whatis - american mathematical society โ€บ journals โ€บ notices โ€บ 201611 โ€บ rnoti-p1252.pdfthe...

3
THE GRADUATE STUDENT SECTION WHAT IS... Symplectic Geometry Tara S. Holm Communicated by Cesar E. Silva Symplectic structures are ๏ฌ‚oppier than holomorphic functions or metrics. In Euclidean geome- try in a vector space over โ„, lengths and angles are the funda- mental measurements, and objects are rigid. In symplectic geome- try, a two-dimensional area measurement is the key ingredient, and the complex numbers are the natural scalars. It turns out that sym- plectic structures are much ๏ฌ‚oppier than holomorphic functions in complex geometry or metrics in Riemannian geometry. The word โ€œsymplecticโ€ is a calque introduced by Hermann Weyl in his textbook on the classical groups. That is, it is a root-by-root translation of the word โ€œcomplexโ€ from the Latin roots com โ€“ plexus, meaning โ€œtogether โ€“ braided,โ€ into the Greek roots with the same meaning, โ€“ รณ. Weyl suggested this word to describe the Lie group that preserves a nondegenerate skew-symmetric bilinear form. Prior to this, that Lie group was called the โ€œline complex groupโ€ or the โ€œAbelian linear groupโ€ (after Abel, who studied the group). A di๏ฌ€erential 2-form on (real) manifold is a gadget that at any point โˆˆ eats two tangent vectors and spits out a real number in a skew-symmetric, bilinear way. That is, it gives a family of skew-symmetric bilinear functions โˆถ ร— โ†’โ„ Tara S. Holm is professor of mathematics at Cornell University and Notices consultant. She is grateful for the support of the Si- mons Foundation. Her e-mail address is [email protected]. For permission to reprint this article, please contact: [email protected]. DOI: http://dx.doi.org/10.1090/noti1450 depending smoothly on the point โˆˆ. A 2-form โˆˆฮฉ 2 () is symplectic if it is both closed (its exterior derivative satis๏ฌes = 0) and nondegenerate (each function is nondegenerate). Nondegeneracy is equiva- lent to the statement that for each nonzero tangent vector โˆˆ , there is a symplectic buddy: a vector โˆˆ so that (,) = 1.A symplectic manifold is a (real) manifold equipped with a symplectic form . Nondegeneracy has important consequences. Purely in terms of linear algebra, at any point โˆˆ we may choose a basis of that is compatible with , using a skew- symmetric analogue of the Gram-Schmidt procedure. We start by choosing any nonzero vector 1 and then ๏ฌnding a symplectic buddy 1 . These must be linearly independent by skew-symmetry. We then peel o๏ฌ€ the two-dimensional subspace that 1 and 1 span and continue recursively, eventually arriving at a basis 1 , 1 ,โ€ฆ, , , which contains an even number of basis vectors. So symplectic manifolds are even-dimensional. This also allows us to think of each tangent space as a complex vector space where each and span a complex coordinate subspace. Moreover, the top wedge power, โˆˆฮฉ 2 (), is nowhere-vanishing, since at each tangent space, ( 1 ,โ€ฆ, ) โ‰  0. In other words, is a volume form, and is necessarily orientable. Symplectic geometry enjoys connections to algebraic combinatorics, algebraic geometry, dynamics, mathemati- cal physics, and representation theory. The key examples underlying these connections include: (1) = 2 = โ„‚ 1 with (, ) = signed area of the parallelogram spanned by and ; (2) any Riemann surface as in Figure 1 with the area for as in (1); โ‹ฏ Figure 1. The area form on a Riemannian surface de๏ฌnes its symplectic geometry. 1252 Notices of the AMS Volume 63, Number 11

Upload: others

Post on 07-Jun-2020

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: WHATIS - American Mathematical Society โ€บ journals โ€บ notices โ€บ 201611 โ€บ rnoti-p1252.pdfTHE GRADUATE STUDENT SECTION (3) ๐‘€=โ„2๐‘‘with๐œ” ๐‘‘=โˆ‘๐‘‘ ๐‘–โˆง๐‘‘ ๐‘–;

THE GRADUATE STUDENT SECTION

WHAT IS. . .

