(COMPLEX DIFFERENTIAL GEOMETRY Robert E. Greene 1. Complex manifolds Let C = the complex numbers and C"= Cx -- XC (nm factors), C” will be referred to as n-dimensional complex euclidean space. Of course, C" is topologically 2n -dimensional; more specifically, C” can and will be identified homeomerphically with R7 as follows: If (r1,...,2,) €C, andif 2) = x, + iy, €R, j = 1 then (2),..52,) is to be identified with (x1,y1,.-..%%_) €R. That is, (21,...,2,) is identified with (Re z;,Im z,,...,Re z,,Im z,), where Re and Im denote real and n imaginary parts, respectively. DEFINITION. A function f:D+C defined on an open subset D of C" is holomorphic if f considered as a function from DC R™* to R® is C™ and satisfies the Cauchy-Riemann equations: ae f _ aA ang AREA) _ _ aim fy for all j = 1,...,.2, Amapping f :D -C™ is holomorphic if each component of f is holomorphic; that is, if f,)...fj4 ate holomorphic, where f;:D -C, j= 1,...m are defined by f(z) = (f:(2),...sfmlz)) € C" for z € D. Note that if D = Uy¢, D, with each D, open, then f:D + C™ is holomorphic if and only if f | D, is holomorphic for every \. Multiplication by J, (z1,.-..2,) + (1,--.,iz,), defines a mapping from C” to C", whose square is multiplication by -1. The induced mapping from R™ to R™, which will be denoted by J, is a real linear endomorphism, whose composition with itself is again multiplication by —1. The following lemma shows that the endomorphism J can be used to characterize holomorphic mappings:229 LEMMA 1. A mapping f:D+C™, D®“ CC", which is C* when considered as a function from D © R™ to R™ is holomorphic if and only if f= Jy» = Inmfs. Here fs is the real Jacobian of f. Proof. Write fz) = (fi(2)s---sfn(2))» f@) €C for j= 1,...,m. In matrix form relative to the standard bases of R2* and R2*, [eat fe Ig» = aRefn) —_a(Ref n) . Oy ayn a(Imf») _a(Imf ») a, am a(Ref m) oy a(imf _) Ox, 0 -1 0230 Ign = Straightforward matrix multiplication shows that f.J = Jf. if and only if aRef) _ Mims) ., sms) __ Ref) ary ay; ax ay; for all k= 1,...9m and j = 1,....7 Lemma 1 implies immediately the following corollaries: COROLLARY. if f:D+C™ and g:D,~=C* are holomorphic mappings and if f(D) C DIP CO, then g af :D~C# is holomorphic. COROLLARY. if f:D+C", DCC", is holomorphic and is a diffeomor- phism onto its image when considered as a mapping from D C R™ to R™ then £1 f(D) =C" is holomorphic. (Note that in this case f(D) is necessarily open in C".) DEFINITION. A complex structure ona C™ manifold M of even dimension 2n is a maximal collection of C* charts indexed by a set A: {(.Uy) 2d € ALU Uy = R™} having Uc, Uy = M_ and satisfying the following condition (holomorphic transition) for all A,p € A, db oY) =U, NU) CR™ is a holomorphic function considered as a function from an open subset of C" to C".231 A complex manifold is a paracompact C* manifold M together with a complex structure. The number n is the complex dimension of M. The coordinate charts of the complex structure of a complex manifold are called holomorphic coordinate charts or holo- morphic coordinate systems. Note that an open subset of a complex manifold is itself a complex manifold in a standard fashion. LEMMA 2. Any collection of C” charts on a paracompact C™ manifold M which cover M and which satisfy the holomorphic transition condition determine one and only one complex structure on M; i.e. any such collection is contained in a unique maximal such col- lection. The proof of this lemma follows exactly the pattern (using the corollaries of Lemma 1) of the standard proof of the corresponding result for real manifolds. Let M bea complex manifold and p bea point of M. Denote the (real) tangent space of M at p by M,. Anendomorphism