On the Riemann-Hilbert problem for hyperplane arrangements with a good line Shunya Adachi, Kazuki Hiroe Abstract We study a variant of the Riemann-Hilbert problem on the complements of hyperplane arrangements. This problem asks whether a given local system on the complement can be realized as the solution sheaf of a logarithmic Pfaffian system with constant coefficients. In this paper, we generalize Katz’s middle convolution as a functor for local systems on hyperplane complements and show that it preserves the solvability of this problem. Introduction Let \mathcal{A} be an arrangement of affine hyperplanes in the affine space V=\mathbb{C}^{l} , and denote its complement M(\mathcal{A})=V\backslash\mathcal{A} . In this paper, we study a variant of the Riemann-Hilbert problem on M(\mathcal{A}) which asks whether a given local system on M(\mathcal{A}) can be realized as the solution sheaf of a logarithmic Pfaffian system with constant coefficients. Since a hyperplane arrangement \mathcal{A} defines a divisor which is not normally crossing in general, this is still a nontrivial problem. The main goal of this paper is to prove that the middle convolution functor on M(\mathcal{A}) preserves the solvability of this problem. 0.1 Riemann-Hilbert problem for logarithmic Pfaffian systems Recall that the de Rham functor \mathrm{DR}\colon\mathrm{Conn}(M(\mathcal{A}))\rightarrow\mathrm{Loc}(M(\mathcal{A}),\mathbb{C});\quad\nabla\mapsto\mathrm{Ker}(\nabla) establishes an equivalence between the category \mathrm{Conn}(M(\mathcal{A})) of flat connections on M(\mathcal{A}) and the category \mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) of local systems on M(\mathcal{A}) . By restricting this functor to the category \mathrm{Pf}(\log\mathcal{A}) defined below, we aim to characterize the essential image of the restriction \mathrm{DR}_{\mathrm{Pf}}\colon\mathrm{Pf}(\log\mathcal{A})\rightarrow\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}). For a finite dimensional \mathbb{C} -vector space E , let us consider an \mathrm{End}_{\mathbb{C}}(E) -valued 1 -form on M(\mathcal{A}) , \Omega_{A}=\sum_{H\in\mathcal{A}}A_{H}\frac{df_{H}}{f_{H}} where A_{H}\in\mathrm{End}_{\mathbb{C}}(E) and f_{H} is the defining linear polynomial of H . Then under the integrability condition \Omega_{A}\wedge\Omega_{A}=0, this 1 -form defines the flat connection \nabla_{A}=d-\Omega_{A}\colon\mathcal{O}_{\mathbb{C}^{l}}\otimes_{\mathbb{C}}E\rightarrow\varOmega^{1}_{\mathbb{C}^{l}}(*\mathcal{A})\otimes_{\mathbb{C}}E. Then we call these flat connections logarithmic Pfaffian systems with constant coefficients on M(\mathcal{A}) , and denote the category of such flat connections by \mathrm{Pf}(\log\mathcal{A}) . The problem can now be formulated as follows. Problem 0.1 (Problem 5.1 ) . Let \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) be a local system. Does there exist a logarithmic Pfaffian system \nabla_{A}\in\mathrm{Pf}(\mathrm{log}(\mathcal{A})) such that \mathrm{DR}_{\mathrm{Pf}}(\nabla_{A})\cong\mathcal{L} ? We call such a logarithmic Pfaffian system \nabla_{A} a solution for \mathcal{L} . We briefly recall related works. Aomoto and Kohno addressed this problem in [ 2 , 15 ] , and proved that every unipotent local system admits a solution. On the other hand, for a smooth complex algebraic variety V with a smooth compactification \overline{V} such that D=\overline{V}\backslash V is a normally crossing divisor, Hain [ 10 ] considered the similar Riemann-Hilbert problem, asking whether a local system on V is realized as the solution sheaf for a connection d-\Omega with \Omega\in\varOmega^{1}_{\overline{V}}(\log D)(\overline{V})\otimes\mathfrak{gl}(n) , and gave a characterization of the solvability of this problem. Here \varOmega^{1}_{\overline{V}}(\log D) denotes the sheaf of meromorphic 1 -forms with logarithmic poles along D . 0.2 Main results We now present main results of this paper. For an affine hyperplane arrangement \mathcal{A} in \mathbb{C}^{l} , we say that a complex line Y in \mathbb{C}^{l} passing through the origin \mathbf{0} is good if X+Y\in L(\mathcal{A})\quad\text{for all }X\in L_{2}(\mathcal{A}), where L(\mathcal{A}) denotes the intersection poset of \mathcal{A} and L_{2}(\mathcal{A}) its subset of rank- 2 elements. This notion of goodness is a natural extension of that for central arrangements introduced by Terao [ 19 ] . Assume that \mathcal{A} admits a good line Y . For a nontrivial character \chi\colon\mathbb{Z}\to\mathbb{C}^{\times} , we construct in Definition 2.3 an endofunctor \mathrm{MC}_{\chi}\colon\mathrm{Loc}(M(\mathcal{A}),\mathbb{C})\longrightarrow\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}), called the middle convolution functor on M(\mathcal{A}) , as a generalization of the middle convolution functor on \mathbb{C}\setminus\{n\text{-points}\} introduced by Katz [ 14 ] . We show that this functor is compatible with Katz’s original construction (Proposition 2.4 ) and satisfies a composition law (Theorem 2.12 ), analogous to the classical case. On the other hand, Haraoka [ 12 ] defined a similar functor for \lambda\in\mathbb{C} , \mathrm{mc}_{\lambda}\colon\mathrm{Pf}(\log\mathcal{A})\longrightarrow\mathrm{Pf}(\log\mathcal{A}), for logarithmic Pfaffian systems with constant coefficients, as a generalization of the middle convolution for Fuchsian ordinary differential equations introduced by Dettweiler–Reiter [ 8 , 9 ] . Our first main result establishes the compatibility of these functors under the following assumption. Assumption 0.2 (Assumption 4.3 ) . Let \mathcal{A} be an affine hyperplane arrangement in \mathbb{C}^{l} with a good line Y , and let \mathcal{A}_{Y} denote the subset of \mathcal{A} consisting of hyperplanes parallel to Y . For a logarithmic Pfaffian system \nabla_{A}=d-\sum_{H\in\mathcal{A}}A_{H}\frac{df_{H}}{f_{H}}\in\mathrm{Pf}(\log\mathcal{A}), we impose the following condition with respect to a parameter \lambda\in\mathbb{C}\setminus\mathbb{Z} : for each H\in\mathcal{A}\setminus\mathcal{A}_{Y} , the matrix A_{H} has no nonzero integer eigenvalue, and the sum \sum_{H\in\mathcal{A}\setminus\mathcal{A}_{Y}}A_{H}+\lambda also has no nonzero integer eigenvalue. Theorem 0.3 (Theorem 4.8 ) . Let \nabla_{A}\in\mathrm{Pf}(\log\mathcal{A}) be a logarithmic Pfaffian system satisfying Assumption 0.2 for some \lambda\in\mathbb{C}\setminus\mathbb{Z} . Let \chi\colon\mathbb{Z}\to\mathbb{C}^{\times} be the character defined by \chi(1)=\exp(2\pi i\lambda) . Then there exists an isomorphism of local systems on M(\mathcal{A}) : \mathrm{MC}_{\chi}\circ\mathrm{DR}_{\mathrm{Pf}}(\nabla_{A})\cong\mathrm{DR}_{\mathrm{Pf}}\circ\mathrm{mc}_{\lambda}(\nabla_{A}). This theorem generalizes Haraoka’s result (Theorem 5.5 in [ 13 ] ) for braid arrangements and Dettweiler–Reiter’s result (Theorem 4.7 in [ 9 ] ) for \mathbb{C}\setminus\{n\text{-points}\} . As a corollary, we obtain the following statement, which ensures that the middle convolution functor preserves the solvability of the Riemann–Hilbert problem. Theorem 0.4 (Theorem 5.2 ) . Let \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) be a local system satisfying property \wp (see Definition 2.6 ), and let \chi\colon\mathbb{Z}\to\mathbb{C}^{\times} be a nontrivial character. Then: 1. If \mathcal{L} admits a solution \nabla_{A}\in\mathrm{Pf}(\log\mathcal{A}) satisfying Assumption 0.2 for some \lambda\in\mathbb{C}\setminus\mathbb{Z} with \chi(1)=\exp(2\pi i\lambda) , then \mathrm{MC}_{\chi}(\mathcal{L}) also admits a solution. 2. Conversely, if \mathrm{MC}_{\chi}(\mathcal{L}) admits a solution \nabla_{A^{\prime}}\in\mathrm{Pf}(\log\mathcal{A}) satisfying Assumption 0.2 for some \lambda^{\prime}\in\mathbb{C}\setminus\mathbb{Z} with \chi(1)=\exp(-2\pi i\lambda^{\prime}) , then \mathcal{L} also admits a solution. 1 Fiber bundle structure on complement of hyperplane arrangement In [ 19 ] , Terao provided a characterization that equips the complements of central hyperplane arrangements with a fiber bundle structure. This section shows that an analogous characterization remains valid even for affine hyperplane arrangements. 1.1 Hyperplane arrangement with modular coatom This section gives a review of the work by Terao [ 19 ] , in which he constructed a correspondence between hyperplane arrangements with a modular coatom and fiber bundle structures on their complements. Let us fix an \ell -dimensional \mathbb{C} -vector space V . A central hyperplane arrangement is a finite collection \mathcal{A} of (\ell-1) -dimensional subspaces of V . Then we associate a lattice L(\mathcal{A}) with \mathcal{A} , called the intersection lattice which is defined as the collection of all intersections of elements of \mathcal{A} . Here the partial order on L(\mathcal{A}) is usually defined by reverse inclusion. Therefore the join of X,Y\in L(\mathcal{A}) is defined as X\lor Y:=X\cap Y, and the meet as X\land Y:=\bigcap_{\begin{subarray}{c}Z\in L(\mathcal{A}),\\ X+Y\subset Z\end{subarray}}Z. Obviously, V and T(\mathcal{A}):=\bigcap_{H\in\mathcal{A}}H are the unique minimal and maximal elements respectively. For X\in L(\mathcal{A}) , the rank of X , denoted by \mathrm{rk}(X) , is defined as the codimension. The subset consisting of elements of rank k is denoted by L_{k}(\mathcal{A}):=\{X\in L(\mathcal{A})\mid\mathrm{rk\,}X=k\}. Rank one elements are usually called atoms , also rank \mathrm{rk\,}T(\mathcal{A})-1 elements are called coatoms . For the intersection lattice L(\mathcal{A}) , the set of all atoms in L(\mathcal{A}) is exactly \mathcal{A} itself. Therefore L(\mathcal{A}) is an atomic , i.e., any X\in L(\mathcal{A}) is a join of atoms. We say that X\in L(\mathcal{A}) is modular if it satisfies \mathrm{rk}(X)+\mathrm{rk}(Y)=\mathrm{rk}(X\land Y)+\mathrm{rk}(X\lor Y) for all Y\in L(\mathcal{A}) . Since the equation \mathrm{codim}_{\mathbb{C}}X+\mathrm{codim}_{\mathbb{C}}Y=\mathrm{codim}_{\mathbb{C}}(X\cap Y)+\mathrm{codim}_{\mathbb{C}}(X+Y) holds any subspaces X,Y\subset V , X\in L(\mathcal{A}) is modular exactly when X+Y\in L(\mathcal{A}) holds for all Y\in L(\mathcal{A}) . We now consider the complement M(\mathcal{A}):=V\,\backslash\bigcup_{H\in\mathcal{A}}H of \mathcal{A} , and explain that a modular coatom yields a fibration on M(\mathcal{A}) . Let Y\subset V be a linear subspace and p_{Y}\colon V\rightarrow V/Y the natural projection. Then the subarrangement \mathcal{A}_{Y}:=\{H\in\mathcal{A}\mid H\supset Y\} defines the arrangement in V/Y , p\mathcal{A}_{Y}:=\{p_{Y}(H)\mid H\in\mathcal{A}_{Y}\}, and \pi_{Y}:=p_{Y}|_{M(\mathcal{A})} defines the surjection \pi_{Y}\colon M(\mathcal{A})\rightarrow M(p\mathcal{A}_{Y}). Here we note that Y is not necessarily an element in L(\mathcal{A}) , and thus consider the maximal element with respect to the order in L(\mathcal{A}) , which contains Y , \langle Y\rangle:=\bigcap_{H\in\mathcal{A}_{Y}}H\in L(\mathcal{A}). We are now ready to recall the theorem by Terao, which relates the following goodness condition for Y and the fiber bundle structure on the projection \pi_{Y}\colon M(\mathcal{A})\rightarrow M(p\mathcal{A}_{Y}) . Definition 1.1 . The above Y is said to be good if either p_{Y}(X)\cap M(p\mathcal{A}_{Y})=\emptyset\quad\text{ or }\quad p_{Y}(X)=V/Y hold for every X\in L(\mathcal{A}) . Theorem 1.2 (Terao, Theorem 2.9 in [ 19 ] ) . Suppose \mathrm{dim}_{\mathbb{C}}Y=1 . Then the following are equivalent. 1. The surjection \pi_{Y}\colon M(\mathcal{A})\rightarrow M(p\mathcal{A}_{Y}) is a fiber bundle. 2. Each fiber of \pi_{Y} is \mathbb{C} with |\mathcal{A}\backslash\mathcal{A}_{Y}| points removed. 3. The line Y is good. 4. \langle Y\rangle=T(\mathcal{A}) , or \langle Y\rangle is modular of rank \mathrm{rk}(T(\mathcal{A}))-1 . By following the argument in [ 4 ] , we can give another characterization of goodness. Proposition 1.3 . Suppose \mathrm{dim}_{\mathbb{C}}Y=1 . Then Y is good if and only if X+Y\in L(\mathcal{A}) for all X\in L_{2}(\mathcal{A}). Before the proof of this proposition, recall elementary properties of \langle Y\rangle . First we notice that the equation holds, \langle Y\rangle=Y+T(\mathcal{A}). Indeed, we have \displaystyle Y+T(\mathcal{A}) \displaystyle=Y+\bigcap_{H\in\mathcal{A}}H=Y+\left(\left(\bigcap_{H\in\mathcal{A}_{Y}}H\right)\cap\left(\bigcap_{H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y}}H^{\prime}\right)\right) \displaystyle=\left(Y+\bigcap_{H\in\mathcal{A}_{Y}}H\right)\cap\bigcap_{H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y}}(Y+H^{\prime})=\bigcap_{H\in\mathcal{A}_{Y}}H\cap\bigcap_{H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y}}V \displaystyle=\bigcap_{H\in\mathcal{A}_{Y}}H=\langle Y\rangle. Here we note that the equation X+(A\cap B)=(X+A)\cap(X+B) holds for any subspaces X,A,B\subset V , and also note the equations Y+H=Y for H\in\mathcal{A}_{Y} , and Y+H^{\prime}=V for H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y}. Next recall the equation X+\langle Y\rangle=X+Y\quad(X\in L(\mathcal{A})), which follows from the fact X+T(\mathcal{A})=X for X\in L(\mathcal{A}) and the above equation. Proof of Proposition 1.3 . Suppose that Y is good. Then since \langle Y\rangle is modular by Theorem 1.2 , we obtain X+Y=X+\langle Y\rangle\in L(\mathcal{A}) for any X\in L(\mathcal{A}) . Thus the only if part holds. Now we show the if part. If Y\subset T(\mathcal{A}) , then \langle Y\rangle=Y+T(\mathcal{A})=T(\mathcal{A}) which implies goodness obviously. Thus we may assume Y\nsubset T(\mathcal{A}) and it suffices to show that \langle Y\rangle is modular of rank \mathrm{rk}(T(\mathcal{A}))-1 . Since the rank condition follows from the equation \langle Y\rangle=Y+T(\mathcal{A}) , we need to show X+\langle Y\rangle\in L(\mathcal{A}) for all X\in L(\mathcal{A}) . Suppose the above holds for L_{k-1}(\mathcal{A}) for k>2 , and take X\in L_{k}(\mathcal{A}) . We may suppose X\nsubset Y since we have