Theory of almost algebraic Poincaré complexes and local combinatorial Hirzebruch formula (Q5953557)

An analog of the Hirzebruch formula with nonflat coefficients for combinatorial manifolds is given. For this purpose, almost algebraic Poincaré complexes, a new algebraic category which has necessary properties for constructing invariants of the signature type for combinatorial manifold \(X\) with a local coefficient system induced by fibres of a vector bundle over the manifold, is introduced. Algebraic study of almost algebraic Poincaré complexes leads algebraic definition of signature of the complexes. If \(X\) is compact, its sufficiently fine simplicial subdivision induces an almost algebraic Poincaré complex. The signature of this complex is the combinatorial version of the left hand side of the Hirzebruch formula, and has an alternative representation in terms of \(L\)-genus (Theorem 7). The outline of the paper is as follows: Signature was defined by using \(H^{2k} (X,\mathbb{C})\), where \(X\) is a compact orientable \(4k\)-dimensional manifold. The cohomology was generalized to the cohomology with coefficients in a flat bundle and then in an almost flat bundle [\textit{A. Connes}, \textit{M. Gromov}, and \textit{H. Moscovici}, C. R. Acad. Sci. Paris, Sér. I, 310, No. 5, 273-277 (1990; Zbl 0693.53007)]. Almost flat bundle is defined by almost representation with respect to \(F\), a finite subset of \(\pi=\pi_1(X)\), \(\sigma:\pi\to U(n)\), \[ \sigma (g^{-1})= \sigma(g)^{-1},\;\|\sigma \|_F= \sup\biggl \{\bigl\|\sigma(gh)- \sigma(g) \sigma(h) \bigr\|: g,h,gh \in F\biggr\}\leq \varepsilon. \] The author also introduced the notion of asymptotic representation, which is a sequence of almost-representation approximating to a representation [\textit{V. M. Manuilov} and \textit{A. S. Mishchenko}, Math. Sb. 189, No. 10, 53-74 (1998) translation from Sb. Math. 189, 1485-1504 (1998; Zbl 0932.46067)]. They are geometric prototype of almost algebraic Poincaré complexes reviewed in Sect. 1 (Introduction). In Sect. 2, algebraic Poincaré complexes [\textit{A. S. Mishchenk}, Izvest. Acad. Sci. USSR 34 (1970) 501-514, English transl. Math. USSR Izv. 4 (1970), 505-519 (1971; Zbl 0232.55015)] is reviewed. An algebraic Poincaré complex is an algebraic complex with the Poincaré duality. It is a chain complex \(C=\oplus^n_{k=0}C_k\) of free \(\Lambda\)-modules, \(\Lambda\) a ring with the unit 1 and unital involution *, with a homomorphism of Poincaré duality \(D:C^*\to C\), \(\deg D=n\), where \(C^*=\text{HOM}_\Lambda (C, \Lambda)\), such that \[ d_kD_k +(-1)^{k+1} D_{k-1}d^a_{n-k+1} st=0,\quad D_k=(-1)^{k(n-k)} D^*_{n-k}. \] Let \(F_k=i^{k(k-1)} D_k\). The cone Cone\((F)\) of \(F\) is \((A,H)\), \(A_k=C_k \oplus C^*_{n-k+1}\), \[ H_k= \left(\begin{matrix} d_k & F_{k-1}\\ 0 & d^*_{n-k+2} \end{matrix}\right). \] \(A\) is decomposed as the direct sum of even and odd components \(A_{ev}= C_{ev}\oplus C^*_{odd}\), \(A_{odd}= C_{odd}\oplus C^*_{ev}= A^*_{ev}\). By this decomposition, \(H\) induces maps \(H_{ev}: A_{ev}\to A_{odd}\) and \(H_{odd}: A_{odd}\to A_{ev}\). Let \(B_0=\ker H_{ev}\). Then we have \(A_{ev}=B_0 \oplus B_1\), \(A^*_{ev}= B_0^*\oplus B^*_1\) and \(H_{ev}\), \(H_{odd}\) induce selfadjoint isomorphisms \(\beta_{ev}: B_1\to B^*_1\), \(\alpha_{odd}: B^*_0 \to B_0\), which are independent from the choice of decomposition of the module \(A_{ev}\). Let \[ S=\beta_{ev} \oplus\alpha_{odd}: (B_1\oplus B^*_0)= M\to M^*=(B_1 \oplus B^*_0)^*. \] Then it defines an element \([M,S]\) of the Hermitian \(K\)-group \(K^h_0(\Lambda)\) of \(\Lambda\). The signature \(\text{sign} (C,d,D)\) of the algebraic Poincaré complex \((C,d,D)\) is defined to be \([M,S]\) (Sect. 2). If \(\Lambda\) is a \(C^*\)-algebra, we have \([C,d+d^*+F] =(C,d,D)\) (Sect. 3). In