Symplectic GeometryTara S. Holm

Communicated by Cesar E. Silva

Symplecticstructures arefloppier thanholomorphicfunctions or

metrics.

In Euclidean geome-try in a vector spaceover โ„, lengths andangles are the funda-mental measurements,and objects are rigid.In symplectic geome-try, a two-dimensionalarea measurement isthe key ingredient, andthe complex numbersare the natural scalars.It turns out that sym-

plectic structures are much floppier than holomorphicfunctions in complex geometry or metrics in Riemanniangeometry.

The word โ€œsymplecticโ€ is a calque introduced byHermann Weyl in his textbook on the classical groups.That is, it is a root-by-root translation of the wordโ€œcomplexโ€ from the Latin roots

com โ€“ plexus,meaning โ€œtogether โ€“ braided,โ€ into the Greek roots withthe same meaning,

๐œŽ๐œ๐œ‡ โ€“ ๐œ‹๐œ†๐œ€๐œ…๐œ๐œ„๐œ…รณ๐œ.Weyl suggested this word to describe the Lie group thatpreserves a nondegenerate skew-symmetric bilinear form.Prior to this, that Lie group was called the โ€œline complexgroupโ€ or the โ€œAbelian linear groupโ€ (after Abel, whostudied the group).

A differential 2-form๐œ” on (real) manifold๐‘€ is a gadgetthat at any point ๐‘ โˆˆ ๐‘€ eats two tangent vectors andspits out a real number in a skew-symmetric, bilinearway. That is, it gives a family of skew-symmetric bilinearfunctions

๐œ”๐‘ โˆถ ๐‘‡๐‘๐‘€ร—๐‘‡๐‘๐‘€ โ†’ โ„

Tara S. Holm is professor of mathematics at Cornell Universityand Notices consultant. She is grateful for the support of the Si-mons Foundation.Her e-mail address is [email protected].

For permission to reprint this article, please contact:[email protected]: http://dx.doi.org/10.1090/noti1450

depending smoothly on the point ๐‘ โˆˆ ๐‘€. A 2-form๐œ” โˆˆ ฮฉ2(๐‘€) is symplectic if it is both closed (its exteriorderivative satisfies ๐‘‘๐œ” = 0) and nondegenerate (eachfunction ๐œ”๐‘ is nondegenerate). Nondegeneracy is equiva-lent to the statement that for each nonzero tangent vector๐‘ฃ โˆˆ ๐‘‡๐‘๐‘€, there is a symplectic buddy: a vector ๐‘ค โˆˆ ๐‘‡๐‘๐‘€so that ๐œ”๐‘(๐‘ฃ,๐‘ค) = 1. A symplectic manifold is a (real)manifold ๐‘€ equipped with a symplectic form ๐œ”.

Nondegeneracy has important consequences. Purely interms of linear algebra, at any point ๐‘ โˆˆ ๐‘€wemay choosea basis of ๐‘‡๐‘๐‘€ that is compatible with ๐œ”๐‘, using a skew-symmetric analogue of the Gram-Schmidt procedure. Westart by choosing any nonzero vector๐‘ฃ1 and then finding asymplectic buddy๐‘ค1. These must be linearly independentby skew-symmetry. We then peel off the two-dimensionalsubspace that ๐‘ฃ1 and ๐‘ค1 span and continue recursively,eventually arriving at a basis

๐‘ฃ1,๐‘ค1,โ€ฆ ,๐‘ฃ๐‘‘,๐‘ค๐‘‘,which contains an even number of basis vectors. Sosymplectic manifolds are even-dimensional. This alsoallows us to think of each tangent space as a complexvector space where each ๐‘ฃ๐‘– and ๐‘ค๐‘– span a complexcoordinate subspace. Moreover, the top wedge power,๐œ”๐‘‘ โˆˆ ฮฉ2๐‘‘(๐‘€), is nowhere-vanishing, since at each tangentspace,

๐œ”๐‘‘(๐‘ฃ1,โ€ฆ ,๐‘ค๐‘‘) โ‰  0.In other words,๐œ”๐‘‘ is a volume form, and๐‘€ is necessarilyorientable.