X+\langle Y\rangle=X+Y=X otherwise. Let us write X=H_{1}\cap\ldots\cap H_{k} by some H_{1},\ldots,H_{k}\in\mathcal{A}. Then we have X+\langle Y\rangle=X+Y=(H_{1}\cap\ldots\cap H_{k})+Y=((H_{1}\cap H_{2})+Y)\cap((H_{2}\cap\ldots\cap H_{k})+Y). Since H_{2}\cap\ldots\cap H_{k}\in L_{k-1}(\mathcal{A}) , the induction hypothesis shows that (H_{2}\cap\ldots\cap H_{k})+Y=(H_{2}\cap\ldots\cap H_{k})+\langle Y\rangle\in L(\mathcal{A}) . Also by the assumption, (H_{1}\cap H_{2})+Y\in L(\mathcal{A}) since H_{1}\cap H_{2}\in L_{2}(\mathcal{A}) . We conclude X+\langle Y\rangle\in L(\mathcal{A}) . Therefore the desired condition follows by the induction on k . ∎ 1.2 Fibration on affine hyperplane arrangement From now on, we remove the assumption that the arrangement is central. Namely, let \mathcal{A} denote a finite collection of affine hyperplanes of the \ell -dimensional affine space V\cong\mathbb{C}^{l} . As well as the central case, the intersection poset L(\mathcal{A}) , which is not a lattice in general. The complement is denoted by M(\mathcal{{A}}) as before. It is well-known that affine arrangements and central arrangements are related to each other by means of their cones and decones which are explained below. Let us fix a coordinate system V\cong\mathbb{C}^{l}=\{(x_{1},\ldots,x_{l})\mid x_{i}\in\mathbb{C}\} . Then an affine hyperplane H\in L(\mathcal{A}) is the zero locus of the polynomial f_{H}(x)=L_{H}(x)+a_{H} with linear homogeneous polynomial L_{H}(x) and a constant a_{H} . We call this f_{H}(x) the defining linear polynomial of H . Then the defining polynomial of the arrangement \mathcal{A} is the product of them, Q_{\mathcal{A}}(x):=\prod_{H\in\mathcal{A}}(L_{H}(x)+a_{H}). Definition 1.4 (cone of affine arrangement) . Let \mathcal{A} be an arrangement of affine hyperplanes in \mathbb{C}^{l} . Then the cone \mathbf{c}\mathcal{A} of \mathcal{A} is the central arrangement of \mathbb{C}^{l+1} with the defining polynomial Q_{\mathbf{c}\mathcal{A}}(x_{0},x_{1},\ldots,x_{l}):=x_{0}^{m+1}\cdot Q_{\mathcal{A}}(x_{1}/x_{0},\ldots,x_{l}/x_{0}), where m is the degree of Q_{\mathcal{A}}(x) . Note that \mathbf{c}\mathcal{A} contains the additional hyperplane H_{0}=\{x_{0}=0\} . We can obviously recover the affine arrangement \mathcal{A} from the cone \mathbf{c}\mathcal{A} by restricting to x_{0}=1 . Definition 1.5 (decone of central arrangement) . Let \mathcal{B} be a central arrangement in \mathbb{C}^{l+1}=\{(x_{0},x_{1},\ldots,x_{l})\mid x_{i}\in\mathbb{C}\} . Then the decone \mathbf{d}\mathcal{\mathcal{B}} of \mathcal{B} is the affine arrangement of \mathbb{C}^{l} with the defining polynomial Q_{\mathbf{d}\mathcal{B}}(x_{1},\ldots,x_{l}):=Q_{\mathcal{B}}(1,x_{1},\ldots,x_{l}). Therefore, we have the natural inclusion M(\mathcal{A})=M(\mathbf{c}\mathcal{A})\cap\{x_{0}=1\}\subset M(\mathbf{c}\mathcal{A}), by regarding \mathcal{A} as the decone of \mathbf{c}\mathcal{A} . Also for intersection posets, there is the natural inclusion L(\mathcal{A})\cong\{X\in L(\mathbf{c}\mathcal{A})\mid X\nsubset\{x_{0}=0\}\}\subset L(\mathbf{c}\mathcal{A}). The characterization of the goodness in Proposition 1.3 allows us to define good lines even for the affine case as follows. Definition 1.6 . Let Y be an affine line in \mathbb{C}^{l} passing through the origin \mathbf{0} . Then Y is said to be good if X+Y\in L(\mathcal{A}) for all X\in L_{2}(\mathcal{A}) . In this case, we also say \mathcal{A} is Y -closed . Remark 1.7 . The terminology Y -closed was introduced in [ 17 ] by Oshima. The purpose of this section is to show that Theorem 1.2 by Terao is still valid even for affine arrangements. For an affine arrangement \mathcal{A} and an affine line Y through \mathbf{0} , we replace the definition of \mathcal{A}_{Y} previously defined for the central case, with the following one, \mathcal{A}_{Y}:=\{H\in\mathcal{A}\mid H\text{ is parallel to }Y\}. Theorem 1.8 . Let \mathcal{A} be an affine arrangement in \mathbb{C}^{l} and Y an affine line in \mathbb{C}^{l} through the origin. The following are equivalent. 1. The surjection \pi_{Y}\colon M(\mathcal{A})\rightarrow M(p\mathcal{A}_{Y}) is a fiber bundle. 2. Each fiber of \pi_{Y} is \mathbb{C} with |\mathcal{A}\backslash\mathcal{A}_{Y}| points removed. 3. The line Y is good in the sense of Definition 1.6 . For the proof of this theorem, we investigate some properties of the good line Y in the cone \mathbf{c}\mathcal{A} . Let us regard M(\mathcal{A}) as a subspace of M(\mathbf{c}\mathcal{A}) as above. Define the 1 -dimensional subspace Y_{\mathbf{0}}:=Y-e_{0} of \mathbb{C}^{l+1} under the parallel translation by e_{0}:=(1,0,\ldots,0)\in\mathbb{C}^{l+1} . Then the diagram {\lx@inpgf@ignorespaces M(\mathcal{A})} {\lx@inpgf@ignorespaces M(\mathbf{c}\mathcal{A})} {\lx@inpgf@ignorespaces M(p\mathcal{A}_{Y})} {\lx@inpgf@ignorespaces M(p(\mathbf{c}\mathcal{A})_{Y_{\mathbf{0}}})} \scriptstyle{\lx@inpgf@ignorespaces p_{Y}} \scriptstyle{\lx@inpgf@ignorespaces p_{Y_{\mathbf{0}}}} (1) is obviously a pullback in the category of topological spaces. Proposition 1.9 . If Y is good in the affine arrangement \mathcal{A} , then Y_{\mathbf{0}} is also good in the central arrangement \mathbf{c}\mathcal{A} . Proof. It suffices to show that Y_{\mathbf{0}} satisfies the condition in Proposition 1.3 . Let us take X\in L_{2}(\mathbf{c}\mathcal{A}) and write X=H_{1}\cap H_{2} by H_{i}\in\mathbf{c}\mathcal{A} . First consider the case that either H_{1} or H_{2} is H_{0}=\{x_{0}=0\} . Since Y_{\mathbf{0}} is contained in H_{0} , Y_{\mathbf{0}}+X equals to X or H_{0} in this case. Thus Y_{\mathbf{0}}+X\in L(\mathcal{A}) . Next we assume that both H_{i} are coming from \mathcal{A} and write the corresponding affine hyperplanes by \mathbf{d}H_{i}\in\mathcal{A} . Then since Y is good, we may assume that Y is parallel to either of \mathbf{d}H_{i} (see (iii) and (v) in Lemma 2.1 in [ 17 ] ). Thus Y_{\mathbf{0}} is contained in either of H_{i} . Then it follows from the same argument as above that Y+X\in L(\mathcal{A}) . ∎ Proof of Theorem 1.8 . The equivalence of 1 and 2 follows from the same argument in Theorem 2.9 in [ 19 ] . Let us see the equivalence of 2 and 3. First suppose that 3 holds. Then the above proposition assures that Y_{\mathbf{0}} is good. Since p_{Y_{\mathbf{0}}} is a fibration by Theorem 1.2 , and the diagram (\ref{eq:pullback}) is a pullback, p_{Y} is also a fibration. Thus 1 holds. Conversely, suppose that 2 holds. Then we can show that 3 holds by a similar argument as Theorem 3.25 in [ 4 ] as follows. Under a linear transformation, we may assume that Y is the x_{l} -axis and the projection p_{Y}\colon M(\mathcal{A})\rightarrow M(\mathcal{A}_{Y}) is given by (x_{1},\ldots,x_{l-1},x_{l})\mapsto(x_{1},\ldots,x_{l-1}). To distinguish x_{l} from the other coordinates, put x_{l}=y . Then the condition 2 implies that the defining polynomial Q_{\mathcal{A}} has no multiple roots as the polynomial of y for any \mathbf{x}=(x_{1},\ldots,x_{l-1})\in M(p\mathcal{A}_{Y}). Let us decompose Q_{\mathcal{A}} as Q_{\mathcal{A}}=\prod_{H\in\mathcal{A}_{Y}}(L_{H}(\mathbf{x})+\alpha_{H})\cdot\prod_{H^{\prime}\notin\mathcal{A}_{Y}}(L_{H^{\prime}}(\mathbf{x},y)+\alpha_{H^{\prime}}). We now take X\in L_{2}(\mathcal{A}) and write X=H_{1}\cap H_{2} by H_{i}\in\mathcal{A} . If both of H_{i} are parallel to Y , then X+Y=X\in L(\mathcal{A}) . If one is parallel to Y , put it as H_{1} , and the other is not, then X+Y=H_{1} . Finally assume both of H_{i} are not parallel to Y . Then for (\mathbf{x}_{0},y_{0})\in H_{1}\cap H_{2} , \prod_{H^{\prime}\notin\mathcal{A}_{Y}}(L_{H^{\prime}}(\mathbf{x}_{0},y)+\alpha_{H^{\prime}}) has y_{0} as a double root. Thus \prod_{H\in\mathcal{A}_{Y}}(L_{H}(\mathbf{x}_{0})+\alpha_{H}) must be zero by the assumption. This means that there exists H\in\mathcal{A}_{Y} such that (\mathbf{x}_{0},y)\in H for all y\in\mathbb{C} , i,e, (x_{0},y_{0})+Y\in H . Since \mathcal{A} is a finite set, this implies that (H_{1}\cap H_{2})+Y=H\in L(\mathcal{A}) . In conclusion, we have X+Y\in L(\mathcal{A}) in all cases, i.e. the condition 3 holds. ∎ 2 Middle convolution on complements of hyperplane arrangements with a good line Let \mathcal{A} be an affine arrangement in \mathbb{C}^{l}=\{(x_{1},\ldots,x_{l})\mid x_{i}\in\mathbb{C}\} with a good line Y . Under a linear transformation, we may assume that Y is the x_{l} -axis and then put y=x_{l} . Let \mathrm{Sh}(M(\mathcal{A}),\mathbb{C}) denote the category of sheaves of \mathbb{C} -vector spaces over M(\mathcal{A}) and D(M(\mathcal{A}),\mathbb{C}) denote its derived category. Also D^{b}(M(\mathcal{A}),\mathbb{C}) denote the full subcategory consisting of bounded complexes in D(M(\mathcal{A}),\mathbb{C}) . Let \mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) denote the full subcategory of \mathrm{Sh}(M(\mathcal{A}),\mathbb{C}) consisting of locally constant sheaves of finite dimensional \mathbb{C} -vector spaces, i.e., the category of finite dimensional local systems over M(\mathcal{A}) . Then we also define a full subcategory of D^{b}(M(\mathcal{A}),\mathbb{C}) by D_{\mathrm{loc}}^{b}(M(\mathcal{A}),\mathbb{C}):=\{\mathcal{F}\in D^{b}(M(\mathcal{A}),\mathbb{C})\mid\mathsf{H}^{k}(\mathcal{F})\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C})\text{ for all }k\in\mathbb{Z}\}. Here \mathsf{H}^{k}\colon D(M(\mathcal{A}),\mathbb{C})\rightarrow\mathrm{Sh}(M(\mathcal{A}),\mathbb{C}) is the k -th cohomology functor. 2.1 Complement of simple Weierstrass polynomial This section introduces simple Weierstrass polynomials and recall some properties of their complements. Let X be a path-connected topological space with the homotopy type of a CW-complex. A simple Weierstrass polynomial was introduced by Hansen (cf. [ 11 ] ) as a function f\colon X\times\mathbb{C}\rightarrow\mathbb{C} of the form f(x,y)=y^{n}+\sum_{i=0}^{n-1}a_{i}(x)y^{i} with continuous functions a_{i}\colon X\rightarrow\mathbb{C} such that for each x_{0}\in X , the polynomial f(x_{0},y) of y has n distinct roots. In particular, if f(x,y) has the global factorization f(x,y)=\prod_{i=1}^{n}(y-t_{i}(x)) with continuous functions t_{i}\colon X\rightarrow\mathbb{C} , we say that the simple Weierstrass polynomial f is completely solvable . The complement of the zero locus C(f):=\{(x,y)\in X\times\mathbb{C}\mid f(x,y)\neq 0\} of the polynomial combined with the natural projection \pi\colon C(f)\rightarrow X was shown to be a fiber bundle as follows. Theorem 2.1 (Hansen [ 11 ] , Møller [ 16 ] , Cohen-Suciu [ 5 ] ) . The above natural projection \pi\colon C(f)\rightarrow X is a locally trivial bundle with the fiber \mathbb{C}\backslash\{n\text{-points}\} , whose structure group is the Artin braid group B_{n} of n -strings. In particular when f is completely solvable, then the structure group reduces to the pure braid group P_{n} . We call the projection \pi\colon C(f)\rightarrow X the polynomial complement fibration associated to the simple Weierstrass polynomial f . 2.2 Middle convolution on complements with a good line We now return to the complement M(\mathcal{A}) of an affine arrangement \mathcal{A} with a good line Y . Let us take the pullback of the fibration \pi_{Y}\colon M(\mathcal{A})\rightarrow M(p\mathcal{A}_{Y}) by itself, \hbox to172.57pt{\vbox to52.87pt{\pgfpicture\makeatletter\hbox{\hskip 86.28574pt\lower-26.32225pt\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke} \lxSVG@begingroup@{fill} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces\nullfont\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} {}{}{}{{}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{{}}{{}}{{}}{{}}}{{{\lx@inpgf@ignorespaces}}}{{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-86.28574pt}{-20.16252pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}}{{{\lx@inpgf@ignorespaces{}}}{{}}{{}}{{}}{{}}}} \lxSVG@closescope }}} {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 8.1 24.44 L 48.47 24.44}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{35.22812pt}{17.66254pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{14.9173pt}{21.37639pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -55.79 12.17 L -55.79 -11.63}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{-40.31735pt}{-8.60281pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-37.96458pt}{-0.82916pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 80.21 12.18 L 80.21 -11.63}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{57.96837pt}{-8.60281pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{60.32114pt}{-1.00415pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -24.6 -24.44 L 40.2 -24.44}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{29.25102pt}{-17.66252pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{0.49356pt}{-14.3042pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope \lxSVG@closescope {\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}\hss}\lxSVG@discardpath\lxSVG@closescope \hss}}\lxSVG@closescope\endpgfpicture}}. Then since \pi_{Y} is a locally trivial fibration by Theorem 1.8 , projections \mathrm{pr}_{i} , i=1,2 , are also locally trivial fibrations. Let us look at the fibrations \mathrm{pr}_{i} , i=1,2 , closer. Under the description M(\mathcal{A})=\{(\mathbf{x},y)\in\mathbb{C}^{l-1}\times\mathbb{C}\mid Q_{\mathcal{A}}(\mathbf{x},y)\neq 0\} , the pullback is written as \displaystyle M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A}) \displaystyle\cong\{((\mathbf{x},y),(\mathbf{x}^{\prime},y^{\prime}))\in(\mathbb{C}^{l-1}\times\mathbb{C})^{2}\mid Q_{\mathcal{A}}(\mathbf{x},y)\neq 0,Q_{\mathcal{A}}(\mathbf{x}^{\prime},y^{\prime})\neq 0,\mathbf{x}=\mathbf{x}^{\prime}\} \displaystyle\cong\{((\mathbf{x},y),z)\in M(\mathcal{A})\times\mathbb{C}\mid Q_{\mathcal{A}}(\mathbf{x},z)\neq 0\}. We moreover consider the open subspace of M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A}) by removing the