J, :M,~M, of the real vector space M, with j2= multiplication by —1 can be defined as follows: Choose a holo- morphic coordinate system y:U -R* defined in a neighborhood U of p. Then Ws, :M, -R™ is an isomorphism (the tangent space to R™" at 1(p) being identified with R™). J, is then defined to be wip'JguW-p. Lemma 2 implies immediately that Jp thus defined does not depend on the choice of ; if @ is a second holomorphic coordinate system around p then at p =O S gah ade because (7'). commutes with Jy by Lemma 2 and the holomorpiicity of w7'. oz gud. = Hence OF Uynds = OF abl gael be = WT gab. . ‘Thus J, is independent of holomorphic coordinate choice; that J? = multiplication by232 —1 is apparent from the fact that Jj» = multiplication by —1. Since holomorphic coordinate: systems are always C™, p + J, considered as a (1,1) tensot on M is C®. Note that this C™ tensor completely determines the complex structure on M since a C” coordinate system : UR” is a holomorphic coordi- nate system if and only if -,J, = Jpmife, for all p € U. DEFINITION. A C” mapping f:M-/M' from one complex manifold M to another M’ is holomorphic if fly = Jypfe- Lemma 2 shows that this definition is equivalent to the requirement that f be holo- morphic when expressed (locally) in holomorphic coordinate systems on M and M’. The definition as given has the advantage of avoiding local coordinate expressions. Note that if M’ is a complex manifold and M a C® real submanifold then M has at most one complex structure such that the injection i: M -~M’ is holomorphic since at most one Jig cam satisfy isJjy = Jypia, ia being injective. Of course such a complex structure on M may fail to exist. If one (and necessarily only one) such does exist, then M is said to be a complex submanifold of M'. Since the (1,1) tensor J with J? = multiplication by —1 on a complex manifold determines the complex structure, it is reasonable to consider such tensors independent of any complex structure: DEFINITION. Let M bea C™ manifold. An almost complex structure on M is a C* (1,1) tensor J on M such that, for every p € M,J2:M,—-M, is multiplication by -1. LEMMA 3. A real finite dimensional vector space which admits an endomorphism J:V=V~ satisfying J? = multiplication by —1 is necessarily even dimensional. Proof. Let X,Y = g(X,Y) be any positive definite inner product on V. Then233 (X,Y) - H(X,¥) defined by H(X,Y) = 9(X,¥) + g(UX,J¥) is a positive definite inner product on V relative to which J acts as an isometry. In particular it follows that any J-invariant subspace has a J-invariant complementary subspace. Now note that if X; €V and X, #0 then X; and JX, are real linearly independent since if JX, = @X;,a€R, then X, = — JX; = - aX, = - aX; or a? = — 1, which is impossible for a € R. Hence X, and JX, spana 2-dimensional subspace, which is obviously J-invariant. ‘This subspace has then a J-invariant comple- ment, say V,, of dimension equal to two less than the dimension of V. The lemma now follows by an obvious induction on dimension argument. REMARK. The existence of a J-invariant complement for every J-invariant sub- space is of course a special case of the complete reducibility of representations of finite groups. Lemma 3 shows that any manifold admitting an almost complex structure must be LEMMA 4. Let M_ be a paracompact (even dimensional) C” manifold with an almost complex structure Jy. in order that J be the almost complex structure on M asso- ciated to complex structure on M, tt is necessary and sufficient that for each point p of BM there exists a C° coordinate chart y:U-=R™, p €U, such that Wudy = Spa everywhere on U. Proof. The necessity of the condition is immediate. To prove the sufficiency, note