Skip to main content
Proceedings of the National Academy of Sciences of the United States of America logoLink to Proceedings of the National Academy of Sciences of the United States of America
. 2003 Oct 20;100(22):12559–12560. doi: 10.1073/pnas.1735527100

Signature quantization

Victor Guillemin †, Shlomo Sternberg ‡,§, Jonathan Weitsman
PMCID: PMC398041  PMID: 14569022

Abstract

We associate to the action of a compact Lie group G on a line bundle over a compact oriented even-dimensional manifold a virtual representation of G using a twisted version of the signature operator. We obtain analogues of various theorems in the more standard theory of geometric quantization.


Let M be a compact oriented manifold of dimension 2d. It has its DeRham complex

graphic file with name M1.gif

If we equip M with a Riemann metric we get a Hodge Inline graphic operator coming from the metric and orientation:

graphic file with name 12559_eqa.jpg

and a δ operator

graphic file with name 12559_eqb.jpg

If we modify Inline graphic by defining

graphic file with name 12559_eqc.jpg

then

graphic file with name M2.gif

and

graphic file with name M3.gif

Hence if we define Ω± to be the ±1 eigenspaces of Inline graphic acting on Inline graphic and let D denote the restriction of d + δ to Ω+, then

graphic file with name M6.gif [1]

This is the definition of the signature operator, and its index is the signature of the manifold M (1–3).

Let Inline graphic be a Hermitian line bundle with compatible connection ▿. This induces a covariant differential

graphic file with name M8.gif

with the defining property

graphic file with name M9.gif

We then also get

graphic file with name 12559_eqd.jpg

and therefore a twisted signature operator (4, 5)

graphic file with name M10.gif

If G is a compact Lie group acting as bundle automorphisms of Inline graphic, we can choose the metric on M and the connection to be G-invariant. This gives representations of G on Inline graphic that are intertwined by Inline graphic. Let Q(M) denote the virtual vector space

graphic file with name M14.gif [2]

This virtual vector space together with the virtual representation of G on it we call the signature quantization of the action of G on (M, Inline graphic). Its definition depends on the choices made: the metric and connection. However, it is clear that up to isomorphism Q(M, Inline graphic) and the virtual representation of G on it are independent of these choices.

The aim of this note is to describe the analogues for signature quantization of a number of theorems in the more standard theory of Inline graphic Dirac operator quantization.

The Generalized Kostant Formula

Let Inline graphic be the standard n-torus. Suppose that the action of G on M has isolated fixed points. For any weight μ of G and any virtual representation W of G let

graphic file with name M19.gif

denote the multiplicity with which μ occurs in W. [It is understood that # (μ, –W) = – #(μ, W).] The Kostant formula for signature quantization asserts that

graphic file with name M20.gif [3]

In this formula the Inline graphic are weighted versions of the Kostant partition function associated with the “polarized weights” of the isotropy representation of G on TMp. These are defined as follows. Because TMp does not have a complex structure, the weights

graphic file with name M22.gif

of the isotropy representation are only determined up to sign. Because p is isolated, none of these weights are zero. So by fixing a Inline graphic on which none of them vanish for any p, we can arrange that for all p and i we have Inline graphic. Then Inline graphic is defined to be the weighted number of solutions of the equation.

graphic file with name M26.gif [4]

where one assigns the weight 2d to a solution if all the ki are positive, the weight 2d–1 if exactly one of the ki is zero, etc. The sign (–1)p occurring in Eq. 3 is defined as follows. The choice of a polarization of the weights of the isotropy representation of G on TMp gives a complex structure on TMp and, hence, an orientation. Then (–1)p is defined to be +1 if this orientation coincides with the orientation coming from that of M and –1 otherwise. Finally νp is the weight of the isotropy representation of G on Inline graphic.

Notice that there are no “ρ shifts” in Eq. 3 in contrast to the usual Kostant formula (see ref. 6).

The Bott–Borel–Weil Theorem

Let G be a compact, simply connected semisimple Lie group, T its maximal torus, and μ a weight of T, which then determines a line bundle Inline graphic where M:= G/T. The standard Bott–Borel–Weil theorem says that if μ is in the positive Weyl chamber, the Inline graphic quantization of (M, Inline graphic) is the irreducible representation Irr(μ) with maximal weight μ. For signature quantization, we get the following version of this theorem.

Proposition 3.1. If μ is in the interior of the positive Weyl chamber then

graphic file with name M31.gif [5]

where ρ is one-half the sum of the positive roots of G, so that Irr(ρ) is the spin representation associated to the orthogonal representation of G on Inline graphic.

Note that our convention for the signature operator differs from that of ref. 7. The choice of conventions as in ref. 7 would eliminate the sign in Proposition 3.1.

The Khovanskii Theorem

Let G be a d-dimensional torus, M = M2d a nonsingular toric variety, Inline graphic its moment polytope, and Inline graphic its weight lattice. For the Inline graphic quantization, dim Q(M) is given by the number of points in Inline graphic. The Khovanskii theorem (see refs. 8 and 9) says that one can compute this number by applying a certain differential operator to the volume of the expanded polytope Δh; see ref. 10 for a formulation of this theorem. For signature quantization, one has an analogous theorem where one uses a weighted count, weighting the interior points by 2d, the points on the codimension one faces by 2d–1, etc. More specifically, suppose the polytope Δ is given by Inline graphic, where the half-spaces Hi are defined by Inline graphic where the Inline graphic are primitive lattice vectors and Inline graphic. For any N-tuple h:= (h1,..., hN) of real numbers, the expanded polytope Δh is then given by Inline graphic where Inline graphic. Define the series L(x) by the power series expansion

graphic file with name M43.gif [6]

where bi denotes the ith Bernoulli number. Let #′(Δ) denote the number of points in Inline graphic counted with weights as above. Then using the Hirzebruch formula for the signature (1) as generalized by Atiyah and Singer (ref. 4; see also ref. 5), we obtain

graphic file with name M45.gif

where L(∂/∂hi) is the infinite-order, constant coefficient differential operator defined by the series Eq. 6. In fact, up to an overall factor of 2d, this is the content of the Euler–Maclaurin formula in refs. 11 and 12 in the case of regular polytopes when applied to the constant function one; the methods of refs. 11 and 12 are purely combinatorial and give formulas for the sum of the values of any smooth function on a lattice points in a simple polytope. In this context the use of L-classes (rather than the Todd classes that appear in the Khovanskii formula) was a key idea of Cappell and Shaneson (13).