diagonal, \displaystyle M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta \displaystyle:=\{(m_{1},m_{2})\in M(\mathcal{A})^{2}\mid m_{1}\neq m_{2},\ \pi_{Y}(m_{1})=\pi_{Y}(m_{2})\} \displaystyle=\{((\mathbf{x},y),z)\in M(\mathcal{A})\times\mathbb{C}\mid Q_{\mathcal{A}}(\mathbf{x},z)(z-y)\neq 0\} \displaystyle=\{((\mathbf{x},y),z)\in M(\mathcal{A})\times\mathbb{C}\mid Q^{\mathrm{red}}_{\mathcal{A}}(\mathbf{x},z)(z-y)\neq 0\}. Here Q^{\mathrm{red}}_{\mathcal{A}}(\mathbf{x},y):=\prod_{H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y}}(L_{H^{\prime}}(\mathbf{x},y)+\alpha_{H^{\prime}}) from the decomposition Q_{\mathcal{A}}(\mathbf{x},y)=\prod_{H\in\mathcal{A}_{Y}}(L_{H}(\mathbf{x})+\alpha_{H})\cdot\prod_{H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y}}(L_{H^{\prime}}(\mathbf{x},y)+\alpha_{H^{\prime}}) of the defining polynomial Q_{\mathcal{A}}(\mathbf{x},y) . Then, since Q^{\mathrm{red}}_{\mathcal{A}}(\mathbf{x},z)(z-y) is a simple Weierstrass polynomial in the variable z over the base space M(\mathcal{A}) , we obtain the following. Proposition 2.2 . The subspace M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta of the pullback M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A}) with the natural projections \mathrm{pr_{i}}\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta\rightarrow M(\mathcal{A}),\quad i=1,2 are locally trivial fibrations with the fiber \mathbb{C} with |\mathcal{A}\backslash\mathcal{A}_{Y}|+1 points removed. Proof. Theorem 2.1 directly shows the claim. ∎ The middle convolution was introduced by Katz [ 14 ] as an endofunctor on the category of local systems on the Riemann sphere with a finite number of punctures. We now give an generalization of the middle convolution functor for local systems over M(\mathcal{A}) . Let us take a nontrivial multiplicative character \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} and consider the associated rank 1 local system \mathcal{K}_{\chi} on \mathbb{C}^{\times} with some fixed base point in \mathbb{C}^{\times} . Then for a local system \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) on M(\mathcal{A}) , we can define the external tensor product of them through the projection maps, \mathrm{pr}_{1}\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta\rightarrow M(\mathcal{A}) and (z-y)\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta\ni((\mathbf{x},y),z)\mapsto z-y\in\mathbb{C}^{\times}. Namely, \mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)}:=\mathrm{pr}_{1}^{*}\mathcal{L}\otimes(z-y)^{*}\mathcal{K}_{\chi}. Regard \mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)} as an object in D_{\mathrm{loc}}^{b}(M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta,\mathbb{C}) in the natural way. Then, since \mathrm{pr}_{2}\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta\rightarrow M(\mathcal{A}) is a locally trivial fibration, the direct image functors \mathrm{pr}_{2\,*} and \mathrm{pr}_{2\,!} define the well-defined right derived functors (cf. Theorem 1.9.5 in [ 1 ] ) R\mathrm{pr}_{2\,*},R\mathrm{pr}_{2\,!}\colon D_{\mathrm{loc}}^{b}(M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta,\mathbb{C})\rightarrow D_{\mathrm{loc}}^{b}(M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta,\mathbb{C}). We can consequently define the local system by \mathrm{MC}_{\chi}(\mathcal{L}):=\mathrm{Im}(R^{1}\mathrm{pr}_{2\,!}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)})\rightarrow R^{1}\mathrm{pr}_{2\,*}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)}))\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}). Definition 2.3 (Middle convolution) . Fix a nontrivial multiplicative character \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} . Then the middle convolution functor with respect to \chi is defined by \mathrm{MC}_{\chi}\colon\mathrm{Loc}(M(\mathcal{A}),\mathbb{C})\rightarrow\mathrm{Loc}(M(\mathcal{A}),\mathbb{C});\quad\mathcal{L}\mapsto\mathrm{MC}_{\chi}(\mathcal{L}). For simplicity, we also use the following notations \displaystyle C_{\chi\,*}(\mathcal{L}):= \displaystyle R^{1}\mathrm{pr}_{2\,*}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)}), \displaystyle C_{\chi\,!}(\mathcal{L}):= \displaystyle R^{1}\mathrm{pr}_{2\,!}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)}). 2.3 Comparison with Katz middle convolution Let us recall the Katz middle convolution for \mathbb{C}\backslash\{n\text{-points}\} by following [ 14 ] . Take \mathcal{L}\in\mathrm{Loc}(\mathbb{C}\backslash\{n\text{-points}\},\mathbb{C}) and a nontrivial character \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} . Let \left(\mathbb{C}\backslash\{n\text{-points}\}\right)^{2}\backslash\Delta be the copy of \mathbb{C}\backslash\{n\text{-points}\} , the diagonal removed. Then the projection on each component, \mathrm{pr}_{i}\colon\left(\mathbb{C}\backslash\{n\text{-points}\}\right)^{2}\backslash\Delta\rightarrow\mathbb{C}\backslash\{n\text{-points}\} , is known to be a locally trivial fibration. We also consider the projection (z-y)\colon\left(\mathbb{C}\backslash\{n\text{-points}\}\right)^{2}\backslash\Delta\ni(y,z)\mapsto z-y\in\mathbb{C}^{\times}. By taking the exterior tensor product \mathcal{L}_{y}\boxtimes\chi_{(z-y)}:=\mathrm{pr}_{1}^{*}\mathcal{L}\otimes(z-y)^{*}\mathcal{K}_{\chi}\in\mathrm{Loc}(\left(\mathbb{C}\backslash\{n\text{-points}\}\right)^{2}\backslash\Delta,\mathbb{C}), we then define the local system \mathrm{MC}_{\chi}^{\mathrm{Katz}}(\mathcal{L}):=\mathrm{Im}(R^{1}\mathrm{pr}_{2\,!}(\mathcal{L}_{y}\boxtimes\chi_{(z-y)})\rightarrow R^{1}\mathrm{pr}_{2\,*}(\mathcal{L}_{y}\boxtimes\chi_{(z-y)})) on \mathbb{C}\backslash\{n\text{-points}\} , similarly as in the previous section. The Katz middle convolution is the endofunctor defined by \mathrm{MC}_{\chi}^{\mathrm{Katz}}\colon\mathrm{Loc}(\mathbb{C}\backslash\{n\text{-points}\},\mathbb{C})\rightarrow\mathrm{Loc}(\mathbb{C}\backslash\{n\text{-points}\},\mathbb{C});\quad\mathcal{L}\mapsto\mathrm{MC}_{\chi}^{\mathrm{Katz}}(\mathcal{L}). We now compare the Katz middle convolution and the one for M(\mathcal{A}) . Let us fix a base point \mathbf{x}_{0}\in M(p\mathcal{A}_{Y}) from the base space of the fibration \pi_{Y}\colon M(\mathcal{A})\rightarrow M(p\mathcal{A}_{Y}) . Then the fiber \pi_{Y}^{-1}(\mathbf{x}_{0}) is written as \pi_{Y}^{-1}(\mathbf{x}_{0})=\{(\mathbf{x}_{0},y)\in M(p\mathcal{A}_{Y})\times\mathbb{C}\mid Q_{\mathcal{A}}^{\mathrm{red}}(\mathbf{x}_{0},y)\neq 0\}. Therefore, denoting the set of n=|\mathcal{A}\backslash\mathcal{A}_{Y}| -distinct roots of Q_{\mathcal{A}}^{\mathrm{red}}(\mathbf{x}_{0},y) by Q_{n}^{\mathbf{x}_{0}}:=\{y\in\mathbb{C}\mid Q_{\mathcal{A}}^{\mathrm{red}}(\mathbf{x}_{0},y)=0\}, we can identify the fiber with \mathbb{C}\backslash Q_{n}^{\mathbf{x}_{0}}. The closed embedding \iota_{\mathbf{x}_{0}}\colon\pi_{Y}^{-1}(\mathbf{x}_{0})=\mathbb{C}\backslash Q_{n}^{\mathbf{x}_{0}}\hookrightarrow M(\mathcal{A}). of the fiber yields the cartesian squares, {\lx@inpgf@ignorespaces\left(\mathbb{C}\backslash Q_{n}^{\mathbf{x}_{0}}\right)^{2}\backslash\Delta} {\lx@inpgf@ignorespaces M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta} {\lx@inpgf@ignorespaces\mathbb{C}\backslash Q_{n}^{\mathbf{x}_{0}}} {\lx@inpgf@ignorespaces M(\mathcal{A})} \scriptstyle{\lx@inpgf@ignorespaces\widetilde{\iota}_{\mathbf{x}_{0}}} \scriptstyle{\lx@inpgf@ignorespaces\mathrm{pr}_{i}} \scriptstyle{\lx@inpgf@ignorespaces\mathrm{pr}_{i}} \scriptstyle{\lx@inpgf@ignorespaces\iota_{\mathbf{x}_{0}}} for i=1,2 . Then by recalling that the pullback functor \iota_{\mathbf{x}_{0}}^{*} is an exact functor and the projections \mathrm{pr}_{i} , i=1,2 , in Proposition 2.2 are locally trivial, we obtain the following isomorphisms, \displaystyle\iota_{\mathbf{x}_{0}}^{*}\mathrm{MC}_{\chi}(\mathcal{L}) \displaystyle=\iota_{\mathbf{x}_{0}}^{*}\mathrm{Im}(R^{1}\mathrm{pr}_{2\,!}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)})\rightarrow R^{1}\mathrm{pr}_{2\,*}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)})) \displaystyle\cong\mathrm{Im}(\iota_{\mathbf{x}_{0}}^{*}R^{1}\mathrm{pr}_{2\,!}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)})\rightarrow\iota_{\mathbf{x}_{0}}^{*}R^{1}\mathrm{pr}_{2\,*}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)})) \displaystyle\cong\mathrm{Im}(R^{1}\mathrm{pr}_{2\,!}(\widetilde{\iota}_{\mathbf{x}_{0}})^{*}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)})\rightarrow R^{1}\mathrm{pr}_{2\,*}(\widetilde{\iota}_{\mathbf{x}_{0}})^{*}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)})). Here for the last isomorphism, we used the proper base change isomorphism \iota_{\mathbf{x}_{0}}^{*}R^{1}\mathrm{pr}_{2\,!}\cong R^{1}\mathrm{pr}_{2\,!}(\widetilde{\iota}_{\mathbf{x}_{0}})^{*} and the base change isomorphism for locally trivial fibrations \iota_{\mathbf{x}_{0}}^{*}R^{1}\mathrm{pr}_{2\,*}\cong R^{1}\mathrm{pr}_{2\,*}(\widetilde{\iota}_{\mathbf{x}_{0}})^{*} . The above diagram moreover gives the following isomorphisms, (\widetilde{\iota}_{\mathbf{x}_{0}})^{*}(\mathcal{L}_{\mathbf{x},y}\boxtimes\chi_{(z-y)})=(\widetilde{\iota}_{\mathbf{x}_{0}})^{*}(\mathrm{pr}_{1}^{*}\mathcal{L}\otimes(z-y)^{*}\chi)\\ \cong\mathrm{pr}_{1}^{*}\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}\otimes\iota_{\mathbf{x}_{0}}^{*}(z-y)^{*}\chi=(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)}. Therefore combining these isomorphisms, we obtain the following. Proposition 2.4 . For a nontrivial character \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} , there exists an isomorphism of functors \iota_{\mathbf{x}_{0}}^{*}\circ\mathrm{MC}_{\chi}\cong\mathrm{MC}_{\chi}^{\mathrm{Katz}}\circ\iota_{\mathbf{x}_{0}}^{*}. Let us look at stalks of the middle convolution functor. Let us take m_{0}=(\mathbf{x}_{0},y_{0})\in M(\mathcal{A}) and consider the inclusion maps i_{m_{0}}\colon\{m_{0}\}\hookrightarrow M(\mathcal{A}) , i_{m_{0}}^{\mathbf{x}_{0}}\colon\{m_{0}\}\hookrightarrow\pi_{Y}^{-1}(\mathbf{x}_{0})=\mathbb{C}\backslash Q_{n}^{\mathbf{x}_{0}} . Then by taking into account the factorization i_{m_{0}}=\iota_{\mathbf{x}_{0}}\circ i_{m_{0}}^{\mathbf{x}_{0}} , we obtain the following diagram, \hbox to299.99pt{\vbox to58.71pt{\pgfpicture\makeatletter\hbox{\hskip 149.99205pt\lower-26.64226pt\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke} \lxSVG@begingroup@{fill} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces\nullfont\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} {}{}{}{{}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{{}}{{}}{{}}{{}}{{}}{{}}}{{{\lx@inpgf@ignorespaces}}}{{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-149.99205pt}{-20.48253pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}}{{{\lx@inpgf@ignorespaces{}}}{{}}{{}}{{}}{{}}{{}}{{}}}} \lxSVG@closescope }}} {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -116.29 24 L -86.18 24}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{-1.0}{0.0}{0.0}{1.0}{-84.04495pt}{17.34253pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@begingroup@{stroke-linecap} \lxSVG@drawpath@unclipped{M 0 2.71 C 0.95 2.71 1.72 2.1 1.72 1.36 C 1.72 0.61 0.95 0 0 0}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-62.08487pt}{17.34253pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-81.73415pt}{22.69527pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope { {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -163.05 11.74 L -163.05 -12.07}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{-117.83846pt}{-8.92282pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 24.79 24 L 54.9 24}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{-1.0}{0.0}{0.0}{1.0}{17.91751pt}{17.34253pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@begingroup@{stroke-linecap} \lxSVG@drawpath@unclipped{M 0 2.71 C 0.95 2.71 1.72 2.1 1.72 1.36 C 1.72 0.61 0.95 0 0 0}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{39.8776pt}{17.34253pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{21.3019pt}{22.69527pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -31.42 11.74 L -31.42 -12.07}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{-22.70364pt}{-8.92282pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-20.35088pt}{-1.1464pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 131.64 11.73 L 131.64 -12.07}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{95.13481pt}{-8.92282pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{97.48758pt}{-1.14917pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -133.8 -24.88 L -66.94 -24.88}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-48.17513pt}{-17.98253pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-79.02359pt}{-13.12979pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 5.55 -24.88 L 99.9 -24.88}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{-1.0}{0.0}{0.0}{1.0}{4.00777pt}{-17.98253pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@begingroup@{stroke-linecap} \lxSVG@drawpath@unclipped{M 0 2.71 C 0.95 2.71 1.72 2.1 1.72 1.36 C 1.72 0.61 0.95 0 0 0}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{72.39456pt}{-17.98253pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{31.90353pt}{-13.12979pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope \lxSVG@closescope {\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}\hss}\lxSVG@discardpath\lxSVG@closescope \hss}}\lxSVG@closescope\endpgfpicture}}, where all squares are cartesian. Here we put Q_{n+1}^{m_{0}}:=Q_{n}^{\mathbf{x}_{0}}\sqcup\{y_{0}\} , and the leftmost vertical arrow is the canonical map to the singleton set \{m_{0}\} which is the terminal object of the category of topological spaces. Then by noting that all vertical arrows are locally trivial fibrations, base change theorems yield the following isomorphism, \displaystyle\mathrm{MC}_{\chi}(\mathcal{L})_{m_{0}}=i_{m_{0}}^{*}\mathrm{MC}_{\chi}(\mathcal{L})=(i_{m_{0}}^{\mathbf{x}_{0}})^{*}\mathrm{MC}^{\mathrm{Katz}}_{\chi}(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}) \displaystyle\cong\mathrm{Im}(H^{1}_{c}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\rightarrow H^{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})). Here we regard (\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)} as the local system on \mathbb{C}\backslash Q_{n+1}^{m_{0}} through the pullback along the inclusion \widetilde{i}_{m_{0}}^{\mathbf{x}_{0}} . Remark 2.5 . By Artin’s vanishing theorem, we know that H^{i}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})=0\quad\text{ if }i\neq 0,1. Moreover if \chi is non-trivial and \mathcal{L} is nonzero, then the local system (\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)} is non-trivial as well. Therefore we also have the vanishing of H^{0} as well. For the compactly supported cohomology groups, it is also known that H^{i}_{c}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})=0\quad\text{ for }i\neq 1, see Proposition B.3.4 in [ 1 ] for example. 