Search the corpus

    Fibrewise isomorphic invertible sheaves differ by a pullback

    P-AGH3124FIBREWISEISO

    Area
    Algebraic Geometry
    Topics
    Flat Morphisms, Invertible Sheaves, Picard Group, Semicontinuity
    Status
    draft

    Let \(Y\) be an integral scheme of finite type over an algebraically closed field \(k\). Let \(f: X \to Y\) be a flat projective morphism whose fibres are all integral schemes. Let \(\mcl, \mcm\) be invertible sheaves on \(X\), and assume for each \(y \in Y\) that \(\mcl_y \cong \mcm_y\) on the fibre \(X_y\).

    Show that there is an invertible sheaf \(\mcn\) on \(Y\) such that \(\mcl \cong \mcm \tensor f^* \mcn\).

    Hint: use the results of this section to show that \(f_*(\mcl \tensor \mcm^{-1})\) is locally free of rank \(1\) on \(Y\).

    Solution

    Put \[ \mcf=\mcl\tensor\mcm^{-1}. \] Then \(\mcf\) is invertible on \(X\), and the hypothesis says \[ \mcf_y\cong\mco_{X_y} \] for every \(y\in Y\).

    1. The sheaf \(\mcf\) is flat over \(Y\), and\[ h^0(X_y,\mcf_y)=1 \]for every \(y\in Y\).

    Since \(f\) is flat, \(\mco_X\) is flat over \(Y\). Locally on \(X\), the invertible sheaf \(\mcf\) is isomorphic to \(\mco_X\), so \(\mcf\) is also flat over \(Y\).

    For every closed point \(y\in Y\), the residue field is \(k\), because \(Y\) is of finite type over the algebraically closed field \(k\). The fibre \(X_y\) is integral and projective over \(k\), hence \[ H^0(X_y,\mco_{X_y})=k. \] Thus \[ h^0(X_y,\mcf_y)=1 \] at every closed point.

    By semicontinuity, the locus where \[ h^0(X_y,\mcf_y)\ge2 \] is closed. If it were nonempty, then, since a finite-type \(k\)-scheme is Jacobson, it would contain a closed point, contrary to the preceding paragraph. On the other hand constants give \[ h^0(X_y,\mco_{X_y})\ge1 \] for every fibre, and \(\mcf_y\cong\mco_{X_y}\). Hence \[ h^0(X_y,\mcf_y)=1 \] for all \(y\in Y\).

    2. The sheaf\[ \mcn=f_*\mcf \]is invertible on \(Y\), and formation of \(f_*\mcf\) commutes with everyresidue-field base change:\[ \mcn\tensor\kappa(y) \xrightarrow{\sim} H^0(X_y,\mcf_y). \]

    The base \(Y\) is integral, hence reduced. By step 1, \(\mcf\) is coherent and flat over \(Y\), and the function \[ y\longmapsto h^0(X_y,\mcf_y) \] is constantly \(1\). The Grauert/base-change statement in cohomology and base change therefore gives that \[ f_*\mcf \] is locally free of rank \(1\), with the displayed base-change isomorphism. A locally free sheaf of rank \(1\) is invertible, so \(\mcn=f_*\mcf\) is the required line bundle candidate on \(Y\).

    3. The canonical evaluation morphism\[ \epsilon:f^*\mcn=f^*f_*\mcf\longrightarrow\mcf \]restricts to an isomorphism on every fibre \(X_y\).

    By the base-change isomorphism of step 2, the restriction of \(\epsilon\) to \(X_y\) is the ordinary evaluation map \[ H^0(X_y,\mcf_y)\tensor_{\kappa(y)}\mco_{X_y} \longrightarrow \mcf_y. \] Choose an isomorphism \[ \mcf_y\cong\mco_{X_y}. \] Step 1 gives \[ H^0(X_y,\mcf_y)\cong\kappa(y), \] and under this identification the evaluation map is \[ \kappa(y)\tensor_{\kappa(y)}\mco_{X_y} \longrightarrow \mco_{X_y}, \qquad a\tensor s\longmapsto as, \] which is an isomorphism.

    4. The evaluation morphism\[ \epsilon:f^*\mcn\longrightarrow\mcf \]is an isomorphism on \(X\).

    Let \[ \mcc=\operatorname{coker}(\epsilon). \] This is coherent. Fix \(x\in X\), put \(y=f(x)\), and take the stalk at \(x\). By step 3, tensoring the stalk sequence with \(\kappa(y)\) kills the cokernel: \[ \mcc_x/\mfm_y\mcc_x=0, \] where \(\mfm_y\mco_{X,x}\subseteq\mfm_x\). Nakayama’s lemma therefore gives \[ \mcc_x=0. \] Since this holds for every \(x\), the map \(\epsilon\) is surjective.

    Both its source and target are invertible sheaves. Locally at \(x\), \(\epsilon_x\) is therefore a surjective endomorphism between free rank-one \(\mco_{X,x}\)-modules. It is multiplication by a generator of the unit ideal, hence by a unit. Thus \(\epsilon_x\) is an isomorphism for every \(x\), so \(\epsilon\) is an isomorphism globally.

    5. There is an invertible sheaf \(\mcn\) on \(Y\) such that\[ \mcl\cong\mcm\tensor f^*\mcn. \]

    By step 4, \[ \mcl\tensor\mcm^{-1} =\mcf \cong f^*\mcn. \] Tensoring by \(\mcm\) gives \[ \boxed{\mcl\cong\mcm\tensor f^*\mcn}. \]

    6. Q.E.D.

    Step 5 is the required conclusion.