betaFeedback

WorksMathematics

Decomposing super unitary representations using Berezin integration

Abstract

I introduce the notion of a family of super unitary representations depending upon odd parameters as well as the notion of a direct Berezin integral of such representations. I use this to decompose the (left-)regular representation of the super Heisenberg group and of a super Heisenberg-like group into smaller (irreducible) subrepresentations. I compare the obtained subrepresentations to the representations obtained by applying (super) geometric quantization of the coadjoint orbits of the corresponding groups. It turns out that for the super Heisenberg group we need coadjoint orbits with an odd symplectic form in order to describe all subrepresentations found in the decomposition and for the Heisenberg-like super group we even need orbits with a non-homogeneous symplectic form.

Introduction/overview

In this paper I present some kind of “experimental” mathematics in the sense that I perform experiments/computations and that I try to find a theory that best fits

Publishing history: an earlier version of this paper [10] was never published and [13] was (for reasons of length) only partially published as [Tuy19]. This new version “thus” is a rewrite of [10], incorporating the notion of a super unitary representation as introduced in [Tuy19] and replacing/improving the discussion of the Berezin-Fourier transform with material from [13] that was excluded in [Tuy19]. these experimental data. More precisely, I do two different experiments: decomposing the left-regular representation of a (super) Lie group into irreducible components and computing the representations obtained by (super) geometric quantization of coadjoint orbits. These experiments start with the standard Heisenberg group HH as template, simply because this is a well understood case/experiment/computation. I then look at the super Heisenberg group SHSH by declaring that two of the three coordinates of HH are odd and I simply imitate/copy the computations done for the standard Heisenberg group. This can be done without any problem, except that in the ungraded case we use direct integrals of representations in the decomposition of HH (Fourier decomposition), which in the super case become Berezin-Fourier integrals. And I finish with a Heisenberg-like super Lie group GG where I imitate the kind of computations done for SHSH (the decomposition of the left-regular representations as well as for a comparison with representations obtained by geometric quantization of coadjoint orbits).

In imitating the decomposition of the Heisenberg group using Fourier transforms in the super context, these Fourier integrals become (partially) integrals over odd coordinates. This has two consequences for a theoretical framework: we have to adapt the underlying Hilbert spaces to odd coordinates and we have to show that, with the chosen definition of these underlying super Hilbert spaces, our Berezin-Fourier transforms are super unitary equivalences. In § 2-§ 4 I provide a suitable theoretical framework that fits all computations presented in this paper.

For the second set of “experiments” — the computation of the representations associated to coadjoint orbits and the link with the irreducible representations found in the decomposition of the left-regular representation—the situation is less straightforward. Not all representations obtained in the decomposition of the regular representation can be linked to coadjoint orbits with an even symplectic form. To obtain a satisfactory link between coadjoint orbits and the irreducible representations in the decomposition of the regular representation, it turns out that for the super Heisenberg group SHSH we need coadjoint orbits with an odd symplectic form and for the Heisenberg-like super group GG we even need coadjoint orbits with a non-homogeneous symplectic form as described in [11].1

When performing the computations for geometric quantization of the coadjoint orbits, I need (a generalisation of) the notion of a polarization. And here appear the first differences with the non-super case: even though the total dimension of a polarization is constant (at least in these examples), its graded dimension is not. Moreover, the representations associated to a same orbit but with polarizations of different graded dimensions are not equivalent. Now this may seem disconcerting, but it fits nicely with computations for the super Heisenberg group SHSH. When decomposing the classical Heisenberg group HH using the Fourier transform, there are two natural choices, choices that are described by physicists as the position representation and the momentum representation respectively. In terms of polarizations this corresponds to the vertical and horizontal polarization. And of course, in the classical case these two representations are equivalent. But for the super Heisenberg group the horizontal and vertical polarizations have different graded dimensions, yielding (obviously) non-equivalent representations. This fits nicely with the two natural choices in the decomposition: the two natural Fourier transforms for HH (which yield the position and momentum representations) become different for SHSH, as one will be done with an even coordinate, whereas the other has to be done with an odd coordinate. We thus have two essentially different ways to decompose SHSH, but these different ways correspond to different choices for a polarization on the same coadjoint orbit.

