4

Trying to understand a bit about the so called Nil-Geometry, I have found two models (are there more?), namely:

  1. The Heisenberg group: we identify the points $(x,y,z)\in \mathbb{R}^3$ with the matrices of the form $\begin{pmatrix}1&x&z\\ 0&1&y\\ 0&0&1\end{pmatrix}$ and consider them as a subgroup of $Gl_3(\mathbb{R})$, endowing $\mathbb{R}^3$ with a new (and nil - sorry for the pun) Lie group structure. The group operation looks like: $$(x_1,y_1,z_1)\ast (x_2,y_2,z_2) = (x_1+x_2,y_1+y_2,z_1+z_2+x_1y_2).$$ Then considering the riemannian metric $$ds_\ast^2:= dx^2+dy^2+(dz-xdy)^2$$ (which is left-invariant), we get a first model for the three-dimensional Nil-geometry.

  2. Consider now another group operation on $\mathbb{R}^3$: $$(x_1,y_1,z_1)\star (x_2,y_2,z_2) = (x_1+x_2,y_1+y_2,z_1+z_2+\frac{1}{2}(x_1y_2-x_2y_1)),$$ and another (also left-invariant) riemannian metric, $$ds_\star^2 = dx^2+dy^2+(dz+\frac{1}{2}(ydx-xdy))^2.$$ This is also said to be a model for three-dimensional Nil-geometry.

One can find these two models in different sources out there and everyone calls them by the same name. Indeed, they somehow look similar and I was able to prove some basic facts, for example, both groups are step-2 nilpotent (therefore considering left-invariant riemannian metrics and calling both Nil-geometry models seems plausible enough).

However, if you optimistically look at the map $((x_1,y_1,z_1),\ast,ds_\ast)\mapsto ((x_1,y_1,z_1),\star,ds_\star)$, hoping it to be an isometry (or at least something like a dilation), you find out it is not even a group homomorphism...

So, what is the precise relation between these two models?

Is there a not so obvious isometry (or dilation) between them?

What does give us the right to call them both Nil-geometry models?

Ѕᴀᴀᴅ
  • 35,369
Derso
  • 2,897
  • 2
    I suppose it's just going to be a coordinate change map. But you have a several typos: two occurrences of $y_1$ in your first equation are wrong; at least one occurrence of $y_1$ in your second equation is wrong; ... – Lee Mosher May 16 '23 at 18:35
  • @LeeMosher I'm sorry for the typos and thank you for pointing them out! I hope everything is correct now. – Derso May 16 '23 at 18:54
  • Do you know what an (X,G)-geometry is? – Moishe Kohan May 27 '23 at 14:57
  • @MoisheKohan I'm not familiar with it, but checked it out. Basically, if G is a group of analytical diffeomorphisms acting on X, then a (X,G)-structure on a manifold M is an atlas whose charts take values on X (so M is a manifold modelled on X) and whose transition maps are restrictions of elements in G. Is that right? – Derso May 29 '23 at 11:03
  • 2
    Almost right. So, what matters is the space plus group action, not a particular choice of an invariant metric. – Moishe Kohan May 29 '23 at 13:07
  • I have a question. The metric $dx^2+dy^2+(dz-xdy)^2 $ defined on $\Bbb R^2\times \Bbb R$ is often called twisted product metric, and denoted by $\Bbb R^2\widetilde\times \Bbb R$. Is $dx^2+dy^2+(dz-xdy)^2 $ isometric to some "standard twisted product" of $(\Bbb R^2, dx^2+dy^2)$ and $(\Bbb R,dz^2)$. – Sumanta Aug 24 '23 at 11:11
  • Here the standard twisted product is defined as follows: Let $(M_1,g_1)$ and $(M_2,g_2)$ be two Riemannian manifolds, and let $f_1,f_2\colon M_1\times M_2\to (0,\infty)$ be two smooth maps. Then define twisted product metric $g$ on $M_1\times M_2$ as $g(X,Y)\colon=f_1^2g_1(\pi_{1}X, \pi_{1}Y)+f_2^2g_1(\pi_{2}X, \pi_{2}Y)$, where $\pi_1,\pi_2$ are canonical projections. https://www.google.co.in/books/edition/Differential_Geometry_Of_Warped_Product/aPAnDwAAQBAJ?hl=en&gbpv=1&dq=twisted+product+metric&pg=PA76&printsec=frontcover – Sumanta Aug 24 '23 at 11:12

1 Answers1

3

Note. This answer is work-in-progress. It will change a lot in the next few minutes. You can also find interesting details at Wikipedia, I am using similar notation as in that article.

In order to detail the relationship between the two models of the Heisenberg group that you have specified, I will first lay out the general Definition of Heisenberg groups on symplectic vector spaces. I will then show how your two models are a special case of this Definition.

Symplectic vector spaces and Darboux basis (TODO: Add references)

Let $V$ be a finite-dimensional vector space over a field $F$ and let $\omega:V\times V\to F$ be a bilinear map such that $\omega(v,v)=0$ for all $v\in V$ and such that for all $u\in V$, we have $\omega(u, v)=0$ for all $v\in V$ if and only if $u=0$ (the latter property is also called "non-degenerateness").

By a standard result from linear algebra, $V$ must have an even dimension.

The standard symplectic vector space is $V=\mathbb R^{2n}$ for some $n\in\mathbb N$ together with the bilinear form $\omega$ given by $\omega(v,w)=\langle v, Aw\rangle$, where $A\in\mathbb R^{2n\times 2n}$ is the matrix $$A=\begin{pmatrix} 0 & \operatorname{Id}_n \\ -\operatorname{Id}_n & 0\end{pmatrix}.$$

For $n=1$, this reduces to $V=\mathbb R^2$ and $\omega((x,y), (\tilde x, \tilde y)) = x \tilde y - \tilde x y$.

The Heisenberg group on symplectic manifolds

Let $(V, \omega)$ be a finite-dimensional real symplectic vector space. The Heisenberg group $H(V)=H(V,\omega)$ on $V$ is defined as the set $V\times\mathbb R$ together with the group operation $\cdot$, mapping $(v,t)$ and $(v', t')$ to

$$(v, t)\cdot (v', t') = \left(v+v', t+t' +\frac 12\omega(v,v')\right).$$

Symplectic representation of the Heisenberg group (TODO: insert references)

By a classical result from linear algebra, every symplectic vector space $(V,\omega)$ of dimension $2n, n\in\mathbb N$, admits a Darboux basis, consisting of vectors $\mathbf e_1,\dots, \mathbf e_n, \mathbf f^1, \dots, \mathbf f^n\in V$, such that $\{\mathbf e_1,\dots,\mathbf e_n, \mathbf f^1,\dots,\mathbf f^n\}$ form a basis of $V$, and such that $\omega(\mathbf e_j, \mathbf f^k)=1$ if $j=k$ and $0$ otherwise.

In the special case $n=1$ with $\omega$ as defined above, such a Darboux basis is given by $\mathbf e_1 = (1,0)$ and $\mathbf f^1 = (0,1)$.

Consider the subspace $\{0\}\times\mathbb R\subset V\times\mathbb R$, and let $E$ denote a basis of this subspace (for example $E=(0,0,\dots, 0, 1)$). (Note that this is wrongly defined in the Wikipedia article.)

A vector $v\in V\times\mathbb R$ can now be written uniquely as (using the Einstein summation convention) $$v=q^a\mathbf e_a + p_a \mathbf f^a + t E$$ for some $q^1,\dots,q^n,p_1,\dots,p_n, t\in\mathbb R$.

Let $p=(p_1,\dots, p_n)$, $q=(q^1,\dots, q^n)$ and denote by $[p, q, t]$ the vector $v$ obtained above. Then the group law can be written as $$[p,q,t]\cdot[p',q',t']=[p+p',q+q',t+t'+\frac 12(\langle p, q'\rangle -\langle p', q\rangle)].$$ This corresponds to the second Definition that you give of the Heisenberg group.

Matrix representation of the Heisenberg group

Consider now all matrices of the form \begin{equation*}\{p,q,u\}=\begin{pmatrix} 1 & p & u \\ 0 & \operatorname{Id}_n & q \\ 0 & 0 & 1 \end{pmatrix} \end{equation*} where $p,q$ are vectors in $\mathbb R^n$ and $u\in\mathbb R$. Denote the set of all such matrices by $M_H(n)$, and define the group $\tilde H$ as $M_H(n)$ equipped with the group operation $$\{p,q,u\}\ \tilde\cdot\ \{p',q',u'\}=\{p+p',q+q',u+u'+\langle p,q'\rangle\}.$$

Equivalence as groups

We now consider the map $T:H(V)\to\tilde H$ given by $T([p,q,t])=\{p,q,t+\frac 12\langle p, q\rangle\}$.

Claim. $T$ is a group isomorphism.

Proof. Proving the fact that $T$ is bijective is left as an exercise to the reader. Now, let $[p,q,t],[p',q',t']\in H(V)$. We have $$T([p,q,t]\cdot[p',q',t'])=T([p+p',q+q',t+t'+\frac 12\langle p, q'\rangle-\frac 12\langle p', q\rangle])=\{p+p',q+q',t+t'+\frac 12 \langle p, q'\rangle - \frac 12\langle p', q\rangle+\frac 12\langle p+p', q+q'\rangle\}=\{p+p',q+q',t+t'+\frac 12\langle p, q\rangle+\frac 12\langle p', q'\rangle+\langle p,q'\rangle\}.$$

On the other hand, $$T([p,q,t])\ \tilde\cdot\ T([p',q',t']) =\{p,q,t+\frac 12\langle p,q\rangle\}\tilde\cdot\{p',q',t'+\frac 12\langle p',q'\rangle\}=\{p+p',q+q',t+\frac 12\langle p,q\rangle+t'+\frac 12\langle p',q'\rangle\}.$$

Since both terms are equal, $T$ is thus indeed a group isomorphism. $\square$

Riemannian metric equivalence (TODO: finish writing paragraph)

We now define Riemannian metrics on $H(V)$ and $\tilde H$, considered as submanifolds of $\mathbb R^{2n+1}$, as $g($