Symplectic geometry enjoys connections to algebraiccombinatorics, algebraic geometry, dynamics, mathemati-cal physics, and representation theory. The key examplesunderlying these connections include:(1) ๐‘€ = ๐‘†2 = โ„‚๐‘ƒ1 with ๐œ”๐‘(๐‘ฃ,๐‘ค) = signed area of the

parallelogram spanned by ๐‘ฃ and ๐‘ค;(2) ๐‘€ any Riemann surface as in Figure 1 with the area

for ๐œ” as in (1);

โ‹ฏFigure 1. The area form on a Riemannian surfacedefines its symplectic geometry.

1252 Notices of the AMS Volume 63, Number 11

Page 2: WHATIS - American Mathematical Society โ€บ journals โ€บ notices โ€บ 201611 โ€บ rnoti-p1252.pdfTHE GRADUATE STUDENT SECTION (3) ๐‘€=โ„2๐‘‘with๐œ” ๐‘‘=โˆ‘๐‘‘ ๐‘–โˆง๐‘‘ ๐‘–;

THE GRADUATE STUDENT SECTION(3) ๐‘€ = โ„2๐‘‘ with ๐œ”๐‘ ๐‘ก๐‘‘ = โˆ‘๐‘‘๐‘ฅ๐‘– โˆง๐‘‘๐‘ฆ๐‘–;(4) ๐‘€ = ๐‘‡โˆ—๐‘‹, the cotangent bundle of any manifold

๐‘‹, thought of as a phase space, with ๐‘-coordinatescoming from ๐‘‹ being positions, ๐‘ž-coordinates inthe cotangent directions representing momentumdirections, and ๐œ” = โˆ‘๐‘‘๐‘๐‘– โˆง๐‘‘๐‘ž๐‘–;

(5) ๐‘€ any smooth complex projective variety with ๐œ”induced from the Fubini-Study form (this includessmooth normal toric varieties, for example);

(6) ๐‘€ = ๐’ช๐œ† a coadjoint orbit of a compact connectedsemisimple Lie group๐บ, equipped with the Kostant-Kirillov-Souriau form ๐œ”. For the group ๐บ = ๐‘†๐‘ˆ(๐‘›),this class of examples includes complex projectivespace โ„‚๐‘ƒ๐‘›โˆ’1, Grassmannians G ๐‘Ÿ๐‘˜(โ„‚๐‘›), the full flagvariety โ„ฑโ„“(โ„‚๐‘›), and all other partial flag varieties.

Plenty of orientable manifolds do not admit a symplecticstructure. For example, even-dimensional spheres that areat least four-dimensional are not symplectic. The reason isthat on a compact manifold, Stokesโ€™ theorem guaranteesthat [๐œ”] โ‰  0 โˆˆ ๐ป2(๐‘€;โ„). In other words, compactsymplectic manifolds must exhibit nonzero topology indegree 2 cohomology. The only sphere with this propertyis ๐‘†2.

Example (3) gains particular importance because of thenineteenth century work of Jean Gaston Darboux on thestructure of differential forms. A consequence of his workisDarbouxโ€™s Theorem. Let๐‘€ be a two-dimensional symplec-tic manifold with symplectic form ๐œ”. Then for every point๐‘ โˆˆ ๐‘€, there exists a coordinate chart ๐‘ˆ about ๐‘ withcoordinates ๐‘ฅ1,โ€ฆ , ๐‘ฅ๐‘‘, ๐‘ฆ1,โ€ฆ ,๐‘ฆ๐‘‘ so that on this chart,

๐œ” =๐‘‘โˆ‘๐‘–=1

๐‘‘๐‘ฅ๐‘– โˆง๐‘‘๐‘ฆ๐‘– = ๐œ”๐‘ ๐‘ก๐‘‘.

Tools from the1970s and 1980s set

the stage fordramatic progress

in symplecticgeometry and

topology.

This makes pre-cise the notion thatsymplectic geome-try is floppy. InRiemannian geome-try there are localinvariants such ascurvature that dis-tinguish metrics.Darbouxโ€™s theoremsays that symplecticforms are all lo-cally identical. Whatremains, then, areglobal topologicalquestions such as, What is the cohomology ring of aparticular symplectic manifold? and more subtle sym-plectic questions such as, How large can the Darbouxcharts be for a particular symplectic manifold?