2.4 Composition low for middle convolution Let us recall the composition law for the Katz middle convolution. Let Q_{n}\subset\mathbb{C} be a set of distinct n -points and j\colon\mathbb{C}\backslash Q_{n}\hookrightarrow\mathbb{C} be the inclusion map. Definition 2.6 (Property \wp ) . For \mathcal{L}\in\mathrm{Loc}(\mathbb{C}\backslash Q_{n},\mathbb{C}) , we say \mathcal{L} has the property \wp if j_{*}\mathcal{L} does not have the constant sheaf \mathbb{C} as a subquotient or a subobject. 1 1 1 Lemmas 2.6.13, 2.6.14, and 2.6.15 in [ 14 ] assure that this property \wp implies the original one defined in (2.6.2) in [ 14 ] . Theorem 2.7 (Katz) . Let \chi,\varphi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} be nontrivial multiplicative characters. Also let \mathcal{L}\in\mathrm{Loc}(\mathbb{C}\backslash Q_{n},\mathbb{C}) be a local system with the property \wp . Then the following hold. 1. The middle convolution preserves the property \wp , i.e., \mathrm{MC}_{\chi}^{\mathrm{Katz}}(\mathcal{L}) has the property \wp . 2. There exist isomorphisms as local systems \displaystyle\mathrm{MC}^{\mathrm{Katz}}_{\varphi}\circ\mathrm{MC}^{\mathrm{Katz}}_{\chi}(\mathcal{L}) \displaystyle\cong\mathrm{MC}^{\mathrm{Katz}}_{\chi\cdot\varphi}(\mathcal{L})\quad\text{ if }\chi\cdot\varphi\text{ is nontrivial}, \displaystyle\mathrm{MC}^{\mathrm{Katz}}_{\chi^{-1}}\circ\mathrm{MC}^{\mathrm{Katz}}_{\chi}(\mathcal{L}) \displaystyle\cong\mathcal{L}. The purpose of this section is to see that the above composition law also holds for the middle convolution functor on M(\mathcal{A}) . For this purpose, we consider a reformulation of the middle convolution functor as a functor on the derived category D_{\mathrm{loc}}^{b}(M(\mathcal{A}),\mathbb{C}) . Consider the full subcategory of D_{\mathrm{loc}}^{b}(M(\mathcal{A}),\mathbb{C}) consisting of shifted local systems, \mathrm{Perv}(M(\mathcal{A}),\mathbb{C}):=\{\mathcal{L}[l]\mid\mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}))\}, where l=\mathrm{dim\,}M(\mathcal{A}) . Take a nontrivial multiplicative character \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} . Then for an object \mathcal{L}[l]\in\mathrm{Perv}(M(\mathcal{A}),\mathbb{C}) represented by a local system \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) , we can define following objects in D_{\mathrm{loc}}^{b}(M(\mathcal{A}),\mathbb{C}) , \displaystyle C_{\chi\,!}(\mathcal{L}[l]) \displaystyle:=R\mathrm{pr}_{2\,!}[-1]((\mathcal{L}[l])_{\mathbf{x},y}\boxtimes\mathcal{K}_{\chi}[1]_{(z-y)}), \displaystyle C_{\chi\,*}(\mathcal{L}[l]) \displaystyle:=R\mathrm{pr}_{2\,*}[-1]((\mathcal{L}[l])_{\mathbf{x},y}\boxtimes\mathcal{K}_{\chi}[1]_{(z-y)}). Here \boxtimes stands for the external tensor product with respect to \mathrm{pr}_{1}\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta\rightarrow M(\mathcal{A}) and (z-y)\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta\rightarrow\mathbb{C}^{\times} . The vanishing results in Remark 2.5 yield the quasi-isomorphisms \displaystyle C_{\chi\,!}(\mathcal{L}[l]) \displaystyle\simeq R^{1}\mathrm{pr}_{2\,!}(\mathrm{pr}_{1}^{*}\mathcal{L}\otimes(z-y)^{*}\mathcal{K}_{\chi})[l], \displaystyle C_{\chi\,*}(\mathcal{L}[l]) \displaystyle\simeq R^{1}\mathrm{pr}_{2\,*}(\mathrm{pr}_{1}^{*}\mathcal{L}\otimes(z-y)^{*}\mathcal{K}_{\chi})[l], which assures that C_{\chi\,!} and C_{\chi\,*} define endofunctors on \mathrm{Prev}(M(\mathcal{A}),\mathbb{C}) . Therefore by taking the image in the abelian category \mathrm{Perv}(M(\mathcal{A}),\mathbb{C}) , we can reformulate the middle convolution functor as follows, \mathrm{MC}_{\chi}\colon\mathrm{Perv}(M(\mathcal{A}),\mathbb{C})\rightarrow\mathrm{Perv}(M(\mathcal{A}),\mathbb{C});\quad\mathcal{L}[l]\mapsto\mathrm{Im}(C_{\chi\,!}(\mathcal{L}[l])\rightarrow C_{\chi\,*}(\mathcal{L}[l])). We moreover need a slight modification. For the closed subspace \Delta:=\{(\mathbf{x},y,y)\in M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\} of M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A}) , we consider the commutative diagrams for i=1,2 , \hbox to294.34pt{\vbox to53.35pt{\pgfpicture\makeatletter\hbox{\hskip 147.16586pt\lower-26.32225pt\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke} \lxSVG@begingroup@{fill} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces\nullfont\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} {}{}{}{{}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{{}}{{}}{{}}{{}}{{}}{{}}}{{{\lx@inpgf@ignorespaces}}}{{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-147.16586pt}{-20.16252pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}}{{{\lx@inpgf@ignorespaces{}}}{{}}{{}}{{}}{{}}{{}}{{}}}} \lxSVG@closescope }}} {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -49.55 24.44 L -19.44 24.44}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{-1.0}{0.0}{0.0}{1.0}{-35.81148pt}{17.66254pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@begingroup@{stroke-linecap} \lxSVG@drawpath@unclipped{M 0 2.71 C 0.95 2.71 1.72 2.1 1.72 1.36 C 1.72 0.61 0.95 0 0 0}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-13.8514pt}{17.66254pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-27.26877pt}{20.0153pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -127.73 12.17 L -127.73 -11.63}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{-92.30862pt}{-8.60281pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-89.95586pt}{-0.82916pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 44.99 12.17 L 44.99 -11.63}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{32.51697pt}{-8.60281pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{34.86974pt}{-2.09164pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 109.43 24.44 L 151.81 24.44}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{-1.0}{0.0}{0.0}{-1.0}{78.88533pt}{17.66254pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{109.71347pt}{17.66254pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@begingroup@{stroke-linecap} \lxSVG@drawpath@unclipped{M 0 2.71 C 0.95 2.71 1.72 2.1 1.72 1.36 C 1.72 0.61 0.95 0 0 0}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{92.85927pt}{21.37639pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 172.72 15.64 L 172.72 -11.63}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{124.82559pt}{-8.60281pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{127.17836pt}{-1.6111pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope { {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -96.54 -24.44 L 13.8 -24.44}{fill:none} \lxSVG@begingroup@{_scopebegin} {\lxSVG@setlinewidth{\pgfinnerlinewidth}\lxSVG@begingroup@{stroke-width} \lxSVG@begingroup@{stroke} \lxSVG@stroke\lxSVG@drawpath@unclipped{M -96.54 -24.44 L 13.8 -24.44}{fill:none} }\lxSVG@closescope {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\lxSVG@closescope \lxSVG@closescope { {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 141.53 -24.44 L 76.18 -24.44}{fill:none} \lxSVG@begingroup@{_scopebegin} {\lxSVG@setlinewidth{\pgfinnerlinewidth}\lxSVG@begingroup@{stroke-width} \lxSVG@begingroup@{stroke} \lxSVG@stroke\lxSVG@drawpath@unclipped{M 141.53 -24.44 L 76.18 -24.44}{fill:none} }\lxSVG@closescope {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\lxSVG@closescope \lxSVG@closescope \lxSVG@closescope {\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}\hss}\lxSVG@discardpath\lxSVG@closescope \hss}}\lxSVG@closescope\endpgfpicture}}, where the top horizontal arrows are the natural inclusions. Also consider the inclusion j\colon\mathbb{C}^{\times}\hookrightarrow\mathbb{C} and the projection \overline{(z-y)}\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\ni(\mathbf{x},y,z)\mapsto z-y\in\mathbb{C}. Then we define \displaystyle\overline{C}_{\chi\,*}(\mathcal{L}[l]):=\mathcal{L}[l]*_{*}j_{*}\mathcal{K}_{\chi}[1] \displaystyle:=R\overline{\mathrm{pr}}_{2\,*}[-1]((\mathcal{L}[l])_{\mathbf{x},y}\boxtimes(j_{*}\mathcal{K}_{\chi}[1])_{z-y}), \displaystyle\overline{C}_{\chi\,!}(\mathcal{L}[l]):=\mathcal{L}[l]*_{!}j_{*}\mathcal{K}_{\chi}[1] \displaystyle:=R\overline{\mathrm{pr}}_{2\,!}[-1]((\mathcal{L}[l])_{\mathbf{x},y}\boxtimes(j_{*}\mathcal{K}_{\chi}[1])_{z-y}), for \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) . Here \boxtimes stands for the external tensor product with respect to the projections \overline{\mathrm{pr}}_{1}\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\rightarrow M(\mathcal{A}) and \overline{(z-y)}\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\rightarrow\mathbb{C} . Now we can relate these functors with the previous ones as follows. The adjunction natural transforms, \delta_{!}\delta^{*}\rightarrow\mathrm{id} , \mathrm{id}\rightarrow\delta_{*}\delta^{*} , yield the natural morphisms C_{\chi\,!}(\mathcal{L}[l])\cong R\overline{\mathrm{pr}}_{2\,!}[-1]\delta_{!}\delta^{*}(\overline{\mathrm{pr}}_{1}^{*}\mathcal{L}[l]\otimes\overline{(z-y)}^{*}j_{*}\mathcal{K}_{\chi}[1])\\ \rightarrow R\overline{\mathrm{pr}}_{2\,!}[1](\overline{\mathrm{pr}}_{1}^{*}\mathcal{L}[l]\otimes\overline{(z-y)}^{*}j_{*}\mathcal{K}_{\chi}[1])\cong\overline{C}_{\chi\,!}(\mathcal{L}[l]), \overline{C}_{\chi\,*}(\mathcal{L}[l])=R\overline{\mathrm{pr}}_{2\,*}[1](\overline{\mathrm{pr}}_{1}^{*}[-1]\mathcal{L}[l]\otimes\overline{(z-y)}^{*}j_{*}\mathcal{K}_{\chi})\\ \rightarrow R\overline{\mathrm{pr}}_{2\,*}[1]\delta_{*}\delta^{*}(\overline{\mathrm{pr}}_{1}^{*}[-1]\mathcal{L}[l]\otimes\overline{(z-y)}^{*}j_{*}\mathcal{K}_{\chi})=C_{\chi\,*}(\mathcal{L}[l]). Proposition 2.8 . For a nontrivial multiplicative character \chi , the above morphisms C_{\chi\,!}(\mathcal{L}[l])\rightarrow\overline{C}_{\chi\,!}(\mathcal{L}[l]) and \overline{C}_{\chi\,*}(\mathcal{L}[l])\rightarrow C_{\chi\,*}(\mathcal{L}[l]) are isomorphisms. Proof. Consider the cartesian square \hbox to149.55pt{\vbox to52.87pt{\pgfpicture\makeatletter\hbox{\hskip 74.77393pt\lower-26.32225pt\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke} \lxSVG@begingroup@{fill} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces\nullfont\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} {}{}{}{{}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{{}}{{}}{{}}{{}}}{{{\lx@inpgf@ignorespaces}}}{{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-74.77393pt}{-20.16252pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}}{{{\lx@inpgf@ignorespaces{}}}{{}}{{}}{{}}{{}}}} \lxSVG@closescope }}} {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 24.58 24.44 L 59.3 24.44}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{-1.0}{0.0}{0.0}{-1.0}{17.56282pt}{17.66254pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{42.85625pt}{17.66254pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@begingroup@{stroke-linecap} \lxSVG@drawpath@unclipped{M 0 2.71 C 0.95 2.71 1.72 2.1 1.72 1.36 C 1.72 0.61 0.95 0 0 0}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{28.76941pt}{21.37639pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -39.86 12.17 L -39.86 -12.55}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{-28.80554pt}{-9.2695pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-26.45277pt}{-3.17499pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope { {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 80.21 15.64 L 80.21 -11.63}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{57.96837pt}{-8.60281pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -21.16 -24.44 L 54.69 -24.44}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{-1.0}{0.0}{0.0}{-1.0}{-15.48892pt}{-17.66252pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{39.5229pt}{-17.66252pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@begingroup@{stroke-linecap} \lxSVG@drawpath@unclipped{M 0 2.71 C 0.95 2.71 1.72 2.1 1.72 1.36 C 1.72 0.61 0.95 0 0 0}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{9.27089pt}{-14.30975pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope \lxSVG@closescope {\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}\hss}\lxSVG@discardpath\lxSVG@closescope \hss}}\lxSVG@closescope\endpgfpicture}}. By recalling that i_{0}^{*}(j_{*}\mathcal{K}_{\chi})=0 since \chi is nontrivial, we obtain \eta^{*}(\overline{\mathrm{pr}}_{1}^{*}\mathcal{L}[l]\otimes\overline{(z-y)}^{*}j_{*}\mathcal{K}_{\chi})=0. Therefore the sequences of adjunction morphisms \delta_{!}\delta^{*}\rightarrow\mathrm{id}\rightarrow R\eta_{*}\eta^{*} and \eta_{*}\eta^{!}\rightarrow\mathrm{id}\rightarrow R\delta_{*}\delta^{*} yield the distinguished triangles C_{\chi\,!}(\mathcal{L}[l])\rightarrow\overline{C}_{\chi\,!}(\mathcal{L}[l])\rightarrow 0\rightarrow,\quad 0\rightarrow\overline{C}_{\chi\,*}(\mathcal{L}[l])\rightarrow C_{\chi\,*}(\mathcal{L}[l])\rightarrow, where we used the equation \eta^{*}=\eta^{!