that if w:U,+R™ and $:U;+R™ satisfy WoJg = Jgzile everywhere on U; and deJy = Jude everywhere on Uy then yb! : (Uy M U2) = Y(U;NU>) satis- fies Jaa"). everywhere on $(U; U3). By Lemma 1, 7! is holomorphic on the open set ("Yada = Paehs Vga = plybo! = Jpinpapst234 $(U; 1 U2), and Lemma 2 now implies the desired conclusion. It is possible to find conditions expressed directly in terms of the almost complex structure tensor J which are necessary and sufficient for J to arise from a complex structure. These conditions are discussed in the Appendix Z. 2. Basic Examples ‘Two complex manifolds M and N are biholomorphic if there is a (real) diffeamor- phism F : MN that is holomorphic. It follows from the results in §1 that F~" is then also holomorphic, From this, it then follows that the relation of being biholo- morphic is an equivalence relation. Two complex manifolds that are biholomorphic are the same object as far as the purposes of complex manifold theory go. ‘The most fundamental examples for the theory are the following: @ C*. This is R™ with the complex structure J determined by R™ = {(e,91---taadl Be) Pe Pe pw Az zl 2) fy fo Bas ‘The holomorphic coordinate system (x, + Y—1y;,.--2%_ + V=1y,) determines a complex manifold structure. (i) B= flea é els bP < i} the unit ball in ©". Since this is an open subset of C", it inherits a complex manifold structure automatically. It is important to note that C" and B” are nor biholomorphic even though they are real diffeomorphic. To sce this, note that a holomorphic mapping F:C"=B", F = (F,,...,.F,), Fy:C*+C_ is necessarily constant because Liouville’s Theorem immediately implies that cach F, is constant. (In detail: z+ F(a + br), a,b €C, z €C is constant so F; is itself constant.)235 (i Variations of B*. If S is a C™ submanifold that is C’ close to {(&1¥i9-+-s%nv¥n) | Ex? + Ey? = 1}, then the interior U, of S is an open subset of C* that is real diffeomorphic to B". If n = 1, then U, is biholomorphic to B1, by the Riemann Mapping Theorem. But if n = 2, U, is generically not biholomorphic to B”. The U, so obtained in fact form a collection of complex manifolds that in a suitable sense give an infinite: dimensional family of biholomorphic equivalence classes. (iv) P,C, complex projective space of (complex) dimension n. The notation CP" is also used. To define P,C, define an equivalence relation on C"*! — {(0,...,0)} as follows: Cie tnt) ~ (1g) if and only if 3 € C such that 2; = Aw; foreach j = 1,2, +1. (Exercise: Check that this is an equivalence relation.) As aset, P,C is the set of all equivalence classes of C**1 — {(0,...,0)}/~. Define a topology on P,C by declaring a set UCP,C to be open if and only if {@1s---tnt1) € C7! = ((0,...,0)} | [21.---Fn42)] € UF is open in C**!, where [ g]_ denotes the equivalence class of (z;,...;2,+1)- (This is the topology usually called the quotient topology relative to the equivalence relation.) P,C_ is compact because the map {G1)---st041) [Ekj? = 1} = PAC obtained by taking equivalence classes is surjective (and continuous by definition). To define a complex structure on P,C, we exhibit coordinate systems as follows: For each j, set Uy = {[(e1s.--s4n+1)] [zy # 0}. Clearly U Uy = P,C. Define Fy:Uy-C™ by [toes tneD) > Grit tlBje---stalti), OF §- The transition map from Fj(Uj) to F,(U;) defined on F/(Uy M U,) is easily seen to256 be holomorphic: it is essentially multiplication by z)/2. (Exercise: Check details of this.) Thus the F,:U;~C” maps define a complex manifold structure on P,C. (v) Submanifolds (of P,C): The definition of (complex) submanifold of a complex manifold runs parallel with the real theory: A subset N of a complex manifold M is an (embedded) submanifold if (1) N isa C* real embedded submanifold of M and (2) TN, all p €N, isa J invariant subspace of T)M (Z,N = real tangent space) of N at p. This definition condition (2) can be checked to be equivalent to the idea that in a neighborhood in M of each point g € N, NM neighborhood is a slice in some holo- morphic coordinate system, i.e., WA abhd = ((2),..