Two tools developed in the 1970s and 1980s set thestage for dramatic progress in symplectic geometry andtopology. Marsden and Weinstein, Atiyah, and Guilleminand Sternberg established properties of the momentummap, resolving questions of the first type. Gromov intro-duced pseudoholomorphic curves to probe questions of

the second type. Let us briefly examine results of eachkind.

If a symplectic manifold exhibits a certain flavor ofsymmetries in the form of a Lie group action, then itadmits a momentum map. This gives rise to conservedquantities such as angular momentum, whence the name.The first example of a momentum map is the heightfunction on a 2-sphere (Figure 2).

Figure 2. The momentum map for ๐‘†1 acting on ๐‘†2 byrotation.

In this case, the conserved quantity is angular momen-tum, and the height function is the simplest example ofa perfect Morse function on ๐‘†2. When the Lie group is aproduct of circles ๐‘‡ = ๐‘†1 ร—โ‹ฏร—๐‘†1, we say that the mani-fold is a Hamiltonian ๐‘‡-space, and the momentum mapis denoted ฮฆ โˆถ ๐‘€ โ†’ โ„๐‘›. In 1982 Atiyah and independentlyGuillemin and Sternberg proved the Convexity Theorem(see Figure 3):Convexity Theorem. If ๐‘€ is a compact Hamiltonian ๐‘‡-space, then ฮฆ(๐‘€) is a convex polytope. It is the convex hullof the images ฮฆ(๐‘€๐‘‡) of the ๐‘‡-fixed points.

Figure 3. Atiyah and Guillemin-Sternberg proved thatif a symplectic manifold has certain symmetries, theimage of its momentum map is a convex polytope.

This provides a deep connection between symplecticand algebraic geometry on the one hand and discretegeometry and combinatorics on the other. Atiyahโ€™s proofdemonstrates that the momentum map provides Morsefunctions on ๐‘€ (in the sense of Bott), bringing the powerof differential topology to bear on global topologicalquestions about ๐‘€. Momentum maps are also used toconstruct symplectic quotients. Lisa Jeffrey will discussthe cohomology of symplectic quotients in her 2017Noether Lecture at the Joint Mathematics Meetings, andmanymore researchers will delve into these topics duringthe special session Jeffrey and I are organizing.

Through example (2), we see that two-dimensional sym-plectic geometry boils down to area-preserving geometry.Because symplectic forms induce volume forms, a naturalquestion in higher dimensions is whether symplectic ge-ometry is as floppy as volume-preserving geometry: can amanifold be stretched and squeezed in any which way so

December 2016 Notices of the AMS 1253

Page 3: WHATIS - American Mathematical Society โ€บ journals โ€บ notices โ€บ 201611 โ€บ rnoti-p1252.pdfTHE GRADUATE STUDENT SECTION (3) ๐‘€=โ„2๐‘‘with๐œ” ๐‘‘=โˆ‘๐‘‘ ๐‘–โˆง๐‘‘ ๐‘–;

THE GRADUATE STUDENT SECTION

Gromov provedthat symplecticmaps are more

rigid thanvolume-

preservingones.

long as volume is pre-served? Using pseudoholo-morphic curves, Gromovdismissed this possibility,proving that symplecticmaps are more rigid thanvolume-preserving ones. Inโ„2๐‘‘ we let ๐ต2๐‘‘(๐‘Ÿ) denotethe ball of radius ๐‘Ÿ. In1985 Gromov proved thenonsqueezing theorem (seeFigure 4):

Nonsqueezing Theorem.There is an embedding

๐ต2๐‘‘(๐‘…) โ†ช ๐ต2(๐‘Ÿ) ร— โ„2๐‘‘โˆ’2

preserving ๐œ”๐‘ ๐‘ก๐‘‘ if and only if ๐‘… โ‰ค ๐‘Ÿ.One direction is straightforward: if ๐‘… โ‰ค ๐‘Ÿ, then

๐ต2๐‘‘(๐‘…) โŠ† ๐ต2(๐‘Ÿ) ร— โ„2๐‘‘โˆ’2. To find an obstruction tothe existence of such a map, Gromov used a pseudo-holomorphic curve in ๐ต2(๐‘Ÿ) ร— โ„2๐‘‘โˆ’2 and the symplecticembedding to produce a minimal surface in ๐ต2๐‘‘(๐‘…),forcing ๐‘… โ‰ค ๐‘Ÿ.