} as functors on constructible sheaves which follows from the smoothness of \eta . Thus we obtain the claim. ∎ Therefore \mathcal{L}[l]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1]:=\mathrm{Im}(\mathcal{L}[l]*_{!}j_{*}\mathcal{K}_{\chi}[1]\rightarrow\mathcal{L}[l]*_{*}j_{*}\mathcal{K}_{\chi}[1]). is isomorphic to the middle convolution \mathrm{MC}_{\chi}(\mathcal{L}[l]) defined above. We now define the property \wp for \mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) . Definition 2.9 (Property \wp ) . We say \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) has the property \wp if for each \mathbf{x}\in M(p\mathcal{A}_{Y}) , the pullback \iota_{\mathbf{x}}^{*}\mathcal{L} along the inclusion \iota_{\mathbf{x}}\colon\pi_{Y}^{-1}(\mathbf{x})\hookrightarrow M(\mathcal{A}) has the property \wp defined in Definition 2.6 . Remark 2.10 . Since all fibers of the fibration \pi_{Y}\colon M(\mathcal{A})\rightarrow M(p\mathcal{A}_{Y}) are isomorphic, it suffice to check the property \wp at a single fiber. That is, a given local system \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) has the property \wp if and only if for a fixed point \mathbf{x}_{0}\in M(p\mathcal{A}_{Y}) , the pullback \iota_{\mathbf{x}_{0}}^{*}\mathcal{L} has the property \wp . We also use the following notation, \displaystyle j_{*}\mathcal{K}_{\chi}[1]*_{!}j_{*}\mathcal{K}_{\varphi}[1] \displaystyle:=R\overline{\mathrm{pr}}_{2\,!}[-1]((j_{*}\mathcal{K}_{\chi}[1])_{y}\boxtimes(j_{*}\mathcal{K}_{\varphi}[1])_{z-y}), \displaystyle j_{*}\mathcal{K}_{\chi}[1]*_{*}j_{!}\mathcal{K}_{\varphi}[1] \displaystyle:=R\overline{\mathrm{pr}}_{2\,*}[-1]((j_{*}\mathcal{K}_{\chi}[1])_{y}\boxtimes(j_{*}\mathcal{K}_{\varphi}[1])_{z-y}), \displaystyle j_{*}\mathcal{K}_{\chi}[1]*_{\mathrm{mid}}j_{!}\mathcal{K}_{\varphi}[1] \displaystyle:=\mathrm{Im}(j_{*}\mathcal{K}_{\chi}[1]*_{!}j_{*}\mathcal{K}_{\varphi}[1]\rightarrow j_{*}\mathcal{K}_{\chi}[1]*_{*}j_{*}\mathcal{K}_{\varphi}[1]), for nontrivial characters \chi,\varphi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} and the projections \overline{\mathrm{pr}}_{i}\colon\mathbb{C}^{2}\rightarrow\mathbb{C} , i=1,2 , and \overline{(z-y)}\colon\mathbb{C}^{2}\ni(y,z)\mapsto z-y\in\mathbb{C} . Then the following is the analogue of Proposition 2.6.5 in [ 14 ] which is a key ingredient for the proof of the composition law. Proposition 2.11 . Let us take \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) with the property \wp , and nontrivial characters \chi,\varphi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} . Then there exist an isomorphism (\mathcal{L}[l]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1])*_{\mathrm{mid}}j_{*}\mathcal{K}_{\varphi}[1]\cong\mathcal{L}[l]*_{\mathrm{mid}}(j_{*}\mathcal{K}_{\chi}[1]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\varphi}[1]) Proof. We follow the argument in Proposition 2.6.5 in [ 14 ] . First note that from the compatibility of the external tensor product and the direct image functors for constructible sheaves, we have the isomorphisms \displaystyle(\mathcal{L}[l]*_{!}j_{*}\mathcal{K}_{\chi}[1])*_{!}j_{*}\mathcal{K}_{\varphi}[1] \displaystyle\cong\mathcal{L}[l]*_{!}(j_{*}\mathcal{K}_{\chi}[1]*_{!}j_{*}\mathcal{K}_{\varphi}[1]), \displaystyle(\mathcal{L}[l]*_{*}j_{*}\mathcal{K}_{\chi}[1])*_{*}j_{*}\mathcal{K}_{\varphi}[1] \displaystyle\cong\mathcal{L}[l]*_{*}(j_{*}\mathcal{K}_{\chi}[1]*_{*}j_{*}\mathcal{K}_{\varphi}[1]). Let us consider the sequence of natural morphisms \mathcal{L}[l]*_{!}j_{*}\mathcal{K}_{\chi}[1]*_{!}j_{*}\mathcal{K}_{\varphi}[1]\rightarrow(\mathcal{L}[l]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1])*_{!}j_{*}\mathcal{K}_{\varphi}[1]\\ \rightarrow(\mathcal{L}[l]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1])*_{\mathrm{mid}}j_{*}\mathcal{K}_{\varphi}[1]. Then the composition of these morphisms is surjective. Indeed, let us take an element m_{0}=(\mathbf{x}_{0},y_{0})\in M(\mathcal{A}) and consider the stalks at m_{0} . Then by Proposition 2.4 , the above sequence of morphisms induces the following sequence, ((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})[1]*_{!}j_{*}\mathcal{K}_{\chi}[1]*_{!}j_{*}\mathcal{K}_{\varphi}[1])_{y_{0}}\rightarrow((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})[1]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1])*_{!}j_{*}\mathcal{K}_{\varphi}[1])_{y_{0}}\\ \rightarrow((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})[1]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1])*_{\mathrm{mid}}j_{*}\mathcal{K}_{\varphi}[1])_{y_{0}}, in which all the morphisms are surjective by Proposition 2.6.5 in [ 14 ] . Next consider another sequence of morphisms (\mathcal{L}[l]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1])*_{\mathrm{mid}}j_{*}\mathcal{K}_{\varphi}[1]\rightarrow(\mathcal{L}[l]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1])*_{*}j_{*}\mathcal{K}_{\varphi}[1]\\ \rightarrow\mathcal{L}[l]*_{*}j_{*}\mathcal{K}_{\chi}[1]*_{*}j_{*}\mathcal{K}_{\varphi}[1], which is injective by taking the stalks and again applying Proposition 2.6.5 in [ 14 ] as above. Therefore (\mathcal{L}[l]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi}[1])*_{\mathrm{mid}}j_{*}\mathcal{K}_{\varphi}[1] is the image of \mathcal{L}[l]*_{!}j_{*}\mathcal{K}_{\chi}[1]*_{!}j_{*}\mathcal{K}_{\varphi}[1] in \mathcal{L}[l]*_{*}j_{*}\mathcal{K}_{\chi}[1]*_{*}j_{*}\mathcal{K}_{\varphi}[1]. Rearranging the parentheses shows that \mathcal{L}[l]*_{\mathrm{mid}}(j_{*}\mathcal{K}_{\chi}[1]*_{\mathrm{mid}}j_{*}\mathcal{K}_{\varphi}[1]) is also this image. ∎ Theorem 2.12 . Let \chi,\varphi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} be nontrivial multiplicative characters. Also let \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) be a local system with the property \wp . Then the following hold. 1. The middle convolution preserves the property \wp , i.e., \mathrm{MC}_{\chi}(\mathcal{L}[l]) has the property \wp . 2. There exist natural isomorphisms as local systems, \displaystyle\mathrm{MC}_{\varphi}\circ\mathrm{MC}_{\chi}(\mathcal{L}) \displaystyle\cong\mathrm{MC}_{\chi\cdot\varphi}(\mathcal{L})\quad\text{ if }\chi\cdot\varphi\text{ is nontrivial}, \displaystyle\mathrm{MC}_{\chi^{-1}}\circ\mathrm{MC}_{\chi}(\mathcal{L}) \displaystyle\cong\mathcal{L}. Proof. The first claim follows from Proposition 2.4 and Theorem 2.7 by looking at stalks. For the second claim, let us recall the isomorphisms obtained by Proposition 2.9.6 in [ 14 ] , \displaystyle j_{*}\mathcal{K}_{\chi}*_{\mathrm{mid}}j_{*}\mathcal{K}_{\varphi}\cong j_{*}\mathcal{K}_{\chi\cdot\varphi}\quad\text{ if }\chi\cdot\varphi\text{ is nontrivial}, \displaystyle j_{*}\mathcal{K}_{\chi}*_{\mathrm{mid}}j_{*}\mathcal{K}_{\chi^{-1}}\cong\delta_{0}, where \delta_{0} is the delta sheaf supported at \{0\}\in\mathbb{C} , the direct image \delta_{0}=i_{0\,*}\mathbb{C} along the inclusion i_{0}\colon\{0\}\hookrightarrow\mathbb{C} . Then since the delta sheaf \delta_{0} is the unit object with respect to the convolution product (see Section 2.5.3 of [ 14 ] ), we obtain the desired isomorphisms by Proposition 2.11 . ∎ 2.5 Middle convolution as coimage The middle convolution functor is defined as the image of the natural morphism from R\mathrm{pr}_{2\,!} to R\mathrm{pr}_{2\,*} . In this section, we give an alternative description of the middle convolution as the coimage of the same morphism. Let \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) be a local system and \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} a nontrivial multiplicative character. Then take a point m_{0}=(\mathbf{x}_{0},y_{0})\in M(\mathcal{A}) and consider the stalks at m_{0} of the above morphism, H^{1}_{c}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\rightarrow H^{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)}), as we saw before Remark 2.5 . Under the Poincaré duality, this morphism is equivalent to the morphism between the homology groups with local coefficients, H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\rightarrow H_{1}^{\mathrm{BM}}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)}), where H_{*}^{\mathrm{BM}} stands for the Borel-Moore homology group. To investigate H_{1}^{\mathrm{BM}}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)}) more closely, we compactify \mathbb{C} to \mathbb{P}^{1}=\mathbb{C}\sqcup\{\infty\} , and set Q_{n+2}^{m_{0},\infty}:=Q_{n+1}^{m_{0}}\sqcup\{\infty\} . Then obviously \mathbb{C}\backslash Q_{n+1}^{m_{0}}=\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty} . For \alpha>0 , let D_{\alpha}(q):=\{z\in\mathbb{C}\mid|z-q|<\alpha\} for q\in\mathbb{C} and D_{\alpha}(\infty):=\{z\in\mathbb{C}\mid|z|>1/\alpha\} . Also denote punctured disks by D^{*}_{\alpha}(q):=D_{\alpha}(q)\backslash\{q\} . Define the union of open disks D_{\alpha}(Q):=\bigcup_{q\in Q}D_{\alpha}(q), for a finite subset Q\subset\mathbb{P}^{1} . Proposition 2.13 . There exists an isomorphism H_{*}^{\mathrm{BM}}(\mathbb{P}^{1}(\mathbb{C})\backslash Q_{n+2}^{m_{0},\infty},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\\ \cong\varprojlim_{\alpha}H_{*}(\mathbb{P}^{1}(\mathbb{C})\backslash Q_{n+2}^{m_{0},\infty},\mathbb{P}^{1}(\mathbb{C})\backslash D_{\alpha}(Q_{n+2}^{m_{0},\infty}),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)}). For sufficiently small \alpha>0 , the right hand side is isomorphic to H_{*}(\mathbb{P}^{1}(\mathbb{C})\backslash Q_{n+2}^{m_{0},\infty},\mathbb{P}^{1}(\mathbb{C})\backslash D_{\alpha}(Q_{n+2}^{m_{0},\infty}),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)}). Proof. We drop the coefficients of the homology groups for simplicity. Let us first notice that for sufficiently small 0<\alpha<\alpha^{\prime}<\!<1 , D_{\alpha}(Q_{n+2}^{m_{0},\infty}) is a deformation retract of D_{\alpha^{\prime}}(Q_{n+2}^{m_{0},\infty}) . Therefore the natural map H_{*}(\mathbb{P}^{1}(\mathbb{C})\backslash Q_{n+2}^{m_{0},\infty},\mathbb{P}^{1}(\mathbb{C})\backslash D_{\alpha^{\prime}}(Q_{n+2}^{m_{0},\infty}))\rightarrow H_{*}(\mathbb{P}^{1}(\mathbb{C})\backslash Q_{n+2}^{m_{0},\infty},\mathbb{P}^{1}(\mathbb{C})\backslash D_{\alpha}(Q_{n+2}^{m_{0},\infty})) is an isomorphism in this case. This shows the second claim, and moreover implies R^{1}\varprojlim_{\alpha}H_{*}(\mathbb{P}^{1}(\mathbb{C})\backslash Q_{n+2}^{m_{0},\infty},\mathbb{P}^{1}(\mathbb{C})\backslash D_{\alpha}(Q_{n+2}^{m_{0},\infty}))=\{0\} . Therefore the first claim follows from Theorem 7.3 in [ 18 ] . ∎ Take \alpha>0 sufficiently small so that D_{\alpha}(Q_{n+2}^{m_{0},\infty}) is the disjoint union of disks D_{\alpha}(q) for q\in Q_{n+2}^{m_{0},\infty} . Then the long exact sequence for relative homology yields the exact sequence \bigoplus_{q\in Q_{n+2}^{m_{0},\infty}}H_{1}(D^{*}_{\alpha}(q))\rightarrow H_{1}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty})\rightarrow H_{1}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty},\mathbb{P}^{1}\backslash D_{\alpha}(Q_{n+2}^{m_{0},\infty})), where the coefficients are omitted for simplicity. Proposition 2.14 . Let \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) be a local system and \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} a nontrivial multiplicative character. Then for m_{0}=(\mathbf{x}_{0},y_{0})\in M(\mathcal{A}) , the stalk of \mathrm{MC}(\mathcal{L})_{m_{0}} is isomorphic to \mathrm{Coker}\left(\bigoplus_{q\in Q_{n+2}^{m_{0},\infty}}H_{1}(D^{*}_{\alpha}(q),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\rightarrow H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\right). Proof. We drop the coefficients of the homology groups as before. Proposition 2.13 and the above long exact sequence show that \displaystyle\mathrm{MC}(\mathcal{L})_{m_{0}} \displaystyle=\mathrm{Im}\left(H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}})\rightarrow H_{1}^{\mathrm{BM}}(\mathbb{C}\backslash Q_{n+1}^{m_{0}})\right) \displaystyle=\mathrm{Im}\left(H_{1}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty})\rightarrow H_{1}^{\mathrm{BM}}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty})\right) \displaystyle\cong\mathrm{Im}\left(H_{1}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty})\rightarrow H_{1}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty},\mathbb{P}^{1}\backslash D_{\alpha}(Q_{n+2}^{m_{0},\infty}))\right) \displaystyle\cong\mathrm{Coker}\left(\bigoplus_{q\in Q_{n+2}^{m_{0},\infty}}H_{1}(D^{*}_{\alpha}(q))\rightarrow H_{1}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty})\right). The last isomorphism is our desired one. ∎ Remark 2.15 . Under our setting, the map \bigoplus_{q\in Q_{n+2}^{m_{0},\infty}}H_{1}(D^{*}_{\alpha}(q),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\rightarrow H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)}) is injective and thus we may regard the source space as a \pi_{1}(M(\mathcal{A}),m_{0}) -submodule of the target space. Indeed, by Proposition 2.13 , we have the isomorphism H_{2}^{\mathrm{BM}}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\cong H_{2}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},\mathbb{C}\backslash D_{\alpha}(Q_{n+1}^{m_{0}}),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)}), and the Poincaré duality and Remark 2.5 tell us that the left hand side vanishes. Therefore the long exact sequence for the relative homology implies the injectivity of the above map. 3 Middle convolution for logarithmic Pfaffian systems with constant coefficients In this section, we recall the definition and fundamental properties of the middle convolution functor for logarithmic Pfaffian systems with constant coefficients on M(\mathcal{A}) , which was introduced by Haraoka [ 12 ] as a generalization of the functor defined by Dettweiler and Reiter in [ 9 ] . Let V=\{(x_{1},\ldots,x_{l})\in\mathbb{C}^{l}\} be a complex affine space of dimension l . For an affine hyperplane H in V , let f_{H}(x) denote a defining affine linear form of H . 