ebook img

Gauge theory for fiber bundles PDF

106 Pages·1991·0.734 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Gauge theory for fiber bundles

GAUGE THEORY FOR FIBER BUNDLES Peter W(cid:1) Michor Mailingaddress(cid:1) Peter W(cid:2) Michor(cid:3) Institut fu(cid:4)r Mathematikder Universit(cid:4)at Wien(cid:3) Strudlhofgasse (cid:5)(cid:3) A(cid:6)(cid:7)(cid:8)(cid:9)(cid:8) Wien(cid:3) Austria(cid:2) E(cid:6)mailmichor(cid:10)awirap(cid:2)bitnet (cid:1)(cid:2)(cid:3)(cid:4)Mathematicssubjectclassi(cid:1)cation(cid:5)(cid:6)(cid:7)(cid:4)(cid:8)(cid:9)(cid:5)(cid:6)C(cid:4)(cid:5)(cid:9)(cid:5)(cid:6)C(cid:1)(cid:4)(cid:9)(cid:5)(cid:3)(cid:7)(cid:4)(cid:8)(cid:9)(cid:5)(cid:3)D(cid:4)(cid:5)(cid:9)(cid:5)(cid:3)A(cid:6)(cid:5) Extendedversionofaseries oflectures heldattheInstitute ofPhysics of the University of Napoli(cid:3)March (cid:11)(cid:12) (cid:13) April (cid:7)(cid:3) (cid:7)(cid:9)(cid:12)(cid:12) (cid:1) Table of contents Introduction (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:11) (cid:7)(cid:2) Notations and conventions (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:14) (cid:11)(cid:2) Calculus of smoothmappings (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:15) (cid:16)(cid:2) Calculus of holomorphicmappings (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:7)(cid:8) (cid:5)(cid:2) Calculus of real analytic mappings (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:7)(cid:11) (cid:14)(cid:2) In(cid:17)nite dimensionalmanifolds (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:7)(cid:5) (cid:15)(cid:2) Manifoldsof mappings (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:7)(cid:18) (cid:18)(cid:2) Di(cid:19)eomorphismgroups (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:11)(cid:11) (cid:12)(cid:2) The Fr(cid:4)olicher(cid:6)Nijenhuis Bracket (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:11)(cid:18) (cid:9)(cid:2) Fiber Bundles and Connections (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:16)(cid:15) (cid:7)(cid:8)(cid:2) Principal Fiber Bundles and G(cid:6)Bundles (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:5)(cid:5) (cid:7)(cid:7)(cid:2) Principal and Induced Connections (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:15)(cid:8) (cid:7)(cid:11)(cid:2) Holonomy (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:18)(cid:15) (cid:7)(cid:16)(cid:2) The nonlinear frame bundle of a (cid:17)ber bundle (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:12)(cid:7) (cid:7)(cid:5)(cid:2) Gauge theory for (cid:17)ber bundles (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:12)(cid:16) (cid:7)(cid:14)(cid:2) A classifying space for the di(cid:19)eomorphismgroup (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:12)(cid:15) (cid:7)(cid:15)(cid:2) A characteristic class (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:9)(cid:8) (cid:7)(cid:18)(cid:2) Self duality (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:9)(cid:12) References (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:7)(cid:8)(cid:7) List of Symbols (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:7)(cid:8)(cid:11) Index (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:2) (cid:7)(cid:8)(cid:16) (cid:8) Introduction Gauge theory usually investigates the space of principal connections on a principal (cid:17)ber bundle (cid:20)P(cid:1)p(cid:1)M(cid:1)G(cid:21) and