On the other hand, a volume-preserving map existsfor any ๐‘Ÿ and ๐‘…. Colloquially, you cannot squeeze asymplectic camel through the eye of a needle.

Gromovโ€™s work has led to many rich theories ofsymplectic invariants with pseudoholomorphic curves

Figure 4. The Nonsqueezing Theorem gives geomet-ric meaning to the aphorism on the impossibility ofpassing a camel through the eye of a needle: A sym-plectic manifold cannot fit inside a space with a nar-row two-dimensional obstruction, no matter how bigthe target is in other dimensions.This cartoon originally appeared in the article โ€œTheSymplectic Camelโ€ by Ian Stewartโ€”published in theSeptember 1987 issue of Natureโ€”we thank him forhis permission to use it. Cosgrove is a well-knowncartoonist loosely affiliated with mathematicians. Hedid drawings for Manifold for a few years and mostrecently did a few sketches for the curious cookbookSimple Scoff:https://www2.warwick.ac.uk/newsandevents/pressreleases/50th_anniversary_cookery.

the common underlying tool. The constructions rely onsubtle arguments in complex analysis and Fredholmtheory. These techniques are essential to current workin symplectic topology and mirror symmetry, and theyprovide an important alternativeperspective on invariantsof four-dimensional manifolds.

Further details on momentum maps may be found in[CdS], and [McD-S] gives anaccountofpseudoholomorphiccurves.

References[CdS] A. Cannas da Silva, Lectures on Symplectic Geometry,

Lecture Notes in Mathematics, 1764, Springer-Verlag,Berlin, 2001. MR 1853077

[McD-S] D. McDuff and D. Salamon, Introduction to Symplec-tic Topology, Oxford Mathematical Monographs, OxfordUniversity Press, New York, 1995. MR 1373431

ABOUT THE AUTHOR

In addition to mathematics, Tara Holm enjoys gardening, cooking with her family, and exploring the Finger Lakes. This column was based in part on her AMSโ€“MAA Invited Address at MAA MathFest 2016 held in August.

Tara S. Holm

ABOUT THE AUTHOR

In addition to mathematics, Tara Holm enjoys gardening, cooking with her family, and exploring the Finger Lakes. This column was based in part on her AMSโ€“MAA Invited Ad-dress at MathFest 2016 held in August. Tara S. Holm

Ph

oto

by

Mel

issa

Totm

an.

A Decade Ago in the Notices: PseudoholomorphicCurves

โ€œWHAT ISโ€ฆSymplectic Geometry?โ€ discusses Gromovโ€™sโ€œnonsqueezing theorem,โ€ a key result in symplecticgeometry. Gromovproved the theoremusing thenotionof a pseudoholomorphic curve, which he introduced in1986.

Readers interested in these topics might also wishto read โ€œWHAT ISโ€ฆa Pseudoholomorphic Curve?โ€ bySimon Donaldson, which appeared in the October 2005issue of the Notices. โ€œThe notion [of pseudoholomor-phic curve] has transformed the field of symplectictopology and has a bearing on many other areas suchas algebraic geometry, string theory and 4-manifoldtheory,โ€ Donaldson writes. Starting with the basic no-tion of a plane curve, he gives a highly accessibleexplanation of what pseudoholomorphic curves are.He notes that they have been used as a tool in sym-plectic topology in two main ways: โ€œFirst, as geometricprobes to explore symplectic manifolds: for examplein Gromovโ€™s result, later extended by Taubes, on theuniqueness of the symplectic structure on the complexprojective planeโ€ฆSecond, as the source of numericalinvariants: Gromovโ€“Witten invariants.โ€

Another related piece โ€œWHAT ISโ€ฆa Toric Variety?โ€by Ezra Miller appeared in the May 2008 issue ofthe Notices.

1254 Notices of the AMS Volume 63, Number 11