the rest, \(C\) is assumed to be finite-dimensional \(\mathbb{C}\)-vector space with a Hermitian structure. Let \(C\) be a graded linear space with an involution \(\eta: C\to C\), \(\eta|_{C_i} =(-1)^i\). A pair \((C,d)\), \(d_i:C_i \to C_i\to C_{i-1}\) is called \((\alpha,A)\)-almost chain complex, if it satisfies \[ \|d\|\leq A,\;\|d^2\|\leq \alpha, \] (Definition 1). \(A>1\) is said to be large variable, and \(0<\alpha <1\) is said to be small variable. An admissible function \(F=(F_s (\alpha,A)\), \(F_l(A))\) is a pair of positive coefficients polynomials of positive rational powers of \(\alpha\) and \(A\) such that \(F_s(0,A)=0\). Admissible functions for the group of small variables \(\alpha_i\) and the group of large variables \(A_j\) are similarly defined. An \((\alpha,A)\)-almost chain complex \((C, d)\) is said to be \(F\)-almost acyclic if there is an admissible function \(F( \alpha, \delta,A)\) such that if \(\|dx\|\leq \delta\|x\|\), then there is an element \(y\) such that \[ \|y\|\leq F_l(A) \|x\|,\;\|x-dy \|\leq F_s(\alpha, \delta;A) \|x\|, \] (Definition 3). It is shown an \((\alpha,A)\)-almost chain complex is almost acyclic if and only if \(d+d^*\) is invertible (Theorems 1 and 2). Almost chain homomorphism \(f:C^{(1)}\to C^{(2)}\) and its cone Cone\((f)= C^{(1)} \oplus C^{(2)}\) are also defined (Definition 2 and (25), (26)). It is shown Cone\((gf)\) is almost acyclic if Cone\((f)\) and Cone\( (g)\) are \(F(\alpha,A)\)-acyclic (Theorem 5). These are explained in Sec. 4. The \(F(\alpha,\delta;A)\)-almost algebraic Poincaré complex of formal dimension \(n\) is the pair \(M=((C,d),D)\), where \((C,d)\) is the \((\alpha,A)\)-almost chain complex, and \(D:C^*\to C\) is the homogeneous homomorphism such that \[ \|D \|\leq A,\;\|Dd^*+dD \eta\|\leq\alpha,\;D^*-= D\eta^{n+1}, \] and Cone \((D)\) is the \(F(\alpha, \delta;A)\)-almost acyclic chain complex Definition 5). After showing, that the tensor product of almost algebraic Poincaré complexes is an almost algebraic Poincaré complex (Theorem 6), almost algebraic Poincaré complex with boundary \(M=((C,d), (C_0,d_0),D)\) is defined similar to almost algebraic Poincaré complex (Definition '7). Denoting \(\partial M=(C_0,d_0)\), \(M_1\) and \(M_2\) are said to be bordant if \(M_1\oplus (-M_2)= \partial W\), for some \(W\), where \(-M=((C,d), -D)\) if \(M=((C,d),D)\). An almost algebraic Poincaré complex \(M=((C,d),D)\), \(C=\oplus^n_{k=0} C_k\) is called elementary if \(C_k=0\) except one or two middle dimensions. If \(M\) is an elementary complex with \(n=2k\), \(D:C^*_k\to C_k\) is invertible and selfadjoint if \(k\) is even, skew-adjoint if \(k\) is odd. The signature of \(M\) is defined by \[ \text{sign}(M)=\text{sign} (D),\;k\text{ is even, sign} (M)=\text{sign}(iD), \;k\text{ is odd}. \] Since any almost algebraic Poincaré complex \(M\) is bordant to an elementary \(M_e\) (Lemma 5), signature of \(M\) is defined to be \(\text{sign}(M_e)\) if \(n\) is even \(\text{(sign}(M)=0\) if \(n\) is odd). By definition, we have \[ \text{sign} (M_1\oplus M_2)=\text{sign} (M_1)+\text{sign} (M_2), \] \[ \text{sign} (M_1\oplus M_2)= \text{sign} (M_1) \times\text{sign} (M_2). \] (Sect. 5). In Sect. 6, the last Section, an almost algebraic Poincaré complex is constructed from a compact combinatorial manifold \(X\) and a vector bundle \(V\) over \(X\). Then apply the preceding discussions, first mentioned analog of the Hirzebruch formula is derived. The author says this paper was written under strong influence of \textit{M. Gromov} [``Positive curvature, macroscopic dimension, spectral gaps and higher signatures'', Progr. Math. 132, 1-213 (1996; Zbl 0945.53022)]. Regrets are lack of proofs related to the bordism of almost algebraic Poincaré complexes and many miss quoting of formula numbers.