its orbit space under the action of the gauge group (cid:20)called the modulispace(cid:21)(cid:3) which is the group of all principal bundle automorphisms of P which cover the identity on the base space M(cid:2) It is the arena for the Yang(cid:6)Mills(cid:6)Higgs equa(cid:6) tions which allows (cid:20)with structure group U(cid:20)(cid:7)(cid:21) SU(cid:20)(cid:11)(cid:21)(cid:21) a satisfactory (cid:1) uni(cid:17)ed description of electromagnetic and weak interactions(cid:3) which was developed by Glashow(cid:3) Salam(cid:3) and Weinberg(cid:2) This electro(cid:6)weak the(cid:6) orypredicted the existence ofmassivevector particles (cid:20)the intermediate (cid:1) (cid:1) bosons W (cid:3) W (cid:3) and Z(cid:21)(cid:3) whose experimental veri(cid:17)cation renewed the interest of physicists in gauge theories(cid:2) On the mathematical side the investigation of self dual and anti self SU(cid:20)(cid:11)(cid:21)(cid:6)connections on (cid:5)(cid:6)manifoldsand of their imagein the orbit space led to stunning topological results in the topology of (cid:5)(cid:6)manifolds by Donaldson(cid:2) The modulispace of anti self dual connections can be com(cid:6) pleted and reworked into a (cid:14) dimensional manifoldwhich is a bordism between the base manifold of the bundle and a simple manifold which depends only on the Poincar(cid:22)e duality form in the second dimensional homology space(cid:2) Combined with results of Freedman this led to the (cid:2) discovery of exotic di(cid:19)erential structures on and on (cid:20)supposedly all (cid:2) but S (cid:21) compact algebraic surfaces(cid:2) In his codi(cid:17)cation of a principal connection (cid:23)Ehresmann(cid:3) (cid:7)(cid:9)(cid:14)(cid:7)(cid:24) be(cid:6) gan with a more general notion of connection on a general (cid:17)ber bundle (cid:20)E(cid:1)p(cid:1)M(cid:1)S(cid:21) with standard (cid:17)ber S(cid:2) This was called an Ehresmann con(cid:6) nection by some authors(cid:3) we will call it just a connection(cid:2) It consists ofthe speci(cid:17)cation of acomplementto the vertical bundle in adi(cid:19)eren(cid:6) tiable way(cid:3) called the horizontal distribution(cid:2) One can conveniently describe such a connection as a one form on the total space E with values in the vertical bundle(cid:3) whose kernel is the horizontaldistribution(cid:2) When combinedwithanother venerable no(cid:6) tion(cid:3) the Fr(cid:4)olicher(cid:6)Nijenhuis bracket for vector valued di(cid:19)erential forms (cid:20)see (cid:23)Fr(cid:1)olicher(cid:2)Nijenhuis(cid:3)(cid:7)(cid:9)(cid:14)(cid:15)(cid:24)(cid:21)oneobtainsaveryconvenientwayto describe curvature and Bianchi identity for such connections(cid:2) Parallel transport along curves in the base space is de(cid:17)ned only locally(cid:3)but one may show that each bundle admits complete connections(cid:3) whose paral(cid:6) lel transport is globally de(cid:17)ned (cid:20)theorem (cid:9)(cid:2)(cid:7)(cid:8)(cid:21)(cid:2) For such connections one can de(cid:17)ne holonomy groups and holonomy Lie algebras and one mayprove afar(cid:6)reachinggeneralizationofthe AmbroseSinger theorem(cid:1) A complete connection tells us whether it is induced from a principal connection on a principal (cid:17)ber bundle (cid:20)theorem (cid:7)(cid:11)(cid:2)(cid:5)(cid:21)(cid:2) Introduction (cid:6) A bundle (cid:20)E(cid:1)p(cid:1)M(cid:1)S(cid:21) without structure