1
$\begingroup$

$\DeclareMathOperator\Spin{Spin}\DeclareMathOperator\SU{SU}\DeclareMathOperator\U{U}\DeclareMathOperator\SL{SL}\DeclareMathOperator\SO{SO}$From Pierre Deligne's Notes on spinors, we can see that there is an injective group homomorphism (embedding): \begin{equation} \Spin(2n) \to \SU(2^{n-1}) \label1\tag1 \end{equation} thus the embedding $\Spin(2n) \subset \SU(2^{n-1})$ for some positive integer $n$. I think this seems to be true for $n \geq 5$ but may not be true for $n \leq 4$.

Thus, below let us focus on $n \geq 5$.

We know that when $n=2k$, the center $Z(\Spin(4k))=\mathbf{Z}/2 \oplus \mathbf{Z}/2$. When $n=2k+1$, the center $Z(\Spin(4k+2))=\mathbf{Z}/4$.

So it is natural to generalize \eqref{1} to have an injective group homomorphism (embedding), $$ \frac{\Spin(4k+2)\times \U(1)}{\mathbf{Z}/4} \to \frac{\SU(2^{2k})\times \U(1)}{\mathbf{Z}/4}. \label{1'}\tag{1'} $$ The right hand side is true as long as ${\mathbf{Z}/4}\subset Z(\SU(2^{2k}))={\mathbf{Z}/2^{2k}}$ in the subgroup of center.

My question:

  1. Do you agree with me that \eqref{1} and \eqref{1'} are correct for $n \geq 5$? where $4k+2 \geq 10$?

  2. I believe that there is an injective group homomorphism (embedding), $$ \frac{\Spin(4k+2)\times \U(1)}{\mathbf{Z}/4} \to \frac{\SU(2^{2k})\times \U(1)}{\mathbf{Z}/2^{2k}}=\U(2^{2k}), \label2\tag2 $$ because the irreducible spinorial representation of $\Spin(4k+2)$ is $2^{2k}$ dimensional, which is the same as the standard fundamental representation of $\U(2^{2k})$. Do you also agree with me on \eqref{2}?

  3. Is there an injective group homomorphism (embedding), $$ \frac{\Spin(4k+2)\times \U(1)}{\mathbf{Z}/2} \to \frac{\SU(2^{2k})\times \U(1)}{\mathbf{Z}/2^{2k}}=\U(2^{2k}). \label3\tag3 $$ Here the $\mathbf{Z}/2$ in the denominator is a normal subgroup of $\mathbf{Z}/4$ in \eqref{2}. The \eqref{2} and \eqref{3} are related by $$ \begin{array}{ccc} \frac{\Spin(4k+2)\times \U(1)}{\mathbf{Z}/2} & & \\ \downarrow &\searrow & \\ \frac{\Spin(4k+2)\times \U(1)}{\mathbf{Z}/4} & \longrightarrow & \U(2^{2k}) \end{array}. \tag{4} $$


Note for some background: The above we concern the $\Spin(2n)=\Spin(2n;\mathbb{R})$ acting on the vector space with the real $\mathbb{R}$ field. It is sufficient to study complex representations $\rho: \Spin(2n; {\mathbb C})\to \SL(2^{n-1}, {\mathbb C})$ or $\SL(2^{n}, {\mathbb C})$ of complex Spin groups $\Spin(2n, {\mathbb C})$: The restriction of such a representation to the compact Spin subgroup $\Spin(2n)$ is automatically unitarizable, i.e. the image is contained in a conjugate of $\SU(N)$ for appropriate $N$.

A representation $\rho$ of $\Spin(2n, \mathbb C)$ is called spinoral if it does not descend to the orthogonal group $\SO(2n, {\mathbb C})$ (equivalently, $\rho$ is injective). There are also half-spin or semi-spin representations: They are also spinoral. This is related to the above choice of $\SL(2^{n-1}, {\mathbb C})$ for irreducible semi-spin representation, and the above choice of $\SL(2^{n}, {\mathbb C})$ for reducible spin representation.

$\endgroup$
4
  • $\begingroup$ Why is $\rho$ injective if and only if it does not descend to $\operatorname{SO}(2n, \mathbb C)$? Do you mean to require only the central character of $\rho$ to be injective? $\endgroup$
    – LSpice
    Aug 21, 2021 at 17:41
  • $\begingroup$ I am not sure why you emphasize "descend to SO(2𝑛,ℂ)". Here I focus on to Spin(2𝑛)=Spin(2𝑛;ℝ). I put a Note p.s. on Spin(2𝑛,ℂ) just for the convenience of sufficient studying complex representations $\endgroup$ Aug 21, 2021 at 17:56
  • 1
    $\begingroup$ I emphasise "descend to $\operatorname{SO}(2n, \mathbb C)$" because that's where you made the claim: "[$\rho$] does not descend to the orthogonal group $\operatorname{SO}(2n, \mathbb C)$ (equivalently, $\rho$ is injective)". But I would be equally sceptical of the real version of the claim; I would definitely believe it on the level of central characters, but it's not clear to me on the level of the entire representation. $\endgroup$
    – LSpice
    Aug 21, 2021 at 18:18
  • $\begingroup$ YES, "A representation 𝜌 of Spin(2𝑛,ℂ) is called spinoral if it does not descend to the orthogonal group SO(2𝑛,ℂ) " --> so here I meant that this injective group homomorphism is pinned down by the spinor or spinoral representation of Spin(2𝑛,ℂ), no the vector representation of SO(2𝑛,ℂ). $\endgroup$ Aug 21, 2021 at 18:31

1 Answer 1

2
$\begingroup$

You really should have a look at F. Reese Harvey's book Spinors and Calibrations, where all of your questions are answered.