The Relation Between Quantization and Reduction

The action τ of the group G on Inline graphic and the connection ▿ give rise to an “abstract moment map” Inline graphic satisfying, for Inline graphic and s Inline graphic,

graphic file with name M50.gif

Moreover, if G is an n-dimensional torus, and MG is finite, there is a cobordism of “G-spaces with abstract moment maps”

graphic file with name 12559_eq7.jpg [7]

See ref. 15 for these notions and a proof of the above formula. Using the above cobordism, the Khovanskii theorem, and the Kostant formula for Inline graphic quantization, one can prove the conjecture “quantization commutes with reduction” as formulated in ref. 14 (see Chapter 8 of ref. 15 for details). The proof in the Inline graphic setting is complicated by the presence of ρ shifts in the Kostant formula. The Kostant formula for signature quantization does not involve these shifts, so the proof of the corresponding result is simpler, but there will be an overall factor of a power of 2 in its formulation, which is as follows: Let, Inline graphic, and let Mμ denote the reduced space of M at a regular value μ of Φ. Then

graphic file with name M54.gif [8]

Thus for signature quantization, the relation between quantization and reduction is more subtle than for Inline graphic quantization.

Additivity Under Cuts

Consider now the case where G = S1, and let Inline graphic be the moment map. Suppose that G acts freely on Φ–1(0). Then the spaces Φ–1([0, ∞)) and Φ–1((–∞, 0]) are compact manifolds with boundary, with a free S1 action on the boundaries. By collapsing the orbits on the boundaries to points, we obtain compact manifolds without boundary that we denote by M+ and M–. From the line bundle Inline graphic and the G action on M, we obtain line bundles on M± equipped with S1 actions. Moreover, the orientation of M induces orientations on M±. The quantizations of M, M+, and M– are then related by

graphic file with name M58.gif

We believe that this identity can be considered as a “characterizing property” of signature quantization.

Acknowledgments

V.G. was supported in part by a National Science Foundation grant. J.W. was supported in part by National Science Foundation Grant DMS 99/71914.

References

  • 1.Hirzebruch, F. (1966) Topological Methods in Algebraic Geometry (Springer, New York); trans. Schwarzenberger, R. L. E. (1956) Neue Topologische Methoden in der Algebraischen Geometrie (Springer, Berlin) (German).
  • 2.Atiyah, M. F. & Singer, I. M. (1968) Ann. Math. 87 546–604. [Google Scholar]
  • 3.Atiyah, M. & Bott, R. (1968) Ann. Math. 88 451–491. [Google Scholar]
  • 4.Atiyah, M. F. & Singer, I. M. (1963) Bull. Am. Math. Soc. 69 422–433. [Google Scholar]
  • 5.Berline, N., Getzler, E. & Vergne, M. (1992) Heat Kernels and Dirac Operators: Fundamental Principles of Mathematical Sciences (Springer, Berlin), p. 298.
  • 6.Guillemin, V., Lerman, E. & Sternberg, S. (1996) Symplectic Fibrations and Multiplicity Diagrams (Cambridge Univ. Press, Cambridge, U.K.).
  • 7.Lawson, H. B. & Michaelson, M. L. (1989) Spin Geometry (Princeton Univ. Press, Princeton).
  • 8.Kantor, J.M. & Khovanskii, A. (1992) Integral Points in Convex Polyhedra, Combinatorial Riemann–Roch Theorem and Generalized Mac-Laurin Formula, preprint IHES/M/932/37.
  • 9.Khovanskii, A. & Pukhlikov, S. (1992) Algebra i Analiz 4 188–216; trans. (1993) St. Petersburg Math. J. 4, 789–812 (Russian). [Google Scholar]
  • 10.Guillemin, V. (1994) Moment Maps and Combinatorial Invariants of Hamiltonian Tn-spaces: Progress in Mathematics (Birkhauser, Boston), p. 122.
  • 11.Karshon, Y., Sternberg, S. & Weitsman, J. (2003) Proc. Natl. Acad. Sci. USA 100 426–433. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Karshon, Y., Sternberg, S. & Weitsman, J. (2003) Euler-Maclaurin with Remainder for a Simple Integral Polytope, preprint math.CO/0307125. [DOI] [PMC free article] [PubMed]
  • 13.Cappell, S. E. & Shaneson, J. L. (1995) C. R. Acad. Sci. Ser. I 321 885–890. [Google Scholar]
  • 14.Guillemin, V & Sternberg, S. (1982) Invent. Math. 67 515–538. [Google Scholar]
  • 15.Ginzburg, V., Guillemin, V. & Karshon, Y. (2002) Moment Maps, Cobordisms, and Hamiltonian Group Actions: Mathematical Surveys and Monographs, Appendix J by Braverman, M. (Am. Math. Soc., Providence, RI), p. 98.

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences

RESOURCES