-529) [y= 2y4y = 0 = ty =O} in some (holomorphic) coordinates (z;,...,2,)- A compact submanifold of P,C is called a (compact) algebraic variety. The fact that a compact complex submanifold of PC is an algebraic variety in the usual sense (of being the common zeroes of a set of homogeneous polynomials) is true, but hard to prove. A converse statement is, however, relatively easy. If P(z1,---,2,41) is. a homogeneous polynomial, then, for h # 0, P(A2 {ss RZpei) = 0 if and only if P(z;,...42441) = 0 since PORE yp Rfpan) = A! Pty. os2n) where d= degree (of homogencity) of P. Thus it makes sense to refer to P vanish- ing (or not) at a point of P,C. If P;,...,P, isa (finite) set of polynomials that are homogeneous, set V(P,,....P,) = the points of P,C at which all the P,,...,P, van- ish. The set V(Pj,..-,P;) need not be a real submanifold: it could have singularities. Bat if it is a real submanifold, then it is in fact necessarily a complex submanifold. Rather than considering this fact in full generality, we illustrate with a concrete exam- ple.237 VeVGPtri +I CPC. Consider VM Uy, j € {1,2,3} where Uy = {[(z1,22.23)]|z) # O} as before, with j = 3, say. The coordinates on U; are (z),22) with (21,22) + [(z1,72,1)] € P2C. The poly- nomial vanishes at [(z1,z2,1)] if and only if 2? + 23 +1=0. So VNUs corresponds in (z,,z2) coordinates to {(z,22) | 2? + 23 + 1 =O}. ‘This is a complex submanifold in C? because, near (w1,w>), w2 #0, with w? + wi+1=0, the functions (21,27 + 23 + 1) form a holomorphic coordinate system and {(z,z2) | zf + 23 + 1 = 0} is obviously a coordinate slice in this coordinate system (what if w, = 0? Exercise). So, checking similarly on VU, and VM U2, we see that V is a complex submanifold of P3C. It is important to realize that the analogue of Whitney’s Embedding Theorem does not hold in the complex case. First of all, a connected, compact complex submanifold of C" must be a point. (Proof outline: z) | is holomorphic if N is a complex submani- fold of C". If W is compact and connected, the Maximum Modulus Principle implies 3) is constant. Think for yourself about why the Maximum Modulus Principle applies in several variables!) So P,C is not realizable as a submanifold of C". You might be inclined to hope that P,C would play some sort of universal embed- ding theorem role. But this does not work, either. There are compact complex mani- folds that are not submanifolds (in the complex sense) of any P,C. It is not trivial to see why, however. See §3. 3, Hermitian and Kahler Metrics DEFINITION. A C* Riemannian metric g ona complex manifold M is an238 Hermitian metric if for each p € M and cach pair of tangent vectors X,Y ¢ M,, the tangent space of M at p, 80GY) = 80XIY) - Every complex manifold admits an Hermitian metric. To verify this fact, note that, since a complex manifold M is paracompact by definition, it necessarily admits a C* Riemannian metric g;. If g is defined by g(X,¥) = gi(X,¥) + 2VXY), then g is clearly an Hermitian metric. ‘The following lemma describes the pointwise structure of Hermitian metrics: LEMMA 1. Let V be areal vector space and J an endomorphism of V satisfying J? = multiplication by 1. Suppose that g is a positive definite inner product on V satisfying g(X,¥) = gUX,J¥) for all X,¥ €V. Then (a) s(XJ¥) = — gUX.Y). (b) there exists an orthonormal basis for V of the