r0,r1,,rn1,sr0,sr1,,srn1}\{r^0,r^1,\ldots,r^{n-1},sr^0,sr^1,\ldots,sr^{n-1}\} and in this case if you define sj=srjs_j = s r^j then you have risj=sjir_i s_j = s_{j-i}.

Adam Topaz (Nov 30 2022 at 22:47):

@Kevin Buzzard yeah that's kind of the point, since wikipedia seems to use sj=rjss_j = r^j s.

Kevin Buzzard (Nov 30 2022 at 22:49):

Oh I see, you were half way through an explanation :-) Yeah, both of these work, and interpreting them as acting on the n-gon you have to decide whether you're acting on the left or the right (or equivalently whether abab means "do aa then bb" or "do bb then aa") which introduces yet another sign ambiguity.

Adam Topaz (Nov 30 2022 at 22:49):

as Junyan's message mentioned, the "standard" convention is that aba \cdot b first acts by bb then by aa.

Adam Topaz (Nov 30 2022 at 22:50):

So for example sr2sr^2 will first apply the rotation rr twice, then apply the reflection, whereas r2sr^2s will first apply ss then apply rr twice.

Junyan Xu (Nov 30 2022 at 22:54):

So if you map r i in mathlib to rir_i in Wikipedia and map sr i to snis_{n-i} ("sis_{-i}"), you get an isomorphism between the two dihedral groups.


Last updated: Dec 20 2023 at 11:08 UTC