Contrary to the situation for the decomposition, I only have a partial theoretical framework (given in § 5) for the computation of a representation associated to a super coadjoint orbit via super geometric quantization. I have a framework for super prequantization, the notion of a super polarization and a way to associate function spaces to these notions on which the (super) group acts. But I do not have a way to generalise the notion of half-densities or half-forms that would allow me to define a (natural) structure of a super Hilbert space on these function spaces. Moreover, the geometric quantization procedure suggests (strongly) to introduce an odd Berezin-Fourier transform and this Berezin-Fourier transform introduces an odd (Fourier) parameter κ\kappa. And although it nicely gives us representations that correspond to representations obtained in the decomposition of the regular representation, it also poses a problem when comparing to p∣qp|q-parameter families of representations that appear in such a decomposition: all odd parameters (all qq of them) have to be real multiples of this same κ\kappa.

Let me finish this introduction with the remark that I will work with the geometric H∞H^\infty version of DeWitt supermanifolds, which is equivalent to the theory of graded manifolds of Leites and Kostant (see [2], [5], [6], [7], [9]). Any reader using a (slightly) different version of supermanifolds should be able to translate the results to her/his version of supermanifolds. Readers unfamiliar with my approach will find an extremely succinct overview of A\mathcal{A}-manifold theory in the appendix.

Super Hilbert spaces

Let me start very naively: a representation of a Lie group GG on a vector space EE is a homomorphism ρ:G→Aut⁡(E)\rho: G \to\operatorname{Aut}(E) and two representations of GG on E1E_1 and E2E_2 are equivalent if there exists an isomorphism (linear bijection) Φ:E1→E2\Phi: E_1 \to E_2 that intertwines the two representations. If the vector space EE is equipped with an inner product, we can define the notion of a unitary representation as a representation that “preserves” the inner product. And then two unitary representations are equivalent if there exists a unitary isomorphism ϕ:E1→E2\phi: E_1 \to E_2 that intertwines the representations.

When we transcribe this to the super setting (more or less putting everywhere the adjective super/graded), we arrive at the description that a (super) representation of an A\mathcal{A}-Lie group GG on a graded vector space EE is a homomorphism ρ:G→Aut⁡(E)\rho: G \to\operatorname{Aut}(E). For a super unitary representation we then will need a super scalar product that should be preserved by the representation, and two super unitary representations are equivalent when there exists a linear bijection ϕ:E1→E2\phi: E_1 \to E_2 that intertwines the representations and is compatible with the super scalar products.

When we remain in the realm of finite dimensional (graded) vector spaces, this transcription does not pose any problems. However, when going to infinite dimensional (graded) vector spaces, the situation becomes less clear. For (ordinary) unitary representations, one requires that the vector space be a Hilbert space, i.e., complete with respect to the topology induced by the inner product. Unfortunately, a super scalar product is not positive definite and there is no natural way to define a topology associated to a super scalar product, hence no way to define a notion of completeness. So we have to ask ourselves the question: is this a problem? Or, changing our point of view, why does one want a Hilbert space in the ungraded case? One (possible) answer is quite obvious: to restrict the number of inequivalent unitary representations. When we allow non-complete vector spaces, any invariant dense subspace of a representation space generates a new unitary representation. And on the other hand, any unitary representation on a non-complete vector space induces a unique unitary representation on its completion. In short: restricting attention to Hilbert spaces does not make an essential difference and reduces the number of inequivalent unitary representations.