form X;,JK;,X2,JX2,....XqJXq- Proof. To prove (a), note that g(XJ¥) = g(JXJGY)) while 8UX,ICY)) = 8X, — ¥) = — 8UX,¥). To prove (b), one follows the method used to prove Lemma 3 of §1: Let X, bea unit vector in V. Then JX, is a unit vector since g(JX;JX;) = g(X;,X;). Also a(X) JX) = — gX1,X)) by part (a) so g(X),JX)) = 0. The subspace generated by X, and JX, is J-invariant, and hence its orthogonal complement is also J-invariant. The desired conclusion now follows by induction on the dimension of V. If g is an Hermitian metric, then the 2-tensor w defined by w(X,Y) = g(UX,Y) is antisymmetric by part (a) of Lemma 1. DEFINITION. If g is an Hermitian metric, the 2-form w defined by w(X,Y) = g(UX,¥) is the Kahler form of g.239 LEMMA 2. If g is an Hermitian metric, then @ \-- Aw (n factors, n= the complex dimension of M) is nowherz zero. Proof. Given p € M, let X;,JX},....X,,JX, be an orthonormal basis relative to g for M,. Such a basis exists by Lemma 1b). Then (XIX) = gUX JK) = 1 while o(X;,JX)) = g UX) JX) = 0 if i# 7. Also o(X;,X)) = oGX,,JX)) = 0 for all i,j. Then (A Aw) (Xp JX,.+-XpWXq) = nl a(X IX) «+ + (XX, In particular, (wA “+ Aw) # 0 at p. PROPOSITION 1. A complex manifold is orientable. Proof. Since any complex manifold admits an Hermitian metric, Lemma 2 shows that there exists a nowhere vanishing 2n form on any complex manifold of real dimen- sion 2n. Actually, it can be shown that a complex manifold is orientable without introducing any metric concepts. In fact, a holomorphic mapping from a domain in C" to C" which is a real diffeomorphism onto its image is necessarily orientation preserving on the underlying R* since its (real) Jacobian determinant is positive (by a calculation with determinants using the Cauchy-Riemann equations). Thus a covering by coordinate sys- tems all of whose transition mappings have positive Jacobian determinant is given directly by the complex structure. DEFINITION. An Hermitian metric g on a complex manifold is a Kahler metric if240 the Kabler form w associated tog is closed, ic. dw = 0. The condition dw = 0 implies a surprisingly close relationship between the metric g and the complex structure of M. One aspect of this relationship is expressed in the following proposition. PROPOSITION 2. Let M be a complex manifold and g an Hermitian metric on M. Then the following conditions on g are equivalent: (a) g isa Kahler metric (b) If D is the Riemannian covariant differentiation associated to g, then DJ = 0. Proof that (b) implies (a): Since Dg = 0 by definition of g and (X,Y) = g(JX,¥) for all X,Y, the vanishing of DJ implies that of Dw: Dzw(X,¥) = Z w(X,¥) — o(DzX,Y) — w(X,Dz¥) = Zg(JX.¥) — g(J(BzX).Y) — g(IX,Dz¥) = (Bzg) (X.Y) + (D2(0X),¥) + gUX.Dz¥) — g((DzX),¥) — gUX.Dz¥) = 0 + g((Da/)X.¥) + gU(OzX).Y) + gUX.Dz¥) — gV(O2X),¥) — sUXDz¥) a(D2)X,Y) =0 if DI=0. Now for any form «, the formula da = 5 du \ Digg o f holds, where (u') is a reai C* local coordinate system. In particular, if Dw = 0 then dw = 0. Proof that (a) implies (b): It suffices to establish the following formula for an arbitrary Hermitian metric g and its Kahler form w: 4 do(tyz) - £ dour) «241 For then the vanishing of dw implies that ¢(Dz/)X,Y) = 0 forall X and Y and hence that DzJ = 0. To establish the formula, note first that, all terms being tensors, the formula need be verified only in the case that X, Y, and Z are linear combina- tions with constant coefficients of the standard vector fields associated toa C™ local coordinate system. In fact, this coordinate system may be taken to be the real and ima- ginary parts of a holomorphic one, say (1,¥1,42,¥2,----tn,¥n)- Note that if X, ¥, and Z are (locally) such constant coefficient linear