3.1 Category of logarithmic Pfaffian systems with constant coefficients Let \mathcal{A} be an affine hyperplane arrangement in \mathbb{C}^{l} . Then for a finite dimensional \mathbb{C} -vector space E , we consider an \mathrm{End}_{\mathbb{C}}(E) -valued logarithmic 1 -form on \mathbb{C}^{l} \Omega_{A}:=\sum_{H\in\mathcal{A}}A_{H}\frac{df_{H}}{f_{H}} with coefficients A_{H}\in\mathrm{End}_{\mathbb{C}}(E) , which also satisfies the integrability condition \Omega_{A}\wedge\Omega_{A}=0. Then \Omega_{A} defines the flat connection \nabla_{A}:=d-\Omega_{A}\colon\mathcal{O}_{\mathbb{C}^{l}}\otimes E\longrightarrow\varOmega_{\mathbb{C}^{l}}^{1}(*\mathcal{A})\otimes E. Here \mathcal{O}_{\mathbb{C}^{l}} is the sheaf of holomorphic functions on \mathbb{C}^{l} , E is regarded as the constant sheaf on \mathbb{C}^{l} , and \varOmega_{\mathbb{C}^{l}}^{1}(*\mathcal{A}) stands for the sheaf of meromorphic 1 -forms on \mathbb{C}^{l} with poles along the Weil divisor \mathcal{A}:=\sum_{H\in\mathcal{A}}H . Then we call such a connection \nabla_{A} a logarithmic Pfaffian system with constant coefficients associated to \mathcal{A} . Let us consider two such connections \nabla_{A_{i}}=d-\sum_{H\in\mathcal{A}}A_{i,H}\frac{df_{H}}{f_{H}} with A_{i,H}\in\mathrm{End}_{\mathbb{C}}(E_{i}) for i=1,2 . Then a morphism from \nabla_{A_{1}} to \nabla_{A_{2}} is defined to be a \mathbb{C} -linear map \phi\colon E_{1}\rightarrow E_{2} such that the following diagram commutes, \hbox to161.83pt{\vbox to57.51pt{\pgfpicture\makeatletter\hbox{\hskip 80.91547pt\lower-27.50508pt\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke} \lxSVG@begingroup@{fill} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces\nullfont\hbox to0.0pt{\lxSVG@begingroup@{_scopebegin} {}{}{}{{}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{{}}{{}}{{}}{{}}}{{{\lx@inpgf@ignorespaces}}}{{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-80.91547pt}{-20.79979pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}}{{{\lx@inpgf@ignorespaces{}}}{{}}{{}}{{}}{{}}}} \lxSVG@closescope }}} {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -31.51 25.19 L 0.59 25.19}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{0.62779pt}{18.20532pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-19.10423pt}{23.06363pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -71.87 12.18 L -71.87 -13.43}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{-51.94383pt}{-9.90677pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-49.59106pt}{-2.40335pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M 56.69 12.18 L 56.69 -11.62}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{0.0}{-1.0}{1.0}{0.0}{40.97162pt}{-8.60005pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{43.32439pt}{-1.75pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{\lx@inpgf@ignorespaces}}{}{}{}{{{}{}}}\lxSVG@begingroup@{_scopebegin} \lxSVG@setlinewidth{\the\pgflinewidth}\lxSVG@begingroup@{stroke-width} \lx@inpgf@ignorespaces{}{}{}{}{{}}{}{}{{}}\lxSVG@stroke\lxSVG@drawpath@unclipped{M -31.51 -25.32 L 0.59 -25.32}{fill:none} {{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}}}{{}{{}}{}{}{{}}{{{\lx@inpgf@ignorespaces}}{{{\lx@inpgf@ignorespaces}}{\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{0.62779pt}{-18.29979pt}\lxSVG@begingroup@{transform} \lxSVG@begingroup@{_scopebegin} \lxSVG@begingroup@{stroke-dasharray,stroke-dashoffset} \lxSVG@begingroup@{stroke-linecap} \lxSVG@begingroup@{stroke-linejoin} \lxSVG@drawpath@unclipped{M -2.88 3.32 C -2.35 1.33 -1.18 0.39 0 0 C -1.18 -0.39 -2.35 -1.33 -2.88 -3.32}{fill:none} \lxSVG@closescope \lxSVG@closescope }}{{\lx@inpgf@ignorespaces}}}}\lx@inpgf@ignorespaces\hbox{\hbox{{\lxSVG@begingroup@{_scopebegin} {{}{}{{ {}{}}}{ {}{}} {{}{{\lx@inpgf@ignorespaces}}}{{}{\lx@inpgf@ignorespaces}}{}{{}{\lx@inpgf@ignorespaces}} {\lx@inpgf@ignorespaces }{{{{\lx@inpgf@ignorespaces}}\lxSVG@begingroup@{_scopebegin} \lxSVG@transformcm{1.0}{0.0}{0.0}{1.0}{-19.10423pt}{-13.44148pt}\lxSVG@begingroup@{transform} \pgfsys@hbox{60}\lxSVG@closescope }}} \lxSVG@closescope }}} \lxSVG@closescope \lxSVG@closescope {\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}{\lx@inpgf@ignorespaces}\hss}\lxSVG@discardpath\lxSVG@closescope \hss}}\lxSVG@closescope\endpgfpicture}}. Then we denote the category of logarithmic Pfaffian systems with constant coefficients associated to \mathcal{A} by \mathrm{Pf}(\mathrm{log}(\mathcal{A})). 3.2 Convolution functor along a line Let us focus on the x_{l} -axis of \mathbb{C}^{l} and denote it by Y as before. Here Y is not necessarily to be good with respect to \mathcal{A} . Let us consider the n=|\mathcal{A}\backslash\mathcal{A}_{Y}| -dimensional vector space, \mathbb{C}^{n}=\bigoplus_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}\mathbb{C}e_{H}. Also consider the following additional set of hyperplanes in \mathbb{C}^{l} , \mathcal{A}^{+Y}:=\{X+Y\mid X\in L_{2}(\mathcal{A}\backslash\mathcal{A}_{Y})\}. Here we note that if Y is good, then \mathcal{A}^{+Y}\subset\mathcal{A} . For an object \nabla_{A}\in\mathrm{Pf}(\mathrm{log}(\mathcal{A})) with the coefficient 1 -form \Omega_{A}=\sum_{H\in\mathcal{A}}A_{H}\frac{df_{H}}{f_{H}}\quad(A_{H}\in\mathrm{End}_{\mathbb{C}}(E)), and a parameter \lambda\in\mathbb{C}\backslash\mathbb{Z} , Haraoka defined in [ 12 ] another logarithmic Pfaffian system \nabla_{c_{\lambda}(A)}\in\mathrm{Pf}(\mathrm{log}(\mathcal{A}\cup\mathcal{A}^{+Y})) with the coefficient 1 -form \Omega_{c_{\lambda}(A)}=\sum_{H\in\mathcal{A}}c_{\lambda}(A)_{H}\frac{df_{H}}{f_{H}}+\sum_{H^{\prime}\in\mathcal{A}^{+Y}}c_{\lambda}(A)_{H^{\prime}}\frac{df_{H^{\prime}}}{f_{H^{\prime}}} for c_{\lambda}(A)_{H},c_{\lambda}(A)_{H^{\prime}}\in\mathrm{End}_{\mathbb{C}}(E\otimes_{\mathbb{C}}\mathbb{C}^{n}) as follows. Let E_{H,H^{\prime}}\in\mathrm{End}_{\mathbb{C}}(\mathbb{C}^{n}) be the matrix units for H,H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y} , i.e., the endomorphism defined by E_{H,H^{\prime}}e_{H^{\prime\prime}}=\delta_{H^{\prime},H^{\prime\prime}}e_{H} for H,H^{\prime},H^{\prime\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y} . Here \delta_{H^{\prime},H^{\prime\prime}} is the Kronecker delta. Then for H\in\mathcal{A} , we set c_{\lambda}(A)_{H}:=\begin{cases}\displaystyle\sum_{H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y}}(A_{H^{\prime}}+\lambda\delta_{H,H^{\prime}}\mathrm{Id}_{E})\otimes E_{H,H^{\prime}}&\text{if }H\in\mathcal{A}\backslash\mathcal{A}_{Y},\\ A_{H}\otimes\mathrm{Id}_{\mathbb{C}^{n}}&\text{if }H\in\mathcal{A}_{Y}.\end{cases} Also for H^{\prime}\in\mathcal{A}^{+Y} , we set c_{\lambda}(A)_{H^{\prime}}:=\displaystyle\sum_{\begin{subarray}{c}H_{1},H_{2}\in\mathcal{A}\backslash\mathcal{A}_{Y},\\ H^{\prime}=(H_{1}\cap H_{2})+Y\end{subarray}}A_{H_{1}}\otimes(E_{H_{2},H_{2}}-E_{H_{2},H_{1}})+A_{H_{2}}\otimes(E_{H_{1},H_{1}}-E_{H_{1},H_{2}}). The integrability condition for \Omega_{c_{\lambda}(A)} can be verified from that for \Omega_{A} (cf. Remark 2.2 in [ 12 ] ). By the definition, the correspondence \mathrm{Pf}(\mathrm{log}(\mathcal{A}))\ni d-\Omega_{A}\longmapsto d-\Omega_{c_{\lambda}(A)}\in\mathrm{Pf}(\mathrm{log}(\mathcal{A}\cup\mathcal{A}^{+Y})) is functorial, therefore it defines a functor c_{\lambda}\colon\mathrm{Pf}(\mathrm{log}(\mathcal{A}))\longrightarrow\mathrm{Pf}(\mathrm{log}(\mathcal{A}\cup\mathcal{A}^{+Y})), called the convolution functor along the line Y with parameter \lambda . 3.3 Middle convolution functor along a line Now we define an analogue of middle convolution functor as a functor between the categories of logarithmic Pfaffian systems. For a logarithmic Pfaffian system d-\Omega_{A}\in\mathrm{Pf}(\mathrm{log}(\mathcal{A})) with A_{H}\in\mathrm{End}_{\mathbb{C}}(E)\ (H\in\mathcal{A}) , let us define the following subspaces of E\otimes_{\mathbb{C}}\mathbb{C}^{n} , K:=\bigoplus_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}\left(\mathrm{Ker\,}A_{H}\otimes\mathbb{C}e_{H}\right),\quad\quad L:=\mathrm{Ker}\left(\sum_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}A_{H}+\lambda\mathrm{Id}_{E}\right)\otimes\mathbb{C}\sum_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}e_{H}. Then it was shown in [ 12 ] and also in [ 9 ] that K\cap L=\{0\} , and the subbundles \mathcal{O}_{\mathbb{C}^{l}}\otimes K and \mathcal{O}_{\mathbb{C}^{l}}\otimes L of \mathcal{O}_{\mathbb{C}^{l}}\otimes(E\otimes_{\mathbb{C}}\mathbb{C}^{n}) are invariant under the connection \nabla_{C_{\lambda}(A)} (see Proposition 2.2 in [ 12 ] ). Namely, we obtain the sub-connections \displaystyle\nabla_{c_{\lambda}(A)}^{K}\colon\mathcal{O}_{\mathbb{C}^{l}}\otimes K\longrightarrow\varOmega_{\mathbb{C}^{l}}^{1}(*(\mathcal{A}\cup\mathcal{A}^{+Y}))\otimes K, \displaystyle\nabla_{c_{\lambda}(A)}^{L}\colon\mathcal{O}_{\mathbb{C}^{l}}\otimes L\longrightarrow\varOmega_{\mathbb{C}^{l}}^{1}(*(\mathcal{A}\cup\mathcal{A}^{+Y}))\otimes L. Therefore, we moreover obtain the quotient-connection \nabla_{\mathrm{mc}_{\lambda}(A)}\colon\mathcal{O}_{\mathbb{C}^{l}}\otimes(E\otimes_{\mathbb{C}}\mathbb{C}^{n}/(K\oplus L))\longrightarrow\varOmega_{\mathbb{C}^{l}}^{1}(*(\mathcal{A}\cup\mathcal{A}^{+Y}))\otimes(E\otimes_{\mathbb{C}}\mathbb{C}^{n}/(K\oplus L)). Then we call the resulting functor \mathrm{mc}_{\lambda}\colon\mathrm{Pf}(\mathrm{log}(\mathcal{A}))\longrightarrow\mathrm{Pf}(\mathrm{log}(\mathcal{A}\cup\mathcal{A}^{+Y}));\quad\nabla_{A}\mapsto\nabla_{\mathrm{mc}_{\lambda}(A)}, the middle convolution functor along the line Y with parameter \lambda . As well as for local systems, this middle convolution functor also satisfies the composition law under the following assumptions: \bigcap_{\begin{subarray}{c}H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y},\\ H^{\prime}\neq H\end{subarray}}\mathrm{Ker\,}A_{H^{\prime}}\cap\mathrm{Ker\,}(A_{H}+\tau\mathrm{Id}_{E})=\{0\}\quad\text{ for any }H\in\mathcal{A}\backslash\mathcal{A}_{Y}\text{ and }\tau\in\mathbb{C}, (2) \sum_{\begin{subarray}{c}H^{\prime}\in\mathcal{A}\backslash\mathcal{A}_{Y},\\ H^{\prime}\neq H\end{subarray}}\mathrm{Im\,}A_{H}+\mathrm{Im\,}(A_{H}+\tau\mathrm{Id}_{E})=E\quad\text{ for any }H\in\mathcal{A}\backslash\mathcal{A}_{Y}\text{ and }\tau\in\mathbb{C}. (3) Theorem 3.1 (Theorem 3.1 in [ 12 ] , Appendix in [ 8 ] ) . Suppose \nabla_{A}\in\mathrm{Pf}(\mathrm{log}(\mathcal{A})) satisfies the assumptions ( 2 ) and ( 3 ). Then the following holds for \lambda,\mu\in\mathbb{C}\backslash\mathbb{Z} , \displaystyle\mathrm{mc}_{\mu}\circ\mathrm{mc}_{\lambda}(\nabla_{A}) \displaystyle\cong\mathrm{mc}_{\lambda+\mu}(\nabla_{A}), \displaystyle\mathrm{mc}_{-\lambda}\circ\mathrm{mc}_{\lambda}(\nabla_{A}) \displaystyle\cong\nabla_{A}. 4 Compatibility for de Rham functor and middle convolution functor along a good line In this section, we retain the notation in the previous sections and moreover assume that Y is a good line for \mathcal{A} . In this case, the goodness implies that \mathcal{A}^{+Y}\subset\mathcal{A} . Therefore convolution and middle convolution functors introduced in the previous section give endofunctors on \mathrm{Pf}(\mathrm{log}(\mathcal{A})) , c_{\lambda},\mathrm{mc}_{\lambda}\colon\mathrm{Pf}(\mathrm{log}(\mathcal{A}))\longrightarrow\mathrm{Pf}(\mathrm{log}(\mathcal{A})). 4.1 De Rham functor for logarithmic Pfaffian systems with constant coefficients Let us denote the category of flat connections on holomorphic vector bundles on M(\mathcal{A}) by \mathrm{Conn}(M(\mathcal{A})) . Namely, it consists of pairs (\mathcal{E},\nabla) of holomorphic vector bundles \mathcal{E} on M(\mathcal{A}) and flat connections \nabla\colon\mathcal{E}\rightarrow\mathcal{E}\otimes\varOmega_{M(\mathcal{A})}^{1} , with usual bundle morphisms preserving the connections as morphisms. Then the de Rham functor is defined by \mathrm{DR}\colon\mathrm{Conn}(M(\mathcal{A}))\longrightarrow\mathrm{Loc}(M(\mathcal{A}),\mathbb{C});\quad(\mathcal{E},\nabla)\longmapsto\mathrm{Ker}(\nabla), which gives an equivalence of categories (see Theorem 2.17 in [ 7 ] ). The open embedding j\colon M(\mathcal{A})\hookrightarrow\mathbb{C}^{l} defines the pullback functor j^{*}\colon\mathrm{Pf}(\mathrm{log}(\mathcal{A}))\longrightarrow\mathrm{Conn}(M(\mathcal{A}));\quad\nabla_{A}\longmapsto j^{*}\nabla_{A}. Then by composing these functors, we define the de Rham functor for logarithmic Pfaffian systems with constant coefficients by \mathrm{DR}_{\mathrm{Pf}}\colon\mathrm{Pf}(\mathrm{log}(\mathcal{A}))\longrightarrow\mathrm{Loc}(M(\mathcal{A}),\mathbb{C});\quad\nabla_{A}\longmapsto\mathrm{DR}(j^{*}\nabla_{A}). 4.2 Multivalued section for flat connection Let us consider a flat connection \nabla\in\mathrm{Conn}(X) defined on a complex manifold X , and also consider the associated local system of horizontal sections \mathcal{L}:=\mathrm{DR}(\nabla) . We recall the description of multivalued horizontal sections for the flat connection following the section 6 in [ 7 ] . Let us take a point x_{0}\in X and let \pi\colon\widetilde{X}_{x_{0}}\rightarrow X be the universal covering of X . Here \widetilde{X}_{x_{0}} is the set of homotopy classes of paths starting from x_{0} in X . Also denote the trivial path at x_{0} by \widetilde{x}_{0}\in\widetilde{X}_{x_{0}} . Definition 4.1 . A multivalued horizontal section of \nabla is a global section of the sheaf \pi^{*}\mathcal{L} . Let us recall that there exists a natural isomorphism between the stalks (\pi^{*}\mathcal{L})_{\widetilde{x}} and (\mathcal{L})_{x} for x\in X and \widetilde{x}\in\pi^{-1}(x) . Definition 4.2 . The germ \tilde{s}_{\widetilde{x}} of a multivalued horizontal section \tilde{s}\in\pi^{*}\mathcal{L}(\widetilde{X}_{x_{0}}) at \widetilde{x} defines the germ s_{x}\in(\mathcal{L})_{x} through the above isomorphism. We call s_{x} the determination of \tilde{s} at x . Conversely, we call \tilde{s} the branch of multivalued horizontal section of s_{x} at \widetilde{x} . In particular, the germ s_{x_{0}} determined by the isomorphism (\pi^{*}\mathcal{L})_{\widetilde{x}_{0}}\cong(\mathcal{L})_{x_{0}} is called the determination of base for \tilde{s} , and \tilde{s} is called the standard branch of multivalued horizontal section of base for s_{x_{0}} . 