72E247F3A7CA
We now pass from the tangent bundle to the dual $V$ of an extension $W$ of $\mathcal O_ M$ by $\Omega ^1_ M$ with nonzero ruling component, for instance the canonical extension of a Kähler class. Sections of $\operatorname {Sym}^ rV$ are fiber-homogeneous polynomial functions on $W$. Compared with the cotangent bundle there is one more fiber coordinate $w$, the coordinate of the quotient $\mathcal O_ M$, and the affine term of the torsor becomes a quadratic relation among the fiber-polynomial functions over the ruling. The weight decomposition of the preceding subsection still applies and shows that the section algebra of $V$ is a polynomial ring in one variable over the tangent algebra. In particular, $\mathcal O_{\mathbb P(V)}(1)$ is big and the section algebra of $V$ is not finitely generated.
Keep the notation of the two preceding subsections. Let $W$ be an extension of $\mathcal O_ M$ by $\Omega _ M^1$ with class $\rho \xi +Kf^*\eta $, where $\rho \ne 0$, and put $V=W^*$. Adding the coordinate $w$ of the quotient $\mathcal O_ M$ to the charts in the proof of theorem 12.30.4, the fiber coordinates over the elliptic cover change by
\[ v=1/t,\qquad u_\infty =u,\qquad y_\infty =-t^2y-\rho t w,\qquad w_\infty =w. \]
For $\lambda =m\tau +n$, the lattice acts by
\[ (z,t,u,y,w)\longmapsto (z+\lambda ,b_\lambda t,u+m(Kw+q_0ty),y/b_\lambda ,w), \qquad q_0=2\pi \mathrm{i}d, \]
with the normalized factors $b_\lambda $. Fiber-homogeneous polynomial functions on $W$ represent symmetric tensors of $V$. In the ruling coordinates they satisfy a quadratic relation coming from the affine term of the torsor, while the elliptic monodromy determines which weights descend to $M$.
Lemma 12.32.1.46169CD2ACA2 Over $\mathbb C\times \mathbb P^1$, the graded algebra of fiber-homogeneous polynomial functions on the pullback of $W$ is
\[ \mathcal O(\mathbb C) [c_-,c_+,h,w,u]/\bigl (c_-c_+-h(h+\rho w)\bigr ), \]
where all the variables have degree one and, in the finite ruling chart,
\[ c_-=y,\qquad c_+=t(ty+\rho w),\qquad h=ty. \]
The lattice fixes $h$ and $w$, multiplies $c_-$ by $b_\lambda ^{-1}$ and $c_+$ by $b_\lambda $, and sends $u$ to $u+m(q_0h+Kw)$.
Proof.
Set $s_1=y$ and $s_2=ty+\rho w$ in the finite chart. Since $\rho \ne 0$, these are linear coordinates that can replace $y,w$, with $w=(s_2-ts_1)/\rho $. In the infinite chart put
\[ (s_{1,\infty },s_{2,\infty }) =(-v y_\infty -\rho w,-y_\infty ). \]
The transition is $(s_{1,\infty },s_{2,\infty })=t(s_1,s_2)$, so $s_1,s_2$ are the fiber coordinates of $\mathcal O_{\mathbb P^1}(-1)^{\oplus 2}$, while $u$ is the coordinate of a trivial line bundle.
By holomorphy in the infinite chart, a polynomial that is homogeneous of degree $j$ in $s_1,s_2$ has coefficients of degree at most $j$ in $t$. Distributing the exponent of $t$ among the $j$ fiber factors, every admissible monomial is a product of $s_1,s_2,ts_1,ts_2$. These four generators satisfy the determinant relation
\[ s_1(ts_2)-(ts_1)s_2=0, \]
and there are no further relations. Indeed, counting the $j+1$ monomials in $s_1,s_2$ and the $j+1$ admissible powers of $t$, the space of functions of degree $j$ has dimension $(j+1)^2$. The polynomial algebra in four variables modulo this nonzero quadratic relation has the same dimension
\[ \binom {j+3}{3}-\binom {j+1}{3}=(j+1)^2 \]
in degree $j$, where the second term is zero for $j<2$. The natural surjective graded map is therefore an isomorphism in every degree. Including $u$ adjoins a polynomial variable. The argument is uniform in $z$, and all coefficient functions are entire.
Identify the four generators as $c_-=s_1$, $h=ts_1$, $h+\rho w=s_2$ and $c_+=ts_2$; the determinant relation becomes the stated quadratic relation. These functions are polynomial also in the infinite chart, where
\[ c_-=-v^2y_\infty -\rho v w,\qquad c_+=-y_\infty ,\qquad h=-v y_\infty -\rho w. \]
The lattice formulas follow by substitution in the finite chart.