combinations of 9/ax; and a/ay; then so are JX, JY, and JZ so that the Lie brackets of any two of the six vector fields X, ¥, Z, JX, JY, JZ all vanish, Then a(Oz)X,¥) = 3(BzUX) — J(D2X),¥) = (0200.2) - U2.) = 8(D200).1) + (OXY) = $M s(12) + 220%) - VeK.2)] + £ lesen) + xelsv.z) - Wrnetx.2)] and on the other hand da(X,¥,Z) = Xw(¥,Z) + Za(X,Y) — ¥o(X,Z) = Xg(J¥,Z) + ZgUX,T) — YgUK,2) and da(JX,J¥,Z) = UX)w UZ) + Zo JY) — U¥)o(IX,Z) = — Og,Z) — ZeK JY) + UY) (XZ) - ‘The desired formula thus follows. Proposition 2 implies that the tensor J is invariant under parallel translation rela-242. tive to a Kahler metric; in particular, the value J, of J ata single point p € M together with the Kabler metric on M determines J everywhere on M (assuming that M is connected) and thus the complex structure of M. If g is a Kahler metric, the Kahler form w, being closed, determines a class in deRham 2-cohomology. Because w/-::\w (n times) is nowhere vanishing, and hence a nonvanishing volume form multiple, « cannot be exact. (Detail: If @ = d0, then wA-A@ (n times) = d(@Aw''-Aw), 2 — Lw’s. So 0 = f d@Awr--Aw) = f wr-rw #0, a contradiction.) Thus the cohomology class of @ is nonzero. Hence any Kahler mani- fold has nonvanishing 2-cohomology in the deRham sense. It can be seen that M = $+! x $44*1 admits a complex structure, all p,q = 1 (cf. [2]). But then M has zero 2-cohomology and hence does not admit Kahler metrics (see $10), so such an M cannot be realized as a submanifold of P,C. 4. Complexification of the Tangent and Cotangent Spaces DEFINITION. Let V be a real vector space. The complexification of V, to be denoted V©, is the complex vector space consisting of all ordered pairs (v,w),v, w € V with operations defined by (rm) + (2m) = Oy + waive + 2) - (@ + iB) ,w) = (av - Bwiaw + By), a BER -~ LEMMA 1. If V isa real vector space of dimension n and v},...,¥, is a basis for V, then (a) VO has complex dimension n and (¥;,0) --- (¥q,0) is a (complex) basis for VS (b) VE considered as a real vector space has dimension 2n and (¥1,0),...,(¥_,0), (0,¥:),.++5(0,¥_) is a (real) basis for ¥°.243, Proof. As an example, the complex linear independence of (¥3,0) ¥_0) will be verified, the other verifications being left to the reader: If for a+, By €R, & = 1,...,n, D (ae + iB) (v4.0) = (0,0) 1 (0,0) = 3 (exeBed = (s aon >, een] é=1 é a1 or SJL; age = 0 and SH, Bry = 0. Hence ay = 0 forall ¢ and B,=0 for all € by the real linear independence of the v;,...,Vq. Since in VC i(w,0) = (0,w) for any w € V, it is reasonable to introduce the notation v + iw for the element (v,w) of V°. DEFINITION, Let M be a complex manifold of complex dimension n, p bea point of M, and (21,...,24) = (81:91+-++2%n.¥n) be a holomorphic coordinate system defined in a neighborhood of p. Then 4/8z;|, and 9/Az;|,,/ = 1,....n, are the ele- ments of Mf given by see az; aa es az, ‘When the point p is clear from the context the abbreviated notations 4/dz;, a/dz; will be used. ‘Since |» + Sie ae 3x, & a 5 and244 gt ay Ba a ay) ax 8%; it follows from Lemma 1 that 8/4z;,...,8/8z,, /A7;,...,0/4%, being thus a spanning set of 2n elements, are a complex basis for Mf. Lemma 2. The complex subspace of ME spanned by 8/32;,...,0/82, is independent of the choice of holomorphic coordinate system (z1,...,Z,)- The same is true of the subspace spanned by 8/8%4,...,8/8Z_. Proof. Let (2*15.