For example, your 'inclusion' (1) is not correct for sufficiently large $n$. In fact, here are the actual embeddings:

There is an embedding $\mathrm{Spin}(8n)\hookrightarrow\mathrm{SO}(2^{4n-1})\times\mathrm{SO}(2^{4n-1})$, but this latter product does not embed into $\mathrm{SU}(2^{4n-1})$, nor does $\mathrm{Spin}(8n)$. Each of the two projections into the $\mathrm{SO}(2^{4n-1})$ factors is a double cover of its image, which is a maximal proper subgroup $\mathrm{SO}'(8n)\subset\mathrm{SO}(2^{4n-1})$ that, for $n>1$, is not isomorphic to $\mathrm{SO}(8n)$. Under the above embedding, the center of $\mathrm{Spin}(4n)$ goes to the subgroup of order $4$ consisting of the elements $(\epsilon_1 I,\epsilon_2 I)\in \mathrm{SO}(2^{4n-1})\times\mathrm{SO}(2^{4n-1})$ where $\epsilon_i^2 = 1$. (The case $n=1$ is, of course, the famous triality isomorphism, $\mathrm{SO}'(8)\simeq \mathrm{SO}(8)$.)

There is an embedding $\mathrm{Spin}(8n{+}4)\hookrightarrow\mathrm{Sp}(2^{4n})\times\mathrm{Sp}(2^{4n})\subset \mathrm{SU}(2^{4n+1})\times\mathrm{SU}(2^{4n+1})$, but, for $n>0$, the projections of $\mathrm{Spin}(8n{+}4)$ into either $\mathrm{Sp}(2^{4n})$ factor is a double cover onto a maximal proper subgroup $\mathrm{SO}'(8n{+}4)\subset\mathrm{Sp}(2^{4n})$ that is not isomorphic to $\mathrm{SO}(8n{+}4)$. In particular, $\mathrm{Spin}(8n{+}4)\hookrightarrow\mathrm{SU}(2^{4n+1})$ is never an embedding. Under the above embedding, the center of $\mathrm{Spin}(8n{+}4)$ goes to the subgroup of order $4$ consisting of the elements $(\epsilon_1 I,\epsilon_2 I)\in \mathrm{Sp}(2^{4n})\times\mathrm{Sp}(2^{4n})$ where $\epsilon_i^2 = 1$. (The case $n=0$ is, of course, different, because, in this case, the embedding $\mathrm{Spin}(4)\hookrightarrow\mathrm{Sp}(1)\times \mathrm{Sp}(1)=\mathrm{SU}(2)\times\mathrm{SU}(2)$ is an isomorphism.)

Finally, for $n\ge 1$ there is an embedding $\mathrm{Spin}(4n{+}2)\hookrightarrow\mathrm{SU}(2^{2n})$. The image of this embedding is a maximal proper subgroup of $\mathrm{SU}(2^{2n})$, and, under this embedding, the center of $\mathrm{Spin}(4n{+}2)$ goes to the multiples of the identity by a power of $i\in\mathbb{C}$. (Of course, this fails for $n=0$.)

$\endgroup$
7
  • 1
    $\begingroup$ Thanks, but I also asked extra U(1) factor with appropriate mod structure. Could you illuminate on that too? $\endgroup$ Aug 21, 2021 at 19:19
  • $\begingroup$ thanks for the refs "F. Reese Harvey's book Spinors and Calibrations" I will read (son)! $\endgroup$ Aug 21, 2021 at 19:19
  • $\begingroup$ I also have this puzzle mathoverflow.net/q/402173/336737 whether you can shed light on it too? $\endgroup$ Aug 21, 2021 at 19:21
  • $\begingroup$ Hi Robert and Marina, I like your post. I find this answer may be partly supporting but also partly contradicting with part of the answer given here: math.stackexchange.com/a/3296484 $\endgroup$ Aug 21, 2021 at 21:54
  • 1
    $\begingroup$ @МаринаMarinaS: I don't really understand your question. When $k\ge 1$, the intersection of the subgroups $\mathrm{Spin}(4k{+}2)$ and the central subgroup $\mathrm{U}(1)\subset\mathrm{U}(2^{2k})$ is the set of $\lambda I$ where $\lambda^4 = 1$, so the first embedding works. However, I don't see how you can get the second map as an embedding. The $\mathrm{U}(1)$ subgroup in $\mathrm{U}(2^{2k})$ has to commute with the subgroup $\mathrm{Spin}(4k{+}2)$, which forces it to be central (i.e., multiples of the identity), and that subgroup intersects $\mathrm{Spin}(4k{+}2)$ in a group of order $4$. $\endgroup$ Aug 22, 2021 at 15:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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