4.3 Period map For a nontrivial character \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} , we take a complex number \lambda\in\mathbb{C}\backslash\mathbb{Z} such that \exp(2\pi i\lambda)=\chi(1) . Let \nabla_{A}=d-\sum_{H\in\mathcal{A}}A_{H}\frac{df_{H}}{f_{H}}\quad(A_{H}\in\mathrm{End}_{\mathbb{C}}(E)) be a logarithmic Pfaffian system with constant coefficients, and \mathcal{L}_{A}:=\mathrm{DR}_{\mathrm{Pf}}(\nabla_{A}) the associated local system. We moreover assume that \nabla_{A} satisfies the following assumption. Assumption 4.3 . For each H\in\mathcal{A}\backslash\mathcal{A}_{Y} , A_{H} has no nonzero integer as an eigenvalue, and the sum \sum_{H\in\mathcal{A}\setminus\mathcal{A}_{Y}}A_{H}+\lambda also has no nonzero integer eigenvalue. The purpose of this section is to show the following theorem which compares the functors C_{\chi\,!} for local systems and c_{\lambda} for logarithmic Pfaffian systems. Theorem 4.4 . Under Assumption 4.3 , there exists an isomorphism of local systems on M(\mathcal{A}) , C_{\chi\,!}\circ\mathrm{DR}_{\mathrm{Pf}}(\nabla_{A})\cong\mathrm{DR}_{\mathrm{Pf}}\circ c_{\lambda}(\nabla_{A}). Let us give some preparations for the proof of this theorem. Let us fix a point m_{0}=(\mathbf{x}_{0},y_{0})\in M(\mathcal{A}) and d_{0}\in\mathbb{C}\backslash Q_{n+1}^{m_{0}}=\mathrm{pr}_{1}^{-1}(m_{0}) . Here \mathrm{pr}_{1}\colon M(\mathcal{A})\times_{\pi_{Y}}M(\mathcal{A})\backslash\Delta\rightarrow M(\mathcal{A}) is the projection onto the first factor. Let us recall that we have isomorphisms \displaystyle C_{\chi\,!}(\mathcal{L}_{A})_{m_{0}}=(R^{1}\mathrm{pr}_{2\,!}((\mathcal{L}_{A})_{\mathbf{x},y}\boxtimes\chi_{(z-y)}))_{m_{0}} \displaystyle\cong H^{1}_{c}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}) \displaystyle\cong H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}), as we saw in Section 2.5 . The isomorphism in Theorem 4.4 will be given by the non-degenerate pairing between the homology and cohomology groups H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})\times H^{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})^{\vee})\rightarrow\mathbb{C}. Here \mathcal{L}^{\vee} denotes the dual local system of a local system \mathcal{L} . Let us look at the cohomology group H^{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})^{\vee}) more closely. Recall that the local system ((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}) on \mathbb{C}\backslash Q_{n+1}^{m_{0}} is defined by the following connection. For H\in\mathcal{A}\backslash\mathcal{A}_{Y} , let q_{H}\in\mathbb{C} be the point defined by the equation f_{H}(\mathbf{x}_{0},y)=0 , and then we can write Q_{n+1}^{m_{0}}=\{q_{H}\mid H\in\mathcal{A}\backslash\mathcal{A}_{Y}\}\sqcup\{y_{0}\}. Let us consider the flat connections on \mathbb{C}\backslash Q_{n+1}^{m_{0}} , \displaystyle\nabla_{A}|_{\mathrm{pr}_{1}^{-1}(m_{0})}:=d-\sum_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}\frac{A_{H}}{y-q_{H}}dy, \displaystyle\nabla_{\lambda}:=d-\frac{\lambda}{y-y_{0}}dy. Then the local system ((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}) on \mathbb{C}\backslash Q_{n+1}^{m_{0}} is defined as the horizontal sections of the tensor product connection \nabla_{A,m_{0}}^{\lambda}:=\nabla_{A}|_{\mathrm{pr}_{1}^{-1}(m_{0})}\otimes\nabla_{\lambda}. Therefore the dual local system ((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})^{\vee} corresponds to the dual connection \nabla_{A,m_{0}}^{\lambda\,\vee} . Since \nabla_{A,m_{0}}^{\lambda\,\vee} has logarithmic singularities along Q_{n+2}^{m_{0},\infty} , it also define the following meromorphic connection on \mathbb{P}^{1} , \overline{\nabla}_{A,m_{0}}^{\lambda\,\vee}\colon\mathcal{O}_{\mathbb{P}^{1}}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee}\longrightarrow\varOmega_{\mathbb{P}^{1}}^{1}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee}, where \mathcal{O}_{\mathbb{P}^{1}}(*Q_{n+2}^{m_{0},\infty}) and \varOmega_{\mathbb{P}^{1}}^{1}(*Q_{n+2}^{m_{0},\infty}) are the sheaves of meromorphic functions and 1 -forms on \mathbb{P}^{1} with poles along Q_{n+2}^{m_{0},\infty} respectively. Then since the Riemann-Roch theorem implies that H^{q}(\mathbb{P}^{1},\mathcal{O}_{\mathbb{P}^{1}}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee})=H^{q}(\mathbb{P}^{1},\varOmega_{\mathbb{P}^{1}}^{1}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee})=0 for q>0 , we have an isomorphism H^{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})^{\vee})\cong H^{1}\Gamma(\mathbb{P}^{1},\varOmega_{\mathbb{P}^{1}}^{*}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee},\overline{\nabla}_{A,m_{0}}^{\lambda\,\vee}) (4) by Proposition I.2.19 in [ 7 ] , see also (2.10.1) in [ 6 ] . Here H^{1}\Gamma(\mathbb{P}^{1},\varOmega_{\mathbb{P}^{1}}^{*}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee},\overline{\nabla}_{A,m_{0}}^{\lambda\,\vee}):=\frac{\Gamma(\mathbb{P}^{1},\varOmega_{\mathbb{P}^{1}}^{1}(*Q_{n+2}^{m_{0},\infty}))\otimes E^{\vee})}{\overline{\nabla}_{A,m_{0}}^{\lambda\,\vee}(\Gamma(\mathbb{P}^{1},\mathcal{O}_{\mathbb{P}^{1}}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee}))}. Then we have the following lemma. Lemma 4.5 . Under Assumption 4.3 , we have \mathrm{dim}_{\mathbb{C}}H^{1}\Gamma(\mathbb{P}^{1},\varOmega_{\mathbb{P}^{1}}^{*}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee},\overline{\nabla}_{A,m_{0}}^{\lambda\,\vee})=n\cdot\mathrm{dim}_{\mathbb{C}}E. and this space is generated by the classes of the following 1 -forms, \frac{dy}{y-q_{H}}\otimes e^{\vee}\quad\text{ for }H\in\mathcal{A}\backslash\mathcal{A}_{Y}\text{ and }e^{\vee}\in E^{\vee}. Proof. Under the poincare duality, the dimension of the cohomology group H^{1}\Gamma(\mathbb{P}^{1},\varOmega_{\mathbb{P}^{1}}^{*}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee},\overline{\nabla}_{A,m_{0}}^{\lambda\,\vee})\cong H^{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})^{\vee}) is same as the compactly supported cohomology group H_{c}^{1}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})). Then the dimension of this comactly supported cohomology group is computed as follows: \displaystyle\mathrm{dim}_{\mathbb{C}}H_{c}^{1}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})) \displaystyle=-\sum_{i}(-1)^{i}\mathrm{dim}_{\mathbb{C}}H^{i}_{c}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})) \displaystyle=-\chi_{c}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}))=-\chi_{c}(\mathbb{P}^{1}\backslash Q_{n+2}^{m_{0},\infty})\cdot\mathrm{dim}_{\mathbb{C}}E \displaystyle=(\chi_{c}(Q_{n+2}^{m_{0},\infty})-\chi_{c}(\mathbb{P}^{1}))\cdot\mathrm{dim}_{\mathbb{C}}E=(n+2-2)\cdot\mathrm{dim}_{\mathbb{C}}E=n\cdot\mathrm{dim}_{\mathbb{C}}E, where the first equality follows from Remark 2.5 . Let us show the second statement. First, we see that the quotient space is generated by the classes of logarithmic 1 -forms of the forms \frac{dy}{y-q}\otimes e^{\vee} and \frac{d\eta}{\eta}\otimes e^{\vee} for e^{\vee}\in E^{\vee} and q\in Q_{n+2}^{m_{0},\infty}\backslash\{\infty\} , where we set \eta=1/y . Indeed, for q\in Q_{n+2}^{m_{0},\infty}\backslash\{\infty\} and k\in\mathbb{Z}_{>0} , we have \overline{\nabla}_{A,m_{0}}^{\lambda\,\vee}\left(\frac{e^{\vee}}{(y-q)^{k}}\right)=\left(-\frac{k}{(y-q)^{k+1}}+\frac{1}{(y-q)^{k}}\left(\sum_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}\frac{A_{H}^{\vee}}{y-q_{H}}+\frac{\lambda\mathrm{Id}_{E^{\vee}}}{y-y_{0}}\right)\right)dy\otimes e^{\vee}, where A_{H}^{\vee}\in\mathrm{End}_{\mathbb{C}}(E^{\vee}) is the dual endomorphism of A_{H}\in\mathrm{End}_{\mathbb{C}}(E) . Then since A_{H}^{\vee}-k and \lambda-k are invertible for any H\in\mathcal{A}\backslash\mathcal{A}_{Y} by Assumption 4.3 , the above equation implies that the class of \frac{dy}{(y-q)^{k+1}}\otimes e^{\vee} can be written as a linear combination of the classes of \frac{dy}{y-q^{\prime}}\otimes(e^{\vee})^{\prime} for q^{\prime}\in Q_{n+2}^{m_{0},\infty}\backslash\{\infty\} and (e^{\vee})^{\prime}\in E^{\vee} . Also, by a similar argument, the class of \frac{d\eta}{\eta^{k+1}}\otimes e^{\vee} can be written as a linear combination of the classes of \frac{dy}{y-q^{\prime}}\otimes(e^{\vee})^{\prime} for q^{\prime}\in Q_{n+2}^{m_{0},\infty}\backslash\{\infty\} and (e^{\vee})^{\prime}\in E^{\vee} . Thus the claim is shown. Next, by the residue theorem on \mathbb{P}^{1} , the class of \frac{d\eta}{\eta}\otimes e^{\vee} can be written as a linear combination of the classes of \frac{dy}{y-q}\otimes e^{\vee} for q\in Q_{n+1}^{m_{0}} . Furthermore, since \lambda\in\mathbb{C}\backslash\mathbb{Z} and the connection \overline{\nabla}_{A,m_{0}}^{\lambda\,\vee} is written as d+\left(\frac{\lambda\cdot\mathrm{Id}_{E^{\vee}}}{y-y_{0}}+(\text{holomorphic at }y_{0})\right)dy near y=y_{0} , the class of \frac{dy}{y-y_{0}}\otimes e^{\vee} can be also written as a linear combination of the classes of \frac{dy}{y-q}\otimes e^{\vee} for q\in Q_{n+2}^{m_{0},\infty}\backslash\{\infty,y_{0}\}=\{q_{H}\mid H\in\mathcal{A}\backslash\mathcal{A}_{Y}\} . ∎ Recall that elements in H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}) are represented by linear combinations of closed chains of the form \sigma\otimes\phi(\mathbf{x},y)(z-y)^{\lambda}, where \sigma is a 1 -chain in the universal covering space \pi_{d_{0}}\colon\widetilde{(\mathbb{C}\backslash Q_{n+1}^{m_{0}})}_{d_{0}}\rightarrow\mathbb{C}\backslash Q_{n+1}^{m_{0}} , and also, \phi(\mathbf{x},y) and (z-y)^{\lambda} are standard branches of multivalued horizontal sections for germs of local systems ((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y})_{d_{0}} and ((z-y)^{*}\mathcal{K}_{\chi})_{d_{0}} respectively. Therefore we can regard the above \phi(\mathbf{x},y)(z-y)^{\lambda} as the function which is multivalued with respect to the variable y and holomorphic with respect to (\mathbf{x},z)\in M(\mathcal{A}) near m_{0} . Now we take a base e_{1},\ldots,e_{r} of E and identify E=\mathbb{C}^{r} with r=\dim_{\mathbb{C}}E . Let e_{1}^{\vee},\ldots,e_{r}^{\vee} be the associated dual base of E^{\vee} . Then we define the map \mathrm{Per}_{Y,m_{0}} which we call the period map along Y at m_{0} , by \displaystyle\mathrm{Per}_{Y,m_{0}}\colon C_{\chi\,!}(\mathcal{L}_{A})_{m_{0}}\longrightarrow(\mathcal{O}_{M(\mathcal{A})}\otimes(E\otimes_{\mathbb{C}}\mathbb{C}^{n}))_{m_{0}}; \displaystyle[\sigma\otimes\phi(\mathbf{x},y)(z-y)^{\lambda}]\longmapsto\left(\int_{\sigma}\langle\phi(\mathbf{x},y)(z-y)^{\lambda},e_{i}^{\vee}\rangle\,\pi^{*}_{d_{0}}\left(\left(\frac{\partial}{\partial y}\log f_{H}(\mathbf{x},y)\right)\,dy\right)\right)_{\begin{subarray}{c}i=1,\ldots,r,\\ H\in\mathcal{A}\backslash\mathcal{A}_{Y}\end{subarray}}. We are now ready for the proof of Theorem 4.4 . Proof of Theorem 4.4 . Let us fix a point m_{0}=(\mathbf{x}_{0},y_{0})\in M(\mathcal{A}) . Then let us show that the period map \mathrm{Per}_{Y,m_{0}} gives an isomorphism between the stalks (R^{1}\mathrm{pr}_{2\,!}((\mathcal{L}_{A})_{\mathbf{x},y}\boxtimes\chi_{(z-y)}))_{m_{0}}\longrightarrow(\mathrm{DR}_{\mathrm{Pf}}(c_{\lambda}(\nabla_{A})))_{m_{0}}, and preserves the monodromy actions. First we note that Proposition 2.1 in [ 12 ] tells us that \mathrm{Im\,}\mathrm{Per}_{Y,m_{0}}\subset(\mathrm{DR}_{\mathrm{Pf}}(c_{\lambda}(\nabla_{A})))_{m_{0}} . Then since \mathrm{Per}_{Y,m_{0}} preserves the monodromy actions by the definition, we only need to check that \mathrm{Per}_{Y,m_{0}} is bijective onto (\mathrm{DR}_{\mathrm{Pf}}(c_{\lambda}(\nabla_{A})))_{m_{0}}. As we saw in Lemma 4.5 , the dimension of the source space is equal to n\cdot\mathrm{dim}_{\mathbb{C}}E , which is also equal to the rank of the connection c_{\lambda}(A)_{H} . Therefore it suffices to show that \mathrm{Per}_{Y,m_{0}} is injective. Under the isomorphism ( 4 ), the non-degenerate pairing H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}))\otimes H^{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})^{\vee})\longrightarrow\mathbb{C} becomes the natural pairing between the homology and cohomology groups, [\sigma\otimes\phi(\mathbf{x},y)(z-y)^{\lambda}]\otimes[\frac{dy}{y-q}\otimes e^{\vee}]\longmapsto\int_{\sigma}\langle\phi(\mathbf{x},y),e^{\vee}\rangle(z-y)^{\lambda}\frac{dy}{y-q}. for \displaystyle[\sigma\otimes\phi(\mathbf{x},y)(z-y)^{\lambda}] \displaystyle\in H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})), \displaystyle[\frac{dy}{y-q}\otimes e^{\vee}] \displaystyle\in H^{1}\Gamma(\mathbb{P}^{1},\varOmega_{\mathbb{P}^{1}}^{*}(*Q_{n+2}^{m_{0},\infty})\otimes E^{\vee},\overline{\nabla}_{A}^{\vee}|_{Y}), see Remark 2.16 in [ 6 ] for instance. Then since Lemma 4.5 tells us that the classes [\frac{dy}{y-q_{H}}\otimes e^{\vee}] for H\in\mathcal{A}\backslash\mathcal{A}_{Y} and e^{\vee} running through a basis of E^{\vee} form a basis of the cohomology group, the injectivity of \mathrm{Per}_{Y,m_{0}} follows from the non-degeneracy of this pairing. ∎ 4.4 Meromorphic solutions for a linear ordinary differential equation with a simple pole Let us consider the following linear ordinary differential equation \frac{d}{dy}F(y)=\left(\frac{A_{-1}}{y}+A_{0}+A_{1}y+\cdots\right)F(y), (5) defined on the punctured disk D^{*}_{\alpha}(0)=\{y\in\mathbb{C}\mid 0<|y|<\alpha\} , where A_{i}\in M_{k}(\mathbb{C}) for i=-1,0,1,\ldots . Let {\mathcal{S}ol}_{D^{*}_{\alpha}(0)} denote the sheaf of holomorphic solutions of ( 5 ) on D^{*}_{\alpha}(0) . Then the stalk ({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}} at a base point y_{0}\in D^{*}_{\alpha}(0) becomes a k -dimensional module over the group ring \mathbb{C}[\pi_{1}(D^{*}_{\alpha}(0),y_{0})] . In this section, we recall some fundamental properties of the space of meromorphic solutions for ( 5 ), i.e., ({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}}^{\pi_{1}(D^{*}_{\alpha}(0),y_{0})}:=\{v\in({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}}\mid\gamma\cdot v=v\text{ for }\gamma\in\pi_{1}(D^{*}_{\alpha}(0),y_{0})\}, which will be used in the next section. Let us first assume that the residue matrix A_{-1} is a nilpotent matrix. Lemma 4.6 . Assume that A_{-1} is a nilpotent matrix. Then there exists a collection of germs of holomorphic solutions F_{1}(y),\ldots,F_{k}(y)\in({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}} which form a basis of ({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}} , such that c_{1}F_{1}(y)+\cdots+c_{k}F_{k}(y)\in({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}}^{\pi_{1}(D^{*}_{\alpha}(0),y_{0})} if and only if (c_{1},\ldots,c_{k})\in\mathrm{Ker\,}A_{-1}. Moreover for (c_{1},\ldots,c_{k})\in\mathrm{Ker\,}A_{-1} , c_{1}F_{1}(y)+\cdots+c_{k}F_{k}(y) is holomorphic at y=0 and satisfies \left.