group can also be viewed as having the whole di(cid:19)eomorphism group Di(cid:19)(cid:20)S(cid:21) of the standard (cid:17)ber S as structure group(cid:3) as least when S is compact(cid:2) We de(cid:17)ne the non linear framebundle for E which is a principle (cid:17)ber bundle over M with structure group Di(cid:19)(cid:20)S(cid:21) and we show that the theory of connections on E corresponds exactlytothe theoryofprincipalconnectionsonthisnon linear frame bundle (cid:20)see section (cid:7)(cid:16)(cid:21)(cid:2) The gauge group of the non lin(cid:6) ear frame bundle turns out to be just the group of all (cid:17)ber respecting di(cid:19)eomorphismsofE which cover the identityon M(cid:3) and one maycon(cid:6) sider the modulispace Conn(cid:20)E(cid:21)(cid:2)Gau(cid:20)E(cid:21)(cid:2) There is hope that it can be strati(cid:17)ed into smooth manifolds corresponding to conjugacy classes of holonomygroups in Di(cid:19)(cid:20)S(cid:21)(cid:3) but this willbe treated in another paper(cid:2) The di(cid:19)eomorphism group Di(cid:19)(cid:20)S(cid:21) for a compact manifoldS admits a smoothclassifyingspace and a classifyingconnection(cid:1) the space ofall (cid:3) embeddingsofS intoaHilbertspace(cid:3) (cid:3)say(cid:3)isthetotalspaceofaprinci(cid:6) palbundlewithstructure groupDi(cid:19)(cid:20)S(cid:21)(cid:3) whosebase manifoldisthenon (cid:3) linear Grassmanian of all submanifolds of (cid:3) of type (cid:20)di(cid:19)eomorphic to(cid:21) S(cid:2) The action of Di(cid:19)(cid:20)S(cid:21) on S leads to the (cid:20)universal or classifying(cid:21) as(cid:6) sociated bundle(cid:3) which admitsaclassifyingconnection(cid:1) Every S(cid:6)bundle over M can be realized as the pull back of the classifying bundle by a classifying smooth mapping from M into the non linear Grassmanian(cid:3) which can be arranged in such a way that it pulls back the classifying connection to a given one on E (cid:13) this is the Narasimhan(cid:6)Ramadas procedure for Di(cid:19)(cid:20)S(cid:21)(cid:2) Characteristic classes owe their existence to invariantpolynomialson the Lie algebra of the structure group like characteristic coe(cid:25)cients of matrices(cid:2) The Lie algebra of Di(cid:19)(cid:20)S(cid:21) for compact S is the Lie algebra X(cid:20)S(cid:21) of vector (cid:17)elds on S(cid:2) Unfortunately it does not admitany invari(cid:6) ants(cid:2) But there are equivariants like X X(cid:4) for some closed form (cid:4) (cid:2)(cid:3)L onS whichcanbeusedtogiveasortofChern(cid:6)Weilconstructionofchar(cid:6) acteristic classes inthecohomologyofthe baseM withlocalcoe(cid:25)cients (cid:20)see sections (cid:7)(cid:15) and (cid:7)(cid:18)(cid:21)(cid:2) Finally we discuss some self duality and anti self duality conditions which depend on some(cid:17)berwise structure on the bundle likea (cid:17)berwise symplectic structure(cid:2) This booklet starts with a short description of analysis in in(cid:17)nite dimensions along the lines of Fr(cid:4)olicher and Kriegl which makes in(cid:17)nite dimensionaldi(cid:19)erential geometry much simpler than it used to be(cid:2) We treatsmoothness(cid:3)realanalyticityandholomorphy(cid:2) Thispartisbasedon (cid:23)Kriegl(cid:2)Michor(cid:3) (cid:7)(cid:9)(cid:9)(cid:8)b(cid:24) and the forthcoming book (cid:23)Kriegl(cid:2)Michor(cid:24)(cid:2) Then we give a more detailed exposition of the theory of manifolds of mappings and the di(cid:19)eomorphism