--52'n) = (4'1¥'1s---7¥'m¥'n) be another holomorphic coordinate system in a neighborhood of p. Then at p for any ¢ = 1,...,0 i fe a n ay; @ } oak ab nes SU 8x 2 (3 aye 8%) aye OY 1a (ey _, % ] Biel ay =23 (2-2 -i se at 2 2 e aye) Oy 2 3 oye xe aye Bap Bay) [de (8s pt z aye) [2 (ay ax'y since ay, a i ah i SL ax'y aye = ayy a's by the Cauchy-Riemann equations. Thus a ax; ax so a/az’, belongs to the complex subspace spanned by 8/421,....8/z,.248 The proof for the subspace spanned by 4/a7j,...,8/8/Z, is similar and will be omit- ted. If f isareal C™ function defined in a neighborhood of p, then each element of ME acts on f by complex lincar extension of the action of the elements of M, on f by differention: ( + ww)f = Of) + iff) v.w € Me, . A second complex linear extension definition gives an action of each clement of M© on complex valued functions defined in a neighborhood of p which are C* considered as functions into R*; (v + iw) () = (v + iw)Ref + i(v + iw)Imt = [p(Ref) — w(imf )] + ilw(Ref) + v(imf)] for v,w € Mp, Ref, Imf as usual. Note that in this sense the last equation in the proof of Lemma 2 can be rewritten a a a oe Fe az'y & (2 4) azy sinee oO ee ae ar, 7 2 a. ae SE Se ee 2 lar, ayy y's ax'y ax" ay'e the last equality following from again applying the Cauchy-Riemann equations. Similar computations can be used to show that where the Z in the parenthetical expression denotes the function xj — iy.246 LEMMA 3. Let f be a complex valued function defined in a neighborhood of p in ‘M which is C* as a function into R? and (2;,...,2,) be a holomorphic coordinate system ina neighborhood of p. Then (a) jor all € (x4 (where ~ denotes complex conjugation). (b) f is holomorphic on the domain of definition of the coordinate system (23, and only if everywhere on the domain f= for alt ¢= ary Proof. The assertion (a) follows immediately from the definitions. To prove (b), recall that f is holomorphic if and only if f.Jy = Jp:fs everywhere on the domain. Since a a a a A ey ee ee (a, aye {e,) axe ff is holomorphic if and only if and247 a)_(~ st (2|- (2 eo. amy] a(S |- (-2 axe (ent) » #, es] 2) Ske Rae nw(] = ( faint), <2 cres)} : ‘Thus the necessary and sufficient condition for f to be holomorphic is that (Ref) = mt) and 52 (int) = - aa aye ie Oe): Bi (Ref) - mn) + + (<2; amy + i, en} aye The process of complexifying the tangent space and relating this complexification to holomorphic coordinate systems has an analogue in the case of the cotangent space: DEFINITION. Let (21,...Z,) bea holomorphic coordinate system defined in a neighborhood of a point p € M. Then the elements dz, dz € (M*,)° are given by ds; + idy ds; — idy 1eak dz; a; Here Mz is the real cotangent space of M at p. Since dx; = (dx, + dijJ/2 and dy, = (dz; — de;)/2, it follows from Lemma 1 that +g, dE \,...,d2, form a complex basis for (M5)°. The previous definition is a special case of the following definition:248 DEFINITION. Let f be a complex-valued function defined in a neighborhood of p €M which is C* considered as a function into R*. Then df ¢ (M*,)° is given by Det af = d(Ref) + id(Imf) . LEMMA 4. For any f as in the previous definition, at the point p € M: reed 3 [-2 -3 (Edad (Ea for any holomorphic coordinate system (z;,...,2,) defined in a neighborhood of p. Proof. [z= (5 oem) sleds bedded 44 bids § [be bd] -4[[E)o- (E444: [ede (eda: (240-(24- from which the formula of the lemma follows. ‘Thus:249 LEMMA 5. The subspace of (M*,)° spanned by dz;,...,d2q Is independent of the holomorphic coordinate system defined in a neighborhood of p. The same is true of the sub- space of (M+p)© spanned by dij,...,di,. Proof. Let (2°js-+:s2',)s (Eiy+++s2q) be two holomorphic coordinate systems each defined in a neighborhood of p ¢ M. Then by Lemma 4 o a oa a w-Seorglede i hay Lemma 3 implies that (a/az)) 2*; = (8/42))2; = 0, so that de"; is a linear combination of the dz/’s. Similarly, dz; is a linear combination of aZ;’s. An clement w; + iw, € (M+,)©, w;,02 € M*,, gives rise to a complex linear fune- tional on (M*,)° by taking (1 + iw) (v + fw) = wy(v) — w2(w) + ia(v) + io,(w) for v,w €M,. (M*,)© can be thus identified with a subspace of the complex dual of MS, and by dimensionality considerations this subspace is in fact the whole of the com- plex dual of M¢. Note that using this identification procedure, dz1,...,d7,, dfj,...,d%, is the complex basis of (°°)© which is dual to the basis 4/dz;,...,4/02,, 0/42;,...,d/4Z, of (Mp)© since li] il 1 Bag sc 4 (a+ [2 +12] u 1 : F Gy - 8) +O=0 while a) eeLice canes [25 = oo «(Z) = 2 a+ (2 i ral = Fy + i) + 10) = by250 and similarly while a & || = 8, . (g)-- (Here 8y=0 f i#j=1 if 1=}.) The definition of df € (M*,)°, where f is complex valued, combined with the definition of the action of the elements of (M*,)° on M{ yields immediately that for al Vee af(¥) = VF where Vf is defined as previously. The duality of de;,...,d¢,, d&,,...@%, and :8/8Z_, 3/97}, ‘az, can be interpreted from this viewpoint also, since (Afaz))z = 0, (afdz)z; = By, ete. 8/825... 8/82; 5. Complex-Valued Differential Forms DEFINITION. A complex-valued r-form (ot complex r-form) on a real vector space V is anelement of (A‘V*)°, where A'V* = the real vector space of real r-forms on V. The (complex) wedge product \: (A’V*)© x (AV*)S = (A’**¥*)© is the complex linear extension of the real wedge product 4: (A'V*) x (A‘V*) = At*! V*. ‘The complex wedge product has the associativity and skewcommutativity properties of the real wedge product. LEMMA 1. If V is areal vector space of dimension N, then (A'V*)© has com- plex dimension (the binomial coefficient) (; ; and if {ey,...,0y} is any (complex) basis for (V*)© then the set of forms281 fw, Av Aw, [LS q =v, if for some (and hence by Lemma 2 any) holomorphic coordinate system (Z1,-++57_) defined in a neighborhood of p, @ belongs to the subspace of (A’M=,)° spanned by the set {dey Av Ade, Ndi, <0 Ads, [LS << Iw = v — iw, vow €V.258 This operation is conjugate linear relative to the complex vector space structure on V°. Anciement of V© is an element of V C V© if and only if it is invariant under the conjugation operation. Note that if an element of (A™M+,)© is of type (p,q) then its conjugate is of type (g,p) since dz, = dz, and 0,40) = 0,40) for any ele- ments 0;,.% of (A’M+,)°. DEFINITION. A (C*) complex-value r-form on M isa mapping © 2M = Use (AME such that (a) w(p) € (A™M*,)© for all p €M (b) @ is C® inthe sense that when w is expressed in terms of a {dey Aoe Ada, Ada), Nv A dey | 1 Sic ge HN de Ao Ads, Ndi, A,256 Bf ej, A Adri, Ade), A> A dz,) = a =BC 1p SE dey hm Bde, A dA A Ady, = i ‘These formulae can be used to provide computational proof of Lemma 4. 6. Hermitian and Kahler Metrics in Complex Notation The subspace of ME spanned by 9/8z;|,,...,/4z,|, (where (2. morphic coordinate system in a neighborhood of p € M) is called the holomorphic tangent space at p. It is a complex vector space of dimension n. By Lemma 2, $4, it is well defined independently of the choice of (z1,.-.,2,) coordinates. Notation: M. If g is a Hermitian metric on M (j.c., a J-invariant Riemannian metric), then g determines a Hermitian metric on each holomorphic tangent space M” in the conven- tional sense of complex linear algebra as follows: First note that g extends by complex linearity to be a complex bilinear form g on Mf. (This form cannot be positive defin- ite!) Define for V, W ¢ Mi a(V,) = av.) . Then g on M# so defined is a positive definite Hermitian metric on M#. (Note: The use—for both M, and MA— of the same symbol g can lead to no confusion, espe- cially since, for real vectors, the conjugation in the definition of g on M3 would be of no effect.) Define, given (21,..-,24) holomorphic coordinates, ag Be) eg (22S 89> 8 azar) ~ Flan) ° Then