\left(c_{1}F_{1}(y)+\cdots+c_{k}F_{k}(y)\right)\right|_{y=0}=(c_{1},\ldots,c_{k})^{T}. Proof. Since A_{-1} is a nilpotent matrix which has 0 as the only eigenvalue, we can take a matrix of fundamental solutions for ( 5 ) of the form X(y)y^{A_{-1}} with X(y)\in\mathrm{GL}_{k}(\mathcal{O}_{D_{\alpha}(0)}(D_{\alpha}(0))) such that X(0)=I_{k} , see Theorem 1 in [ 3 ] for instance. Then for a simple loop \gamma around the origin in D^{*}_{\alpha}(0) with the base point y_{0} , the monodromy action on the germ (X(y)y^{A_{-1}})_{y_{0}} is given by \gamma\cdot(X(y)y^{A_{-1}})_{y_{0}}=(X(y)y^{A_{-1}})_{y_{0}}\exp(2\pi iA_{-1}). (6) Thus for v\in\mathbb{C}^{k} , we have (X(y)y^{A_{-1}})_{y_{0}}\cdot v\in({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}}^{\pi_{1}(D^{*}_{\alpha}(0),y_{0})} if and only if \exp(2\pi iA_{-1})v=v. Since \mathrm{Ker}(\exp(2\pi iA_{-1})-\mathrm{Id}_{k})=\mathrm{Ker\,}A_{-1}, we have (X(y)y^{A_{-1}})_{y_{0}}\cdot v\in({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}}^{\pi_{1}(D^{*}_{\alpha}(0),y_{0})}\iff v\in\mathrm{Ker\,}A_{-1}, as desired. Moreover note that y^{A_{-1}}v=v for v\in\mathrm{Ker\,}A_{-1} . Indeed, we have y^{A_{-1}}v=\sum_{j=0}^{\infty}\frac{(\log y)^{j}}{j!}A_{-1}^{j}v=v, since A_{-1}^{j}v=0 for all j\geq 1 . Thus for v\in\mathrm{Ker\,}A_{-1} , X(y)y^{A_{-1}}v=X(y)v is holomorphic at y=0 and satisfies \left.X(y)y^{A_{-1}}v\right|_{y=0}=X(0)v=I_{k}v=v. ∎ Next we assume that the residue matrix A_{-1} has no nonzero integer as an eigenvalue. Then we can show the same as above. Proposition 4.7 . Assume that A_{-1} has no nonzero integer as an eigenvalue. Then there exists a collection of germs of holomorphic solutions F_{1}(y),\ldots,F_{k}(y)\in({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}} which form a basis of ({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}} , such that c_{1}F_{1}(y)+\cdots+c_{k}F_{k}(y)\in({\mathcal{S}ol}_{D^{*}_{\alpha}(0)})_{y_{0}}^{\pi_{1}(D^{*}_{\alpha}(0),y_{0})} if and only if (c_{1},\ldots,c_{k})\in\mathrm{Ker\,}A_{-1}. Moreover for (c_{1},\ldots,c_{k})\in\mathrm{Ker\,}A_{-1} , c_{1}F_{1}(y)+\cdots+c_{k}F_{k}(y) is holomorphic at y=0 and satisfies \left.\left(c_{1}F_{1}(y)+\cdots+c_{k}F_{k}(y)\right)\right|_{y=0}=(c_{1},\ldots,c_{k})^{T}. Proof. Under a linear transformation, we may assume that A_{-1} is a block diagonal matrix of the form A_{-1}=\begin{pmatrix}A_{-1}^{\prime}&0\\ 0&A_{-1}^{\prime\prime}\end{pmatrix}, where A_{-1}^{\prime} is a nilpotent matrix and A_{-1}^{\prime\prime} is an invertible matrix. Namely \mathbb{C}^{k} is written as the sum \mathbb{C}^{k}=V^{\prime}\oplus V^{\prime\prime} of A_{-1} -invariant subspaces such that A_{-1}^{\prime}=A_{-1}|_{V^{\prime}} , A_{-1}^{\prime\prime}=A_{-1}|_{V^{\prime\prime}} , and A_{-1}=A_{-1}^{\prime}\oplus A_{-1}^{\prime\prime} . Then since A_{-1}^{\prime} has 0 as the only eigenvalue and A_{-1}^{\prime\prime} has no integer as an eigenvalue, there exists an invertible matrix \tilde{A}_{-1}^{\prime\prime} of the same size as A_{-1}^{\prime\prime} such that the differential equation (\ref{eq:ode}) has a matrix of fundamental solutions of the form X(y)\begin{pmatrix}y^{A_{-1}^{\prime}}&0\\ 0&y^{\tilde{A}_{-1}^{\prime\prime}}\end{pmatrix} with X(y)\in\mathrm{GL}_{k}(\mathcal{M}_{D_{\alpha}(0)}(D_{\alpha}(0))) such that (X(y)|_{V^{\prime}})|_{y=0}=\mathrm{id}_{V^{\prime}} , see Proposition 3.4 in [ 3 ] for instance. Here \mathcal{M}_{D_{\alpha}(0)}(D_{\alpha}(0)) is the sheaf of meromorphic functions on D_{\alpha}(0). Then since \mathrm{Ker}\left(\begin{pmatrix}\exp(2\pi iA_{-1}^{\prime})&0\\ 0&\exp(2\pi i\tilde{A}_{-1}^{\prime\prime})\end{pmatrix}-I_{k}\right)=\mathrm{Ker\,}A_{-1}, the same argument as in the previous lemma shows the desired result. ∎ 4.5 Middle convolution functor and de Rham functor In this section, we give a proof of the following theorem. Theorem 4.8 . Let \nabla_{A}\in\mathrm{Pf}(\log\mathcal{A}) be a logarithmic Pfaffian system satisfying Assumption 4.3 with respect to a parameter \lambda\in\mathbb{C}\backslash\mathbb{Z} . Let \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} be the character defined by \chi(1)=\exp(2\pi i\lambda) . Then there exists an isomorphism of local systems on M(\mathcal{A}) , \mathrm{MC}_{\chi}\circ\mathrm{DR}_{\mathrm{Pf}}(\nabla_{A})\cong\mathrm{DR}_{\mathrm{Pf}}\circ\mathrm{mc}_{\lambda}(\nabla_{A}). Let us recall the \mathbb{C}[\pi_{1}(M(\mathcal{A}),m_{0})] -module map \bigoplus_{q\in Q_{n+2}^{m_{0},\infty}}H_{1}(D^{*}_{\alpha}(q),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})\rightarrow H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}), in Proposition 2.14 . Since, as we noted in Remark 2.15 , this map is injective, we regard the source space as a submodule of the target space. Since D^{*}_{\alpha}(q) is homotopy equivalent to S^{1} , the homology group H_{1}(D^{*}_{\alpha}(q),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}) has a simple description as follows. Lemma 4.9 . Let \mathcal{L} be a local system on S^{1} and \gamma be a simple loop around S^{1} with a base point d\in S^{1} . Let \tilde{\gamma} be a lift of \gamma to the universal covering space \pi_{d}\colon\widetilde{S^{1}}_{d}\rightarrow S^{1} . Denote the \pi_{1}(S^{1},d) -invariant subspace of the stalk \mathcal{L}_{d} by \mathcal{L}_{d}^{\pi_{1}(S^{1},d)} . Then the map \mathcal{L}_{d}^{\pi_{1}(S^{1},d)}\longrightarrow H_{1}(S^{1},\mathcal{L});\quad v\longmapsto[\tilde{\gamma}\otimes v] is an isomorphism of vector spaces. Therefore, in particular for q\in Q_{n+1}^{m_{0},\infty} , we have \mathrm{dim}_{\mathbb{C}}H_{1}(D^{*}_{\alpha}(q),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L})_{y}\boxtimes\chi_{(z-y)})\\ =\begin{cases}\mathrm{dim}_{\mathbb{C}}\mathrm{Ker\,}A_{H}&\text{ if}\quad q=q_{H}\text{ for }H\in\mathcal{A}\backslash\mathcal{A}_{Y},\\ 0&\text{ if}\quad q=y_{0},\\ \mathrm{dim}_{\mathbb{C}}\mathrm{Ker}\left(\sum_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}A_{H}+\lambda\right)&\text{ if}\quad q=\infty,\end{cases} under Assumption 4.3 . Proof. The first statement is standard, see [ 20 ] for instance. For the second statement, let us recall that the local system ((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}) on \mathbb{C}\backslash Q_{n+1}^{m_{0}} is defined as the horizontal sections of the connection \nabla_{A,m_{0}}^{\lambda}=\nabla_{A}|_{\mathrm{pr}_{1}^{-1}(m_{0})}\otimes\nabla_{\lambda}, as we saw in Section 4.3 . Then Proposition 4.7 shows that \mathrm{dim}_{\mathbb{C}}((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})_{q_{0}}^{\pi_{1}(D^{*}_{\alpha}(q),q_{0})}\\ =\begin{cases}\mathrm{dim}_{\mathbb{C}}\mathrm{Ker\,}A_{H}&\text{ if}\quad q=q_{H}\text{ for }H\in\mathcal{A}\backslash\mathcal{A}_{Y},\\ \mathrm{dim}_{\mathbb{C}}\mathrm{Ker\,}\lambda&\text{ if}\quad q=y_{0},\\ \mathrm{dim}_{\mathbb{C}}\mathrm{Ker}\left(\sum_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}A_{H}+\lambda\right)&\text{ if}\quad q=\infty.\end{cases} Here q_{0}\in D^{*}_{\alpha}(q) are suitably chosen base points for q\in Q_{n+1}^{m_{0},\infty} . Therefore the result follows from the first statement. ∎ Proposition 4.10 . Under Assumption 4.3 , the period map along Y at m_{0} gives an isomorphism \bigoplus_{q\in Q_{n+2}^{m_{0},\infty}}H_{1}(D^{*}_{\alpha}(q),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})\cong\mathrm{DR}_{\mathrm{Pf}}\left(\nabla^{K}_{c_{\lambda}(A)}\oplus\nabla^{L}_{c_{\lambda}(A)}\right) as local systems. Proof. For each q\in Q_{n+2}^{m_{0},\infty}\backslash\{y_{0}\} , let X_{q}(y) be the matrix of fundamental solutions for \nabla_{A,m_{0}}^{\lambda}=\nabla_{A}|_{\mathrm{pr}_{1}^{-1}(m_{0})}\otimes\nabla_{\lambda} near a base point q_{0}\in D^{*}_{\alpha}(q) , which satisfies the condition in Proposition 4.7 . Then we have ((\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})_{q_{0}}^{\pi_{1}(D^{*}_{\alpha}(q),q_{0})}\\ =\begin{cases}\left\{X_{q}(y)v\,\middle|\,v\in\mathrm{Ker\,}A_{H}\right\}&\text{ if }q=q_{H}\text{ for }H\in\mathcal{A}\backslash\mathcal{A}_{Y},\\ \left\{X_{q}(y)v\,\middle|\,v\in\mathrm{Ker}\left(\sum_{H\in\mathcal{A}\backslash\mathcal{A}_{Y}}A_{H}+\lambda\right)\right\}&\text{ if }q=\infty.\end{cases} Since the above X_{q}(y)v is holomorphic at y=q satisfying \left.X_{q}(y)v\right|_{y=q}=v, Cauchy’s integral formula and the residue theorem show that \int_{\gamma_{q}}X_{q}(y)v(y_{0}-y)^{\lambda}\,\frac{dy}{y-q_{H}}=\begin{cases}2\pi i(y_{0}-q_{H})^{\lambda}v&\text{ if }q=q_{H},\\ 2\pi iy_{0}^{-\lambda}v&\text{ if }q=\infty,\\ 0&\text{otherwise},\end{cases} with the simple loop \gamma_{q} around q in D^{*}_{\alpha}(q) with the base point q_{0} . This shows that the period map along Y at m_{0} gives an injection \bigoplus_{q\in Q_{n+2}^{m_{0},\infty}}H_{1}(D^{*}_{\alpha}(q),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})\xhookrightarrow{\mathrm{Per}_{Y,m_{0}}}\mathrm{DR}_{\mathrm{Pf}}\left(\nabla^{K}_{c_{\lambda}(A)}\oplus\nabla^{L}_{c_{\lambda}(A)}\right)_{m_{0}}. Moreover, since the dimension of both sides are equal by the previous lemma, this map gives an isomorphism as desired. ∎ Proof of Theorem 4.8 . First recall that the functor \mathrm{DR}_{\mathrm{Pf}} is an exact functor, since \mathrm{DR} is an equivalence of categories and the pullback functor j^{*} is exact. Also recall that the middle convolution functor \mathrm{MC}_{\chi} can be obtained as the cokernel of the map \bigoplus_{q\in Q_{n+2}^{m_{0},\infty}}H_{1}(D^{*}_{\alpha}(q),(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)})\rightarrow H_{1}(\mathbb{C}\backslash Q_{n+1}^{m_{0}},(\iota_{\mathbf{x}_{0}}^{*}\mathcal{L}_{A})_{y}\boxtimes\chi_{(z-y)}), and similarly, the middle convolution functor \mathrm{mc}_{\lambda} can be obtained as the cokernel of the map \nabla_{c_{\lambda}(A)}^{K}\oplus\nabla_{c_{\lambda}(A)}^{L}\hookrightarrow\nabla_{c_{\lambda}}(A) . Therefore, the exactness of \mathrm{DR}_{\mathrm{Pf}} implies the desired isomorphism by Theorem 4.4 and Proposition 4.10 . ∎ 5 Riemann-Hilbert problem for logarithmic Pfaffian systems with constant coefficients As in the previous section, we consider an affine hyperplane arrangement \mathcal{A} in \mathbb{C}^{l} with good line Y and the category \mathrm{Pf}(\log\mathcal{A}) of logarithmic Pfaffian systems with respect to \mathcal{A} . Let us consider the following variant of the Riemann-Hilbert problem, which asks for a characterization of the essential image of the de Rham functor \mathrm{DR}_{\mathrm{Pf}}\colon\mathrm{Pf}(\log\mathcal{A})\longrightarrow\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}). Problem 5.1 . Let \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) be a local system. Does there exist a logarithmic Pfaffian system \nabla_{A}\in\mathrm{Pf}(\mathrm{log}(\mathcal{A})) such that \mathrm{DR}_{\mathrm{Pf}}(\nabla_{A})\cong\mathcal{L} ? We call such a logarithmic Pfaffian system \nabla_{A} a solution for \mathcal{L} . By combining results obtained in the previous sections, we can show the middle convolution functor \mathrm{MC}_{\chi} preserves the solvability of the above Riemann-Hilbert problem under Assumption 4.3 . Theorem 5.2 . Let \mathcal{L}\in\mathrm{Loc}(M(\mathcal{A}),\mathbb{C}) be a local system satisfying the property \wp , and \chi\colon\mathbb{Z}\rightarrow\mathbb{C}^{\times} be a nontrivial character. Then the following hold: 1. If \mathcal{L} admits a solution \nabla_{A}\in\mathrm{Pf}(\log\mathcal{A}) satisfying Assumption 4.3 with respect to a parameter \lambda\in\mathbb{C}\backslash\mathbb{Z} such that \chi(1)=\exp(2\pi i\lambda) , then the local system \mathrm{MC}_{\chi}(\mathcal{L}) also admits a solution as well. 2. If \mathrm{MC}_{\chi}(\mathcal{L}) admits a solution \nabla_{A^{\prime}}\in\mathrm{Pf}(\log\mathcal{A}) satisfying Assumption 4.3 with respect to a parameter \lambda^{\prime}\in\mathbb{C}\backslash\mathbb{Z} such that \chi(1)=\exp(-2\pi i\lambda^{\prime}) , then the local system \mathcal{L} also admits a solution as well. Proof. Suppose that \mathcal{L} admits a solution \nabla_{A}\in\mathrm{Pf}(\log\mathcal{A}) satisfying Assumption 4.3 with respect to a parameter \lambda\in\mathbb{C}\backslash\mathbb{Z} such that \chi(1)=\exp(2\pi i\lambda) . Then Theorem 4.8 shows that \mathrm{MC}_{\chi}(\mathcal{L})\cong\mathrm{MC}_{\chi}\circ\mathrm{DR}_{\mathrm{Pf}}(\nabla_{A})\cong\mathrm{DR}_{\mathrm{Pf}}(\mathrm{mc}_{\lambda}(\nabla_{A})). Thus \mathrm{mc}_{\lambda}(\nabla_{A}) gives a solution for \mathrm{MC}_{\chi}(\mathcal{L}) , which proves the first statement. The second statement can be shown in the same way by using the fact that \mathrm{MC}_{\chi^{-1}}\circ\mathrm{MC}_{\chi}(\mathcal{L})\cong\mathcal{L}, under the property \wp , see Theorem 2.12 . ∎