$\square$
Theorem 12.32.2.77C35E15A510 There is a degree-preserving isomorphism of complex algebras
\[ \bigoplus _{r\geq 0}H^0(M,\operatorname {Sym}^ rV) \simeq \left(\bigoplus _{r\geq 0}H^0(M,\operatorname {Sym}^ rT_ M)\right)[w], \]
where the variable $w$ on the right has degree one.
Proof.
Step 1: weights. Give $c_-$ weight $-1$, $c_+$ weight $1$, and $h,w,u$ weight zero. These weights come from the ruling action and are compatible with the quadratic relation and the lattice action. Using the relation to eliminate products $c_-c_+$, a polynomial of weight $-k$, $k\geq 1$, can be written as $c_-^ kF(z,h,w,u)$, and a polynomial of weight $k$ as $c_+^ kF(z,h,w,u)$, where $F$ has degree $r-k$ in tensor degree $r$; weight zero consists of the polynomials $F(z,h,w,u)$. These expressions are unique. Indeed, the substitution $c_+\mapsto h(h+\rho w)/c_-$ embeds the algebra of lemma 12.32.1 into
\[ \mathcal O(\mathbb C)[h,w,u,c_-,c_-^{-1}], \]
the reduced expressions involve distinct powers of $c_-$, and their coefficients vanish only if the original coefficients vanish, because $h(h+\rho w)$ is a nonzero polynomial. This proves that the weight decomposition is a direct sum.
Step 2: coefficient bundles. Let $U_3$ be the rank-three bundle on $E$ of linear polynomials in $h,w,u$ with the transformation rule
\[ F(z+\lambda ,h,w,u) =F(z,h,w,u-m(q_0h+Kw)). \]
By the transformation of $c_-$ and $c_+$ in lemma 12.32.1, the block of weight $-k$ is $H^0(E,A^ k\otimes \operatorname {Sym}^{r-k}U_3)$, and the block of weight $k$ is $H^0(E,A^{-k}\otimes \operatorname {Sym}^{r-k}U_3)$; the inverse translation of $u$ appears when invariance is written out. The signs of the twists agree with those in lemma 12.31.1.
Step 3: reduction to the tangent bundle. Since $q_0\ne 0$, put $H=h+(K/q_0)w$. Then the translation of $u$ is $u\mapsto u-mq_0H$, and consequently
\[ U_3\simeq U\oplus \mathcal O_ Ew, \]
where the variable $h$ of the bundle $U$ defined above is replaced by $H$. The symmetric powers of $U_3$ are the direct sums of the bundles $w^ j\operatorname {Sym}^{s-j}U$, $0\leq j\leq s$. The vanishing for negative twists in lemma 12.31.2 eliminates every positive weight, and the same lemma identifies the algebra of sections of weight zero with $\mathbb C[H,w]$.
In each negative weight, expanding in powers of $w$ identifies the remaining spaces with the corresponding blocks for the tangent bundle in lemma 12.31.1, with a polynomial variable $w$ adjoined. Sending the Euler variable $h$ of those blocks to $H$ and their leading factor $y^ k$ to $c_-^ k$ gives bijections in every tensor degree and weight. Products of surviving blocks multiply their coefficients and add their powers of $c_-$, and the eliminated positive weights contribute no products. The bijections therefore preserve multiplication and give the asserted isomorphism of graded algebras.
$\square$
Corollary 12.32.3.37D7893F0652 For every integer $r\geq 0$,
\[ h^0(M,\operatorname {Sym}^ rV) =r+1+d\binom {r+3}{4}. \]
The tautological line bundle on $\mathbb P(V)$ is big, with volume $d$, and the section algebra of $V$ is not finitely generated. These conclusions hold for the canonical extension of every Kähler class on $M$.
Proof.
The isomorphism of theorem 12.32.2 and theorem 12.31.3 give
\[ h^0(M,\operatorname {Sym}^ rV) =\sum _{j=0}^ r h^0(M,\operatorname {Sym}^ jT_ M) =r+1+d\sum _{j=0}^ r\binom {j+2}{3} =r+1+d\binom {r+3}{4}, \]
where $\binom {2}{3}=0$. The surface $M$ is projective, and by coherent GAGA the holomorphic bundle $V$ is algebraic. Thus $\mathbb P(V)$ is a projective fourfold, and the volume of its tautological line bundle is the limit of $24h^0(M,\operatorname {Sym}^ rV)/r^4$, which equals $d>0$. Hence this line bundle is big. If the section algebra of $V$ were finitely generated, so would be its quotient by $w$, contradicting theorem 12.31.4.
For the canonical extension $0\to \mathcal O_ M\to V\to T_ M\to 0$ of a Kähler class, the dual extension of $\mathcal O_ M$ by $\Omega ^1_ M$ has nonzero ruling component, as shown in the proof of theorem 12.30.4. Its class therefore has $\rho \ne 0$ in the conventions used here, and the same argument applies.
$\square$
Comments
Corrections, questions, and mathematical discussion are welcome. Sign in with GitHub to comment. Comments are public.