group(cid:2) This part is adapted from (cid:10) Introduction (cid:23)Michor(cid:3) (cid:7)(cid:9)(cid:12)(cid:8)(cid:24) to the use of the Fr(cid:4)olicher(cid:6)Kriegl calculus(cid:2) Then we present acarefulintroductiontotheFr(cid:4)olicher(cid:6)Nijenhuisbracket(cid:3)to(cid:17)ber bundles and connections(cid:3) G(cid:6)structures and principal connection which emphasizes the construction and recognition of induced connections(cid:2) Thematerialintherestofthebook(cid:3)fromsection(cid:7)(cid:11)onwards(cid:3)hasbeen published in (cid:23)Michor(cid:3) (cid:7)(cid:9)(cid:12)(cid:12)(cid:24)(cid:2) The version here is much more detailed and contains moreresults(cid:2) The material in this booklet is a much extended version of a series of lectures held at the Institute of Physics of the University of Napoli(cid:3) March (cid:11)(cid:12)(cid:26)April(cid:7)(cid:3)(cid:7)(cid:9)(cid:12)(cid:12)(cid:2) IwanttothankG(cid:2)Marmoforhishospitality(cid:3) hisinterestinthissubject(cid:3)andforthesuggestiontopublishthisbooklet(cid:2) Wien(cid:3) August (cid:15)(cid:3) (cid:7)(cid:9)(cid:9)(cid:8) P(cid:2) Michor (cid:5) (cid:1)(cid:2) Notations and conventions (cid:3)(cid:4)(cid:3)(cid:4) De(cid:5)nition(cid:4) A Lie group G is a smooth manifold and a group such that the multiplication(cid:5) (cid:1) G G G is smooth(cid:2) Then also the (cid:1) (cid:3) inversion (cid:6) (cid:1)G G turns out to be smooth(cid:2) (cid:3) We shall use the followingnotation(cid:1) (cid:5)(cid:1)G G G(cid:3) multiplication(cid:3)(cid:5)(cid:20)x(cid:1)y(cid:21)(cid:27)x(cid:7)y(cid:2) (cid:1) (cid:3) (cid:8)a (cid:1)G G(cid:3) left translation(cid:3)(cid:8)a(cid:20)x(cid:21)(cid:27)a(cid:7)x(cid:2) (cid:3) (cid:9)a (cid:1)G G(cid:3) right translation(cid:3) (cid:9)a(cid:20)x(cid:21)(cid:27)x(cid:7)a(cid:2) (cid:3) (cid:1)(cid:4) (cid:6) (cid:1)G G(cid:3) inversion(cid:3) (cid:6)(cid:20)x(cid:21)(cid:27)x (cid:2) (cid:3) e G(cid:3) the unit element(cid:2) (cid:4) If g(cid:27)TeG is the Lie algebra of G(cid:3) we use the followingnotation(cid:1) AdG (cid:1)G AutLie(cid:20)g(cid:21) and so on(cid:2) (cid:3) (cid:3)(cid:4)(cid:6)(cid:4) Let (cid:3) (cid:1) G M M be a left action(cid:3) so (cid:3)(cid:28)(cid:1) G Di(cid:19)(cid:20)M(cid:21) is a (cid:1) (cid:3) (cid:3) group homomorphism(cid:2) Then we have partial mappings (cid:3)a (cid:1) M M x x (cid:3) and (cid:3) (cid:1)G M(cid:3) given by (cid:3)a(cid:20)x(cid:21)(cid:27)(cid:3) (cid:20)a(cid:21)(cid:27)(cid:3)(cid:20)a(cid:1)x(cid:21)(cid:27)a(cid:7)x(cid:2) (cid:3) M For any X g we de(cid:17)ne the fundamental vector (cid:1)eld (cid:10)X (cid:27) (cid:10)X (cid:4) x (cid:4) X(cid:20)M(cid:21) by (cid:10)X(cid:20)x(cid:21)(cid:27)Te(cid:20)(cid:3) (cid:21)(cid:7)X (cid:27)T(cid:5)e(cid:1)x(cid:6)(cid:3)(cid:7)(cid:20)X(cid:1)(cid:8)x(cid:21)(cid:2) Lemma(cid:4) In this situation the followingassertions hold(cid:1) (cid:20)(cid:7)(cid:21) (cid:10) (cid:1)g X(cid:20)M(cid:21) is a linear mapping(cid:2) (cid:3) (cid:20)(cid:11)(cid:21) Tx(cid:20)(cid:3)a(cid:21)(cid:7)(cid:10)X(cid:20)x(cid:21)(cid:27)(cid:10)Ad(cid:5)a(cid:6)X(cid:20)a(cid:7)x(cid:21)(cid:2) (cid:20)(cid:16)(cid:21) RX (cid:8)M X(cid:20)G M(cid:21) is (cid:3)(cid:3)related to (cid:10)X X(cid:20)M(cid:21)(cid:2) (cid:1) (cid:4) (cid:1) (cid:4) (cid:20)(cid:5)(cid:21) (cid:23)(cid:10)X(cid:1)(cid:10)Y(cid:24)(cid:27) (cid:10)(cid:7)X(cid:1)Y(cid:8)(cid:2) (cid:5) (cid:3)(cid:4)(cid:7)(cid:4) Let