The preceding discussion (strongly?) suggests that we should introduce a topology and/or an inner product on a representation space for a super unitary representation in order to achieve a similar reduction of super unitary representations. Once we admit that there should be two structures on a graded vector space in order to be a representation space for a super unitary representation: a super scalar product S\mathcal{S} for the notion of super-unitary and an inner product ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle to define a topology and the notion of completeness, we should specify if and how these two structures are related. The definition usually adopted requires a direct link between the super scalar product S\mathcal{S} and the inner product ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle. In [Tuy19] I argued that in this approach for most A\mathcal{A}-Lie groups the (left) regular representation will not be super unitary. To circumvent this restriction I weakened this condition by requiring only that S\mathcal{S} be continuous with respect to the topology induced by ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle. In [3.5] I will give another reason to weaken the link between the super scalar product and the inner product by showing that when there is a fixed (algebraic) relation between the super scalar product S\mathcal{S} and the inner product ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle, then a simple change of coordinates on an underlying super manifold will yield inequivalent super unitary representations. Having defined a super inner product space as a graded vector space equipped with two structures S\mathcal{S} and ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle, we can define a super Hilbert space as a super inner product space that is complete with respect to the inner product ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle. This gives us the following definitions.

Definitions: super inner product spaces

Let K\mathcal{K} denote any of the four rings/fields R\mathbb{R}, C\mathbb{C}, A\mathcal{A} or AC\mathcal{A}^{\mathbb{C}} and let VV be a graded vector space over K\mathcal{K}.

  • I will reserve the name inner product for the ungraded notion of a positive definite symmetric sesquilinear form on a real or complex vector space (with or without grading).2

  • A super scalar product S\mathcal{S} on a graded vector space VV (over K\mathcal{K}) is a smooth graded symmetric non-degenerate graded right-sesquilinear form on VV, i.e., a map S\mathcal{S} defined on V×VV\times V and with values in K\mathcal{K} satisfying the conditions (i) S(v1+v2λ,w)=S(v1,w)+S(v2,λw)\mathcal{S}(v_1+v_2\lambda,w)=\mathcal{S}(v_1,w)+\mathcal{S}(v_2,\lambda w) (right-anti-linear in the first entry); (ii) S(v,w1+w2λ)=S(v,w1)+S(v,w2)⋅λ\mathcal{S}(v,w_1+w_2\lambda)=\mathcal{S}(v,w_1)+\mathcal{S}(v,w_2)\cdot\lambda (right-linear in the second entry); (iii) S(w,v)=(−1)∣w∣∣v∣S(v,w)‾\mathcal{S}(w,v)=(-1)^{|w||v|}\overline{\mathcal{S}(v,w)} (graded symmetry); (iv) ∀w≠0 ∃v:S(v,w)≠0\forall w\ne0\ \exists v:\mathcal{S}(v,w)\ne0 (non-degeneracy); (v) S(BV,BV)⊂BK\mathcal{S}(BV,BV)\subset B\mathcal{K} (smoothness);

    Note that we explicitly do not require S\mathcal{S} to be even nor homogeneous (its “parity” is “concentrated” in the symbol S\mathcal{S}).

  • A right-linear map Φ:V→V\Phi:V\to V is said to be graded skew-symmetric with respect to a super scalar product S\mathcal{S} on VV if we have (see (41) for the definition of C\mathcal{C})

S(Φv,w)+S(v,Φ0w)+S(C(v),Φ1w)=0,\mathcal{S}(\Phi v,w)+\mathcal{S}(v,\Phi_0w)+\mathcal{S}(\mathcal{C}(v),\Phi_1w)=0,

where Φ=Φ0+Φ1\Phi=\Phi_0+\Phi_1 is the decomposition of Φ\Phi into its homogeneous parts.

  • A super inner product space is a triple (E,⟨⋅,⋅⟩,S)(E,\langle\cdot,\cdot\rangle,\mathcal{S}) in which EE is a (complex) graded vector space, ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle is an inner product on EE (i.e., (E,⟨⋅,⋅⟩)(E,\langle\cdot,\cdot\rangle) is a hermitian space) and S\mathcal{S} is a super scalar product on EE such that (i) ⟨E0,E1⟩=0\langle E_0,E_1\rangle=0;

(ii) S\mathcal{S} is continuous with respect to the topology of (E,⟨⋅,⋅⟩)(E,\langle\cdot,\cdot\rangle).

