The binary tetrahedral group $\mathbb T$ is an interesting 24-element group. For instance it can be expressed as the subgroup $$ \mathbb T = \left\{ \pm 1, \pm i, \pm j, \pm k, \dfrac{\pm 1 \pm i \pm j \pm k}2 \right\} \subseteq \mathbb{H}^\times$$ of the multiplicative group of the quaternions (which also gives you a fascinating regular 24-vertex 4-polyhedron).

Alternatively, the binary tetrahedral group can be thought of as the inverse image of $\mathfrak A(4)$ through the 2:1 morphism $\rm{SU}_2 \to \rm{SO}_3$ ($\mathfrak A(4) \subseteq \rm{SO}_3$ being the group of direct isometries of the regular tetrahedron).

The property which puzzles me is the following: $\mathbb T$ is isomorphic to the matrix group $\rm{SL}_2(\mathbb F_3)$.

I'm able to give a proof of this result, but not an enlightening one. Essentially, a 24-element group with a normal subgroup isomorphic to the 8-element quaternion group $\rm Q_8$ is either isomorphic to $\rm Q_8 \times \rm C_3$ or to the unique nontrivial semidirect product $\rm Q_8 \rtimes \rm C_3$; it's not that hard to prove that both $\mathbb T$ and $\rm{SL}_2(\mathbb F_3)$ satisfy this property and belong to the second case. The only interesting thing about this proof is the description of that normal $\rm Q_8$: on the one hand $\{\pm 1, \pm i, \pm j, \pm k\} \subset \mathbb T$ is the inverse image of the quite exceptional Vierergruppe $V_4 = \{ \rm{id}, (12)(34), (13)(24), (14)(23)\} \lhd \mathfrak A(4)\subseteq \rm{SO}_3$; on the other hand, the elements of $\mathrm{SL}_2(\mathbb F_3)$ which are diagonalisable over $\mathbb F_9$ — or, equivalently, $\pm I_2$ and the matrices whose trace is zero — form a normal subgroup isomorphic to $\rm Q_8$ (and it's quite astonishing they do form a subgroup!) whose elements are $$\begin{array}{l} \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}, \begin{pmatrix} -1 & 0 \\ 0 & -1 \end{pmatrix}, \begin{pmatrix} 0 & 1 \\ -1 & 0 \end{pmatrix}, \begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix},\\ \begin{pmatrix} 1 & 1 \\ 1 & -1 \end{pmatrix}, \begin{pmatrix} 1 & -1 \\ -1 & -1 \end{pmatrix}, \begin{pmatrix} -1 & 1 \\ 1 & 1 \end{pmatrix}, \begin{pmatrix} -1 & -1 \\ -1 & 1 \end{pmatrix}. \end{array}$$

As I said earlier, this proof isn't very satisfying. Hence my question:

Is there an enlightening proof of the isomorphism $\mathbb T \simeq \rm{SL}_2(\mathbb F_3)$?

  • $\begingroup$ I suspect that the binary tetrahedral group is conjugate to a subgroup of $\text{SL}_2(\mathbb{Z})$. If so, it's known that if $G$ is a finite subgroup of $\text{GL}_n(\mathbb{Z})$ then $G$ injects into $\text{GL}_n(\mathbb{F}_3)$, and that provides a pretty straightforward isomorphism. $\endgroup$ Apr 24 '13 at 17:41
  • 2
    $\begingroup$ The isomorphism $\mathfrak A(4) \simeq \rm{PSL}_2(\mathbb F_3)$ is quite easy to see: the action of $\rm{PSL}_2(\mathbb F_3)$ on the 4-element projective line $\rm P^1(\mathbb F_3)$ provides an injective morphism $\rm{PSL}_2(\mathbb F_3) \to \mathfrak S(4)$ which gives easily the desired isomorphism. What is unclear to me is why the double cover gives precisely $\rm{SL}_2(\mathbb F_3)$. Maybe there are direct cohomological arguments? $\endgroup$
    – PseudoNeo
    Apr 24 '13 at 19:35
  • 2
    $\begingroup$ @QiaochuYuan: I doubt it: finite subgroups of $\rm{GL}_2(\mathbb Z)$ preserve a scalar product, so they are cyclic or dihedral. $\endgroup$
    – PseudoNeo
    Apr 24 '13 at 19:38
  • 3
    $\begingroup$ Well, this is probably not as geometric as you want, but $SL(2,3)$ is the only nontrivial "double cover" (central extension of order 2) over $PSL(2,3)$. $\endgroup$
    – user641
    Apr 24 '13 at 23:24
  • 1
    $\begingroup$ There is a 2-dimensional complex representation of $Q_8$, but it cannot be written over the rationals, so $Q_8$ does not embed in ${\rm GL}_2({\mathbb Z})$. (But ${\rm SL}_2(3)$, which contains $Q_8$, does embed in ${\rm SL}_2(p)$ for all odd primes $p$.) $\endgroup$
    – Derek Holt
    Apr 25 '13 at 8:59

Let's start with the ring $H$ of Hurwitz quaternions (quaternions s.t. either all coefficients are integers or all coefficients are half-integers). The ring $H\otimes\mathbb F_p$ is a quaternion algebra over a finite field, i.e. a matrix algebra $Mat_{2\times 2}(\mathbb F_p)$ (with determinant as the norm).

Now the binary tetrahedral group is the group of norm $1$ elements in $H$ and it maps injectively (for $p>2$) to the group of norm $1$ elements in $H\otimes\mathbb F_p$, i.e. to $SL_2(\mathbb F_p)$. In particular, for $p=3$ we get an isomorphism $\mathbb T\to SL_2(\mathbb F_3)$.

(Ref.: TWF 198.)


We know that $\mathbb T= <x,y,z| x^2=y^3=z^3=xyz>$. We also know that $ord(xyz)=2$, $\mathbb{T}/<xyz> \simeq A_4$ and $\mathbb{T}$ has no subgroup of order 12.

In one of my problem sheets, we are intended to give a guided proof of the fact that $\mathbb T \simeq SL_2( \mathbb F_3)$ by showing that the map $\phi: \mathbb T \to SL_2(\mathbb F_3)$ that sends $$x \longrightarrow \left( \begin{array}{cc} 0 & 2 \\ 1 & 0 \end{array} \right), y \longrightarrow \left( \begin{array}{cc} 2 & 0 \\ 1 & 2 \end{array} \right) \text{ and } z \longrightarrow \left( \begin{array}{cc} 0 & 1 \\ 2 & 1 \end{array} \right)$$ extends to an isomorphism between $\mathbb T$ and $SL_2( \mathbb{F}_3)$. This map is clearly a homomorphism, because it is easy to see that the images of the generators respect the relations.

We can just count the elements of $SL_2(\mathbb F_3)$ to see that $|SL_2(\mathbb F_3)|= | \mathbb{T}|=24$, so the only thing that we still have to show is either that $\phi$ is injective or surjective. Anybody can think of an easy way to prove this? I think this might lead to an easy proof of the result, without even making use of the notion of semidirect products.


Maybe the confusion arrises from the fact that $\mathbb{T}$ has no subgroup of order $12$. Rather it has the quaternion group $Q=\{\pm1,\pm i,\pm j, \pm k\}$ as a subgroup of order $8$.

I think a good way to grasp this intuitively is to see it in the context of the well known action of $\mathbb{H}^1$ on the pure quaternions $\mathbb{H}_0\simeq\mathbb{R}^3$ by rotation via $q(p)=qp\overline{q}$, which gives a double cover of $\mathrm{SO}(3)$. In particular, $\mathbb{H}^1/\{\pm1\}\cong\mathrm{SO}(3)$.

We can use this to realize the action of $\mathbb{T}$ on a tetrahedron explicitly, as follows. Let $T\subset\mathbb{H}_0$ be the tetrahedron with vertices $$i+j+k,\quad i-j-k,\quad -i+j-k,\quad -i-j+k.$$ An element of $Q$ acts on this via rotation by $\pi$ about a line through the midpoints of opposite sides of $T$. An element of $\mathbb{T}-Q$ acts on this by rotation by $2\pi/3$ about a line through a vertex and the midpoint of its opposite face. Any pair of elements of $\mathbb{T}$ differing only by sign give the same action. Now, by passing from $Q$ to $\mathbb{T}$ we have tripled the size of the group, and included everything necessary to get the full tetrahedral group, but with redundancy. In particular, the kernel of the action of $\mathbb{T}$ on $T$ is $\{\pm1\}$. So $\mathbb{T}$ gives a double cover of the tetrahedral group.

As you mentioned, there's a straightforward hands-on way to see that the tetrahedral group $\mathfrak{U}(4)$ is isomorphic to $\mathrm{PSL}(2,\mathbb{F}_3)$. So if we pull back to the full group, we've got $\mathbb{T}\cong\mathrm{SL}_2(\mathbb{F}_3)$.

This way of thinking about it is not my own invention. It's from a preprint I have of the book Quaternion Algebras by John Voight, which is expected to go into print very soon and which I highly recommend.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.