r (cid:1) M G M be a right action(cid:3) so r(cid:28) (cid:1) G Diff(cid:20)M(cid:21) (cid:1) (cid:3) (cid:3) is a group anti homomorphism(cid:2) We will use the following notation(cid:1) a a r (cid:1)M M and rx (cid:1)G M(cid:3) given by rx(cid:20)a(cid:21)(cid:27)r (cid:20)x(cid:21)(cid:27)r(cid:20)x(cid:1)a(cid:21)(cid:27)x(cid:7)a(cid:2) (cid:3) (cid:3) M For any X g we de(cid:17)ne the fundamental vector (cid:1)eld (cid:10)X (cid:27) (cid:10)X (cid:4) (cid:4) X(cid:20)M(cid:21) by (cid:10)X(cid:20)x(cid:21)(cid:27)Te(cid:20)rx(cid:21)(cid:7)X (cid:27)T(cid:5)x(cid:1)e(cid:6)r(cid:7)(cid:20)(cid:8)x(cid:1)X(cid:21)(cid:2) Lemma(cid:4) In this situation the followingassertions hold(cid:1) (cid:20)(cid:7)(cid:21) (cid:10) (cid:1)g X(cid:20)M(cid:21) is a linear mapping(cid:2) (cid:20)(cid:11)(cid:21) Tx(cid:20)ra(cid:3)(cid:21)(cid:7)(cid:10)X(cid:20)x(cid:21)(cid:27)(cid:10)Ad(cid:5)a(cid:1)(cid:1)(cid:6)X(cid:20)x(cid:7)a(cid:21)(cid:2) (cid:20)(cid:16)(cid:21) (cid:8)M LX X(cid:20)M G(cid:21) is r(cid:3)related to (cid:10)X X(cid:20)M(cid:21)(cid:2) (cid:1) (cid:4) (cid:1) (cid:4) (cid:20)(cid:5)(cid:21) (cid:23)(cid:10)X(cid:1)(cid:10)Y(cid:24)(cid:27)(cid:10)(cid:7)X(cid:1)Y(cid:8)(cid:2) (cid:11) (cid:3)(cid:2) Calculus of smooth mappings (cid:6)(cid:4)(cid:3)(cid:4) The traditional di(cid:19)erential calculus works well for (cid:17)nite dimen(cid:6) sional vector spaces and for Banach spaces(cid:2) For more general locally convex spaces a whole(cid:29)ock of di(cid:19)erent theories were developed(cid:3) each of themrathercomplicatedandnonereallyconvincing(cid:2) Themaindi(cid:25)culty isthatthecompositionoflinearmappingsstopstobejointlycontinuous at the level of Banach spaces(cid:3) for any compatible topology(cid:2) This was the originalmotivationfor the development of a whole new (cid:17)eld within general topology(cid:3)convergence spaces(cid:2) Thenin(cid:7)(cid:9)(cid:12)(cid:11)(cid:3)AlfredFr(cid:4)olicherandAndreas Krieglpresented indepen(cid:6) dently the solution to the question for the right di(cid:19)erential calculus in in(cid:17)nite dimensions(cid:2) They joined forces in the further development of the theory and the (cid:20)up to now(cid:21) (cid:17)nal outcome is the book (cid:23)Fr(cid:1)olicher(cid:2) Kriegl(cid:3) (cid:7)(cid:9)(cid:12)(cid:12)(cid:24)(cid:3) which is the general reference for this section(cid:2) See also the forthcomingbook (cid:23)Kriegl(cid:2)Michor(cid:24)(cid:2) In this section I willsketch the basic de(cid:17)nitions and the most impor(cid:6) tant results of the Fr(cid:4)olicher(cid:6)Kriegl calculus(cid:2) (cid:2) (cid:6)(cid:4)(cid:6)(cid:4) The c (cid:2)topology(cid:4) Let E be a locally convex vector space(cid:2) A (cid:2) curve c (cid:1) E is called smooth or C if all derivatives exist and (cid:3) (cid:2) are continuous (cid:6) this is a concept without problems(cid:2) Let C (cid:20) (cid:1)E(cid:21) be (cid:2) the space of smooth functions(cid:2) It can be shown that C (cid:20) (cid:1)E(cid:21) does not depend on the locally convex topology of E(cid:3) only on its associated bornology (cid:20)system of bounded sets(cid:21)(cid:2) The (cid:17)nal topologies with respect to the following sets of mappings into E coincide(cid:1) (cid:2) (cid:20)(cid:7)(cid:21) C (cid:20) (cid:1)E(cid:21)(cid:2) c(cid:5)t(cid:6)(cid:1)c(cid:5)s(cid:6) (cid:20)(cid:11)(cid:21) Lipschitz