The triple (E,⟨⋅,⋅⟩,S)(E,\langle\cdot,\cdot\rangle,\mathcal{S}) will be called a super Hilbert space when (E,⟨⋅,⋅⟩)(E,\langle\cdot,\cdot\rangle) is an ordinary Hilbert space, i.e., complete with respect to the inner product ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle.

  • Let (E,⟨⋅,⋅⟩E,SE)(E,\langle\cdot,\cdot\rangle_E,\mathcal{S}_E) and (F,⟨⋅,⋅⟩F,SF)(F,\langle\cdot,\cdot\rangle_F,\mathcal{S}_F) be two super inner product spaces and let kk be a non-zero real constant. We will say that these two spaces are kk-isomorphic3 if there exists a linear bijection Φ:E→F\Phi:E\to F satisfying the following conditions (where Φ=Φ0+Φ1\Phi=\Phi_0+\Phi_1 and x=x0+x1x=x_0+x_1 are the decompositions of Φ\Phi respectively xx into homogeneous parts).

IsoSH 1. The homogeneous parts of Φ\Phi do not interfere in the sense that ker⁡(Φ0)+ker⁡(Φ1)=E\ker(\Phi_0)+\ker(\Phi_1)=E (and thus E=ker⁡(Φ0)⊕ker⁡(Φ1)E=\ker(\Phi_0)\oplus\ker(\Phi_1)).

IsoSH 2. Φ\Phi intertwines the inner products: ∀x,y∈E: ⟨Φx,Φy⟩F=⟨x,y⟩E\forall x,y\in E:\ \langle\Phi x,\Phi y\rangle_F=\langle x,y\rangle_E.

IsoSH 3. Φ\Phi intertwines the super scalar products up to the constant kk in the sense that for all x,y∈Ex,y\in E we have

k⋅SF(Φx,Φy)={SE(x,y)x∈E, y∈ker⁡(Φ1)SE(Cx,y)x∈ker⁡(Φ1), y∈ker⁡(Φ0)iSE(Cx,y)x,y∈ker⁡(Φ0).k\cdot\mathcal{S}_F(\Phi x,\Phi y)= \begin{cases} \mathcal{S}_E(x,y) & x\in E,\ y\in\ker(\Phi_1)\\ \mathcal{S}_E(\mathcal{C}x,y) & x\in\ker(\Phi_1),\ y\in\ker(\Phi_0)\\ i\mathcal{S}_E(\mathcal{C}x,y) & x,y\in\ker(\Phi_0). \end{cases}

Alternatively we will say that a map Φ\Phi satisfying these conditions is a kk-isomorphism between the two super inner product spaces. In the case k=1k=1 we will simply say that Φ\Phi is an isomorphism instead of a 1-isomorphism and that EE and FF are isomorphic instead of 1-isomorphic.

Remark. As (by definition) any graded vector space VV over A\mathcal{A} respectively AC\mathcal{A}^{\mathbb{C}} is of the form V=E⊗AV=E\otimes\mathcal{A} (resp. E⊗ACE\otimes\mathcal{A}^{\mathbb{C}}) for a graded vector space EE over R\mathbb{R} (resp. C\mathbb{C}), the smoothness condition for a super scalar product implies that restriction to E≅E⊗{1}⊂VE\cong E\otimes\{1\}\subset V establishes a bijection between super scalar products on VV and super scalar products on EE.