curves (cid:20)so that t(cid:1)s (cid:1)t(cid:27)s is bounded in E(cid:21)(cid:2) f (cid:6) g (cid:20)(cid:16)(cid:21) EB E (cid:1) B bounded absolutely convex in E (cid:3) where EB is f (cid:3) g the linear span of B equipped with the Minkowski functional pB(cid:20)x(cid:21)(cid:1)(cid:27)inf (cid:8)(cid:11)(cid:8)(cid:1)x (cid:8)B (cid:2) f (cid:4) g (cid:20)(cid:5)(cid:21) Mackey(cid:6)convergent sequences xn x (cid:20)there exists a sequence (cid:3) (cid:8)(cid:12)(cid:8)n with (cid:8)n(cid:20)xn x(cid:21) bounded(cid:21)(cid:2) (cid:7)(cid:8) (cid:5) (cid:2) (cid:2) This topologyiscalled the c (cid:6)topologyon E andwe write c E forthe resulting topologicalspace(cid:2) In general (cid:20)on the space of test functions D for example(cid:21) it is (cid:17)ner than the given locally convex topology(cid:3) it is not a vector space topology(cid:3) since scalar multiplication is no longer jointly continuous(cid:2) The (cid:17)nest among all locally convex topologies on E which (cid:2) arecoarser thanc E isthe bornologi(cid:17)cationofthegivenlocallyconvex (cid:2) topology(cid:2) If E is a Fr(cid:22)echet space(cid:3) then c E (cid:27)E(cid:2) (cid:8)(cid:12) Calculusofsmoothmappings (cid:13) (cid:6)(cid:4)(cid:7)(cid:4) Convenient vector spaces(cid:4) Let E be a locally convex vector space(cid:2) E is said to be a convenient vector space if one of the following equivalent (cid:20)completeness(cid:21) conditions is satis(cid:17)ed(cid:1) (cid:20)(cid:7)(cid:21) AnyMackey(cid:6)Cauchy(cid:6)sequence (cid:20)sothat (cid:20)xn xm(cid:21) isMackey con(cid:6) (cid:5)(cid:2) vergent to (cid:8)(cid:21) converges(cid:2) This is also called c (cid:6)complete(cid:2) (cid:20)(cid:11)(cid:21) If B is bounded closed absolutely convex(cid:3) then EB is a Banach space(cid:2) (cid:20)(cid:16)(cid:21) Any Lipschitz curve in E is locally Riemannintegrable(cid:2) (cid:2) (cid:2) (cid:3) (cid:20)(cid:5)(cid:21) For any c(cid:4) C (cid:20) (cid:1)E(cid:21) there is c(cid:3) C (cid:20) (cid:1)E(cid:21) with c(cid:4) (cid:27) c(cid:3) (cid:4) (cid:4) (cid:20)existence of antiderivative(cid:21)(cid:2) (cid:6)(cid:4)(cid:8)(cid:4) Lemma(cid:4) Let E be a locally convex space(cid:2) Then the following properties are equivalent(cid:1) (cid:2) (cid:20)(cid:7)(cid:21) E is c (cid:3)complete(cid:2) k k k (cid:20)(cid:11)(cid:21) If f (cid:1) E is scalarwise Lip (cid:4) then f is Lip (cid:4) for k (cid:11)(cid:7)(cid:2) (cid:3) (cid:2) (cid:20)(cid:16)(cid:21) If f (cid:1) E is scalarwise C then f is di(cid:5)erentiable at (cid:6)(cid:2) (cid:3) (cid:2) (cid:2) (cid:20)(cid:5)(cid:21) If f (cid:1) E is scalarwise C then f is C (cid:2) (cid:3) k k Here a mapping f (cid:1) E is called Lip if all partial derivatives (cid:3) n (cid:2) up to order k exist and are Lipschitz(cid:3) locally on (cid:2) f scalarwise C (cid:2) means that (cid:8) f is C for all continuous linear functionals on E(cid:2) (cid:9) This lemma says that a convenient vector space one can recognize smoothcurvesbyinvestigatingcompositionswithcontinuouslinearfunc(cid:6) tionals(cid:2) (cid:6)(cid:4)(cid:9)(cid:4) Smoothmappings(cid:4) LetEandF belocallyconvexvectorspaces(cid:2) (cid:2) (cid:2) A mappingf (cid:1) E F is called smooth or C (cid:3) if f c C (cid:20) (cid:1)F(cid:21) for (cid:2) (cid:3) (cid:2) (cid:2) (cid:9) (cid:4) all c C (cid:20) (cid:1)E(cid:21)(cid:30) so f(cid:4) (cid:1) C (cid:20) (cid:1)E(cid:21) C (cid:20) (cid:1)F(cid:21) makes sense(cid:2) Let (cid:2) (cid:4) (cid:3) C (cid:20)E(cid:1)F(cid:21) denote the space of all smooth mappingfromE to F(cid:2) For E and F (cid:17)nite dimensionalthis gives the usual notion of smooth mappings(cid:1) thishas been (cid:17)rst provedin (cid:23)Boman(cid:3)(cid:7)(cid:9)(cid:15)(cid:18)(cid:24)(cid:2) Constant map(cid:6) pings are smooth(cid:2) Multilinear mappings are smooth if and only if they are bounded(cid:2) Therefore we denote by L(cid:20)E(cid:1)F(cid:21) the space of allbounded linear mappingsfrom E to F(cid:2) (cid:2) (cid:2) (cid:6)(cid:4)(cid:10)(cid:4) Structureon C (cid:20)E(cid:1)F(cid:21)(cid:4) Weequipthe space C (cid:20) (cid:1)E(cid:21)withthe bornologi(cid:17)cation of the topology of uniform convergence on compact (cid:2) sets(cid:3) in all derivatives separately(cid:2) Then we equip the space C (cid:20)E(cid:1)F(cid:21) withthebornologi(cid:17)cationoftheinitialtopologywithrespect toallmap(cid:6) (cid:4) (cid:2) (cid:2) (cid:4) (cid:2) pings c (cid:1)C (cid:20)E(cid:1)F(cid:21) C (cid:20) (cid:1)F(cid:21)(cid:3)c (cid:20)f(cid:21) (cid:1)(cid:27)f c(cid:3) for allc C (cid:20) (cid:1)E(cid:21)(cid:2) (cid:3) (cid:9) (cid:4) (cid:6)(cid:4)(cid:11)(cid:4) Lemma(cid:4) For locally convex spaces E and F we have(cid:1) (cid:2) (cid:20)(cid:7)(cid:21) If F is convenient(cid:4) then also C (cid:20)E(cid:1)F(cid:21) is convenient(cid:4) for any E(cid:2) (cid:2) The space L(cid:20)E(cid:1)F(cid:21)is a closed linear subspace ofC (cid:20)E(cid:1)F(cid:21)(cid:4)so it (cid:3) (cid:8)(cid:12) Calculusofsmoothmappings also convenient(cid:2) (cid:20)(cid:11)(cid:21) IfE isconvenient(cid:4) then a curve c(cid:1) L(cid:20)E(cid:1)F(cid:21)is smoothifand (cid:3) only if t c(cid:20)t(cid:21)(cid:20)x(cid:21) is a smoothcurve in F for all x E(cid:2) (cid:2)(cid:3) (cid:4) (cid:6)(cid:4)(cid:12)(cid:4) Theorem(cid:4) Cartesian closedness(cid:4) The category of convenient vector spaces and smooth mappings is cartesian closed(cid:2) So we have a natural bijection (cid:2) (cid:2) (cid:2) C (cid:20)E F(cid:1)G(cid:21)(cid:27)C (cid:20)E(cid:1)C (cid:20)F(cid:1)G(cid:21)(cid:21)(cid:1) (cid:1) (cid:10) which is even a di(cid:5)eomorphism(cid:2) (cid:2) Ofcoarsethisstatementisalsotrueforc (cid:6)opensubsets ofconvenient vector spaces(cid:2) (cid:6)(cid:4)(cid:13)(cid:4) Corollary(cid:4) Let all spaces be convenient vector spaces(cid:2) Then the followingcanonical mappingsare smooth(cid:2) (cid:2) ev (cid:1)C (cid:20)E(cid:1)F(cid:21) E F(cid:1) ev(cid:20)f(cid:1)x(cid:21)(cid:27)f(cid:20)x(cid:21) (cid:2) (cid:1) (cid:3) ins(cid:1)E C (cid:20)F(cid:1)E F(cid:21)(cid:1) ins(cid:20)x(cid:21)(cid:20)y(cid:21) (cid:27)(cid:20)x(cid:1)y(cid:21) (cid:5) (cid:3)(cid:2) (cid:2) (cid:1) (cid:2) (cid:20) (cid:21) (cid:1)C (cid:20)E(cid:1)C (cid:20)F(cid:1)G(cid:21)(cid:21) C (cid:20)E F(cid:1)G(cid:21) (cid:6) (cid:2) (cid:3)(cid:2) (cid:2)(cid:1) (cid:20) (cid:21) (cid:1)C (cid:20)E F(cid:1)G(cid:21) C (cid:20)E(cid:1)C (cid:20)F(cid:1)G(cid:21)(cid:21) (cid:2) (cid:1) (cid:2)(cid:3) (cid:2) comp(cid:1)C (cid:20)F(cid:1)G(cid:21) C (cid:20)E(cid:1)F(cid:21) C (cid:20)E(cid:1)G(cid:21) (cid:2) (cid:2) (cid:1) (cid:3) (cid:2) (cid:3)(cid:3) (cid:2) (cid:2) (cid:2) (cid:3) (cid:3) C (cid:20) (cid:1) (cid:21)(cid:1)C (cid:20)F(cid:1)F (cid:21) C (cid:20)E (cid:1)E(cid:21) C (cid:20)C (cid:20)E(cid:1)F(cid:21)(cid:1)C (cid:20)E (cid:1)F (cid:21)(cid:21) (cid:1) (cid:3) (cid:20)f(cid:1)g(cid:21) (cid:20)h f h g(cid:21) (cid:2)(cid:3) (cid:2)(cid:3) (cid:9) (cid:9) (cid:2) (cid:2) (cid:1) C (cid:20)Ei(cid:1)Fi(cid:21) C (cid:20) Ei(cid:1) Fi(cid:21) (cid:3) Y Y Y Y (cid:6)(cid:4)(cid:3)(cid:14)(cid:4) Theorem(cid:4) Let E and F be convenient vector spaces(cid:2) Then the di(cid:5)erential operator (cid:2) (cid:2) d(cid:1)C (cid:20)E(cid:1)F(cid:21) C (cid:20)E(cid:1)L(cid:20)E(cid:1)F(cid:21)(cid:21)(cid:1) (cid:3) f(cid:20)x(cid:31)tv(cid:21) f(cid:20)x(cid:21) df(cid:20)x(cid:21)v (cid:1)(cid:27) lim (cid:5) (cid:1) t(cid:7)(cid:9) t exists and is linear and bounded (cid:7)smooth(cid:8)(cid:2) Also the chain rule holds(cid:1) d(cid:20)f g(cid:21)(cid:20)x(cid:21)v (cid:27)df(cid:20)g(cid:20)x(cid:21)(cid:21)dg(cid:20)x(cid:21)v(cid:7) (cid:9) (cid:6)(cid:4)(cid:3)(cid:3)(cid:4) Remarks(cid:4) Notethattheconclusionoftheorem(cid:11)(cid:2)(cid:12)isthestarting point of the classical calculus of variations(cid:3) where a smooth curve in a space of functions was assumed to be just a smooth function in one variable more(cid:2) (cid:8)(cid:12) Calculusofsmoothmappings (cid:2) If one wants theorem (cid:11)(cid:2)(cid:12) to be true and assumes some other obvious properties(cid:3) then the calculus of smooth functions is already uniquely determined(cid:2) Thereare(cid:3)however(cid:3)smoothmappingswhicharenotcontinuous(cid:2) This is unavoidableand not so horrible as it mightappear at (cid:17)rst sight(cid:2) For (cid:3) example the evaluation E E is jointly continuous if and only if (cid:1) (cid:3) E is normable(cid:3) but it is always smooth(cid:2) Clearly smooth mappings are (cid:2) continuous for the c (cid:6)topology(cid:2) For Fr(cid:22)echet spaces smoothness in the sense described here coincides (cid:2) with the notion Cc of (cid:23)Keller(cid:3) (cid:7)(cid:9)(cid:18)(cid:5)(cid:24)(cid:2) This is the di(cid:19)erential calculus used by (cid:23)Michor(cid:3)(cid:7)(cid:9)(cid:12)(cid:8)(cid:24)(cid:3)(cid:23)Milnor(cid:3)(cid:7)(cid:9)(cid:12)(cid:5)(cid:24)(cid:3)and (cid:23)Pressley(cid:2)Segal(cid:3)(cid:7)(cid:9)(cid:12)(cid:15)(cid:24)(cid:2) Aprevalent opinionincontemporarymathematicsis(cid:3)that forin(cid:17)nite dimensionalcalculus each serious application needs its own foundation(cid:2) Byaseriousapplicationoneobviouslymeanssomeapplicationofahard inversefunctiontheorem(cid:2) These theoremscanbeproved(cid:3)ifbyassuming enoughaprioriestimatesone creates enough Banachspace situationfor some modi(cid:17)ed iteration procedure to converge(cid:2) Many authors try to build their platonic idea of an a priori estimate into their di(cid:19)erential calculus(cid:2) Ithinkthatthis makesthe calculusinapplicableandhidesthe origin of the a priori estimates(cid:2) I believe(cid:3) that the calculus itself should be as easy to use as possible(cid:3) and that all further assumptions (cid:20)which most often come from ellipticity of some nonlinear partial di(cid:19)erential equation of geometric origin(cid:21) should be treated separately(cid:3) in a setting depending on the speci(cid:17)c problem(cid:2) I am sure that in this sense the Fr(cid:4)olicher(cid:6)Krieglcalculus as presented here andits holomorphicandreal analytic o(cid:19)springs in sections (cid:11) and (cid:16) below are universally usable for most applications(cid:2)

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.