Lemma

(i) Let (H,⟨⋅,⋅⟩H)(\mathcal{H},\langle\cdot,\cdot\rangle_{\mathcal{H}}) be the completion of (E,⟨⋅,⋅⟩E)(E,\langle\cdot,\cdot\rangle_E). If the extension SH\mathcal{S}_{\mathcal{H}} of SE\mathcal{S}_E to H\mathcal{H} by continuity is non-degenerate, then (H,⟨⋅,⋅⟩H,SH)(\mathcal{H},\langle\cdot,\cdot\rangle_{\mathcal{H}},\mathcal{S}_{\mathcal{H}}) is a super Hilbert space.

(ii) If F⊂EF\subset E is a dense subspace, then (F,⟨⋅,⋅⟩E∣F,SE∣F)(F,\langle\cdot,\cdot\rangle_{E|F},\mathcal{S}_{E|F}) is a super inner product space.

Remark. Let (E,⟨⋅,⋅⟩,S)(E,\langle\cdot,\cdot\rangle,\mathcal{S}) be a super inner product space. In view of [2.2], it might be better to replace the condition that S\mathcal{S} should be continuous with respect to the topology of (E,⟨⋅,⋅⟩)(E,\langle\cdot,\cdot\rangle) by the condition that there exist two constants c,C>0c,C>0 such that

∀x∈E∖{0}:c≤sup⁡y∈E∖{0}∣S(x,y)∣∥x∥⋅∥y∥≤C.\forall x\in E\setminus\{0\}: c\leq\sup_{y\in E\setminus\{0\}}\frac{|\mathcal{S}(x,y)|}{\lVert x\rVert\cdot\lVert y\rVert}\leq C.

References

  1. [1][CCTV06] C. Carmeli, G. Cassinelli, A. Toigo, and V.S. Varadarajan, Unitary representations of super Lie groups and applications to the classification and multiplet structure of super particles, Commun. Math. Phys. 263 (2006), 217–258, arXiv:hep-th/0501061.DOI
  2. [2]B. DeWitt, Supermanifolds, Cambridge UP, Cambridge, 1984.
  3. [3]Axel de Goursac and Jean-Philippe Michel, Superunitary representations of Heisenberg supergroups, International Mathematics Research Notices 2020 (2018), no. 19, 5926–6006, doi.org/10.1093/imrn/rny184, arXiv:1601.07387 [math.RT].
  4. [5]B. Kostant, Graded manifolds, graded Lie theory, and prequantization, Differential geometric methods in mathematical physics (K. Bleuler & A. Reetz, ed.), LNM 570, Springer-Verlag, Berlin, 1977, Proceedings Conference, Bonn 1975, pp. 177–306.DOI
  5. [6]D.A. Leites, Introduction to the theory of supermanifolds, Russian Math. Surveys 35 (1980), 1–64.DOI
  6. [7]Alice Rogers, Supermanifolds: Theory and applications, World Scientific, Singapore, 2007.
  7. [8]Michael Reed and Barry Simon, Methods of modern mathematical physics I: Functional analysis, Academic Press, Inc., Boston, 1980.
  8. [9]Gijs M. Tuynman, Supermanifolds and supergroups: Basic theory, Mathematics and Its Applications 570, Kluwer Academic Publishers, Dordrecht, 2004.DOI
  9. [10]———, Geometric quantization of superorbits: a case study, Preprint, UFR de Mathématiques, Université de Lille I, Lille, 2009, arXiv:0901.1811 [math-ph].arxiv.org/abs/0901.1811
  10. [11]———, Super symplectic geometry and prequantization, J. Geom. Phys. 60 (2010), 1919–1939, arXiv:math-ph/0306049.
  11. [12]———, Integrating infinitesimal (super) actions, J. of Lie Theory 26 (2016), 297–358, arXiv:1302.2823 [math.DG].arxiv.org/abs/1302.2823
  12. [13]———, Super unitary representations revisited, Preprint, UFR de Mathématiques, Université de Lille I, Lille, 2017, arXiv:1711.00233 [math.DG].arxiv.org/abs/1711.00233

Paper details

vlnc:W_jdq7exh3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents