Search the corpus

    Definitions

    Contents

    Point-set topology

    (The prefix "locally")

    For a property \(P\) of spaces, "\(X\) is locally \(P\)" has two definitions in use: every \(x\in X\) has some neighborhood with property \(P\), or every \(x\in X\) has a neighborhood basis of sets with property \(P\). Local connectedness and local path connectedness use the second.

    Limit point

    Let \(X\) be a topological space, \(A\subseteq X\), and \(x\in X\). The point \(x\) is a limit point of \(A\) if every neighborhood of \(x\) meets \(A\sm\ts{x}\). A limit point is also called an accumulation point.

    A point \(x\in X\) is a limit point of \(A\subseteq X\) if and only if \(x\in \cl_{X}(A\sm\ts{x})\), the closure of \(A\sm\ts{x}\) in \(X\).

    What is a limit point of a subset $A \subseteq X$?

    Basis for a topology

    Let \(X\) be a set. A collection \(\mcb\) of subsets of \(X\) is a basis for a topology on \(X\) if

    • every \(x\in X\) lies in some \(B\in\mcb\), and

    • whenever \(x \in B_1 \intersect B_2\) with \(B_1, B_2\in\mcb\), there exists \(B_3\in\mcb\) with \(x\in B_3 \subseteq B_1 \intersect B_2\).

    The topology generated by \(\mcb\) is the collection of all \(U\subseteq X\) such that for each \(x\in U\) there exists \(B\in\mcb\) with \(x\in B \subseteq U\).

    Let \(\mcb\) be a basis for a topology on a set \(X\). The topology generated by \(\mcb\) is a topology on \(X\), and its open sets are exactly the unions of elements of \(\mcb\).

    [@Mun00].

    Boundary of a subset

    Let \(X\) be a topological space and \(A\subseteq X\). The boundary of \(A\) is \(\bd A \coloneqq \cl_X(A) \sm A^\circ\), where \(\cl_X(A)\) is the closure of \(A\) and \(A^\circ\), its interior, is the set of interior points of \(A\).

    A point \(p\in X\) lies in \(\bd A\) if and only if every neighborhood of \(p\) intersects both \(A\) and \(X\sm A\).

    Bounded set in a metric space

    Let \((X, d)\) be a metric space. A subset \(S \subseteq X\) is bounded if there exists \(m\in \RR\) such that \(d(x, y) < m\) for all \(x, y\in S\).

    Comparability of topologies

    Let \(\tau_1\) and \(\tau_2\) be topologies on a set \(X\). The topology \(\tau_1\) is finer than \(\tau_2\) if \(\tau_2\subseteq\tau_1\), and coarser than \(\tau_2\) if \(\tau_1\subseteq\tau_2\). The topologies \(\tau_1\) and \(\tau_2\) are comparable if \(\tau_1\subseteq\tau_2\) or \(\tau_2\subseteq\tau_1\).

    See [@Mun00].

    Characterizations of connectedness

    Let \(X\) be a topological space. A separation of \(X\) is a pair \((U, V)\) of disjoint nonempty open subsets of \(X\) with \(X = U\union V\). The space \(X\) is disconnected if it has a separation, and connected otherwise.

    For a topological space \(X\), the following are equivalent:

    1. \(X\) is connected.

    2. The only subsets of \(X\) that are both open and closed are \(\emptyset\) and \(X\).

    3. Every continuous map from \(X\) to the discrete space \(\ts{0, 1}\) is constant.

    Let \(X\) be a topological space and \(Y\subseteq X\) a subspace. Then \(Y\) is disconnected if and only if there exist disjoint nonempty subsets \(A, B\subseteq Y\) with \(Y = A\union B\) such that neither contains a limit point of the other, that is, \(\cl_X(A)\intersect B = \emptyset\) and \(A\intersect\cl_X(B) = \emptyset\).

    The subspace \(\QQ\subseteq\RR\) is disconnected, and every connected subspace of \(\QQ\) has at most one point, so the connected components of \(\QQ\) are its singletons. For rationals \(p < q\) and an irrational \(r\) with \(p<r<q\), the sets \(\QQ\intersect(-\infty, r)\) and \(\QQ\intersect(r, \infty)\) form a separation of \(\QQ\).

    [@Mun00].

    Comparability of topologies

    Let \(\tau_1\) and \(\tau_2\) be topologies on a set \(X\). The topology \(\tau_1\) is finer (also stronger or larger) than \(\tau_2\) if \(\tau_2\subseteq\tau_1\), and coarser (also weaker or smaller) than \(\tau_2\) if \(\tau_1\subseteq\tau_2\). The topologies \(\tau_1\) and \(\tau_2\) are comparable if \(\tau_1\subseteq\tau_2\) or \(\tau_2\subseteq\tau_1\).

    Let \(X\) be a set and let \(\mathrm{Top}(X)\) be the set of topologies on \(X\), partially ordered by inclusion. Then \(\mathrm{Top}(X)\) is a complete lattice [@Mun00]:

    • the infimum of a nonempty family \((\tau_i)_{i\in I}\) is the intersection \(\bigcap_{i\in I}\tau_i\);

    • the supremum of \((\tau_i)_{i\in I}\) is the topology generated by \(\bigcup_{i\in I}\tau_i\), the intersection of all topologies containing every \(\tau_i\);

    • the greatest element is the discrete topology \(\mathcal P(X)\) and the least element is the indiscrete topology \(\ts{\emptyset,X}\).

    The union of two topologies need not be a topology. On \(X=\ts{a,b,c}\), the union of \(\ts{\emptyset,\ts{a},X}\) and \(\ts{\emptyset,\ts{b},X}\) contains \(\ts{a}\) and \(\ts{b}\) but not \(\ts{a,b}\), so the supremum is in general strictly larger than the union.

    Let \(f\colon X\to Y\) be a map of sets, let \(\tau_1\subseteq\tau_2\) be topologies on \(X\), and let \(\sigma_1\subseteq\sigma_2\) be topologies on \(Y\).

    1. If \(f\colon(X,\tau_1)\to(Y,\sigma_2)\) is continuous, then \(f\colon(X,\tau_2)\to(Y,\sigma_1)\) is continuous.

    2. If \(f\colon(X,\tau_2)\to(Y,\sigma_1)\) is an open map, then \(f\colon(X,\tau_1)\to(Y,\sigma_2)\) is an open map.

    3. If a sequence \((x_k)\) converges to \(x\) in \((X,\tau_2)\), then it converges to \(x\) in \((X,\tau_1)\).

    For \(n\geq1\), the Zariski topology \(\tau_{\zar}\) on \(\CC^n\), whose closed sets are the common zero sets of families of polynomials, is strictly coarser than the Euclidean topology \(\tau_{\an}\). Zero sets of polynomials are closed in \(\tau_{\an}\) because polynomials are continuous, so \(\tau_{\zar}\subseteq\tau_{\an}\). The open unit ball is in \(\tau_{\an}\) but not in \(\tau_{\zar}\): every nonempty Zariski-open subset of \(\CC^n\) contains the complement of the zero set of a nonzero polynomial, which is dense in \(\tau_{\an}\) and hence not contained in the unit ball.

    Connected components

    Let \(X\) be a topological space. For \(x, y\in X\) write \(x\sim y\) if there is a connected subspace \(C\subseteq X\) with \(x, y\in C\). The equivalence classes of \(\sim\) are the connected components of \(X\).

    The relation \(\sim\) is an equivalence relation on \(X\).

    See [@Mun00].

    Closed and open maps

    Let \(X\) and \(Y\) be topological spaces and \(f\colon X\to Y\) a map. The map \(f\) is a closed map if \(f(A)\) is closed in \(Y\) for every closed \(A\subseteq X\), and an open map if \(f(U)\) is open in \(Y\) for every open \(U\subseteq X\) [@Mun00].

    Neither condition implies or is implied by continuity, and neither implies the other.

    The projection \(\pi_1\colon\RR\times\RR\to\RR\), \((x,y)\mapsto x\), is continuous and open but not closed: the hyperbola \(\ts{(x,y)\st xy=1}\) is closed in \(\RR\times\RR\), and its image \(\RR\setminus\ts{0}\) is not closed in \(\RR\) [@Mun00].

    Let \(f\colon X\to Y\) be a continuous map of topological spaces.

    1. If \(f\) is surjective and open or closed, then \(f\) is a quotient map; a quotient map need not be open or closed [@Mun00].

    2. If \(X\) is compact and \(Y\) is Hausdorff, then \(f\) is closed [@Mun00].

    3. If \(X\) is compact, \(Y\) is Hausdorff, and \(f\) is bijective, then \(f\) is a homeomorphism [@Mun00].

    Compact space

    A topological space \((X, \tau)\) is compact if every open cover of \(X\) has a finite subcover: for every family \((U_j)_{j\in J}\) in \(\tau\) with \(X = \bigcup_{j\in J} U_j\), there is a finite subset \(J'\subseteq J\) with \(X = \bigcup_{j\in J'} U_j\).

    Closed set

    Let \(X\) be a topological space. A subset \(U \subseteq X\) is closed in \(X\) if its complement \(X\sm U\) is open in \(X\). If \(Y \subseteq X\) carries the subspace topology and \(U \subseteq Y\), then \(U\) is closed in the subspace \(Y\) if \(Y\sm U\) is open in \(Y\).

    Let \(X\) be a topological space and \(U\subseteq X\).

    1. \(U\) is closed in \(X\) if and only if every limit point of \(U\) in \(X\) lies in \(U\).

    2. \(U\) is closed in \(X\) if and only if \(\cl_X(U) = U\), where \(\cl_X(U)\) is the closure of \(U\) in \(X\).

    3. If \(Y\subseteq X\) is a subspace and \(U\subseteq Y\), then \(U\) is closed in \(Y\) if and only if \(U = Y \intersect V\) for some closed subset \(V\) of \(X\).

    4. If \(Y\subseteq X\) is a subspace and \(U\subseteq Y\), then the closure of \(U\) in \(Y\) is \(\cl_Y(U) = \cl_X(U)\intersect Y\).

    [@Mun00].

    Closed map

    Let \(X\) and \(Y\) be topological spaces. A map \(f\colon X\to Y\) is closed if \(f(C)\) is closed in \(Y\) for every closed subset \(C\subseteq X\).

    Closure of a set

    Let \(X\) be a topological space and \(U\subseteq X\). The closure of \(U\) in \(X\) is the intersection of all closed subsets of \(X\) containing \(U\), \[ \cl_X(U) \coloneqq \Intersect_{\substack{B\supseteq U \\ B \text{ closed in } X}} B ,\] also written \(\bar U\).

    If \(Y\subseteq X\) is a subspace containing \(U\), the closure of \(U\) in \(Y\) is \(\cl_Y(U) = \cl_X(U) \intersect Y\) [@Mun00]. A point \(x\in X\) lies in \(\cl_X(U)\) if and only if every neighborhood of \(x\) in \(X\) intersects \(U\) [@Mun00].

    Compact as Hausdorff and quasicompact

    A topological space \((X, \tau)\) is compact if \(X\) is Hausdorff and quasicompact, that is, every open cover of \(X\) has a finite subcover: for every family \((U_j)_{j\in J}\) in \(\tau\) with \(X = \bigcup_{j\in J} U_j\), there is a finite subset \(J'\subseteq J\) with \(X = \bigcup_{j\in J'} U_j\).

    This is the Bourbaki convention. A two-point set with the indiscrete topology satisfies the open-cover condition, so it is compact when compactness is defined by that condition alone; it is not Hausdorff, so in the Bourbaki convention it is quasicompact and not compact.

    Continuous map

    Let \(X\) and \(Y\) be topological spaces. A map \(f\colon X\to Y\) is continuous if \(\inverseof{f}(U)\) is open in \(X\) for every open set \(U \subseteq Y\).

    Cover of a space or subset

    Let \(X\) be a set and \(\mathcal U = \ts{U_\alpha}_{\alpha\in I}\) a collection of subsets of \(X\). The collection \(\mathcal U\) is a cover of \(X\) if \(X = \Union_{\alpha\in I} U_\alpha\), and a cover of a subset \(A\subseteq X\) if \(A\subseteq \Union_{\alpha\in I} U_\alpha\). If \(X\) is a topological space and every \(U_\alpha\) is open, \(\mathcal U\) is an open cover.

    Dense subspace

    Let \(X\) be a topological space. A subset \(Q\subseteq X\) is dense in \(X\) if every neighborhood of every point \(x\in X\) intersects \(Q\).

    A subset \(Q\subseteq X\) is dense in \(X\) if and only if its closure is \(\cl_X(Q) = X\).

    See [@Mun00].

    In terms of closure, when is $A \subset X$ dense in $X$?

    Diameter of a subset of a metric space

    Let \((X, d)\) be a metric space and \(A\subseteq X\) a nonempty subset. The diameter of \(A\) is \(\diam(A)\coloneqq\sup_{p, q\in A} d(p, q)\in[0,\infty]\).

    Discrete topology

    Let \(X\) be a set. The discrete topology on \(X\) is the topology \(\tau = \mathcal{P}(X)\), in which every subset of \(X\) is open.

    What is the discrete topology on a set $X$?
    Which sets are open in the discrete topology on $X$?

    Indiscrete topology

    Let \(X\) be a set. The indiscrete topology on \(X\) is the topology \(\tau = \ts{\emptyset, X}\).

    What is the indiscrete topology on a set $X$?

    First countable space

    A topological space \(X\) is first countable if every point \(x\in X\) has a countable neighborhood basis.

    What does it mean for a space to be first countable?

    Hausdorff space

    A topological space \(X\) is Hausdorff, or \(T_2\), if for all \(p, q \in X\) with \(p\neq q\) there exist disjoint open sets \(U, V\subseteq X\) with \(p\in U\) and \(q\in V\).

    In a Hausdorff space, every sequence converges to at most one point [@Mun00].

    What does it mean for a space to be Hausdorff?

    Injective map

    Let \(A\) and \(B\) be sets. A map \(\iota\colon A\to B\) is injective if for all \(x, y\in A\), \(\iota(x) = \iota(y)\) implies \(x = y\).

    If \(A\neq\emptyset\), then \(\iota\colon A\to B\) is injective if and only if it has a left inverse: a map \(p\colon B\to A\) with \(p\circ\iota = \id_A\). The empty map \(\emptyset\to B\) with \(B\neq\emptyset\) is injective and has no left inverse. For a continuous injective map of topological spaces, a left inverse need not be continuous: the inclusion \(S^1\injects D^2\) has no continuous left inverse.

    [@DF04].

    Injective map

    Let \(A\) and \(B\) be sets. A map \(\iota\colon A\to B\) is injective if for all \(x, y\in A\), \(\iota(x) = \iota(y)\) implies \(x = y\).

    If \(A\neq\emptyset\), then \(\iota\colon A\to B\) is injective if and only if it has a left inverse: a map \(p\colon B\to A\) with \(p\circ\iota = \id_A\). The empty map \(\emptyset\to B\) with \(B\neq\emptyset\) is injective and has no left inverse. For a continuous injective map of topological spaces, a left inverse need not be continuous: the inclusion \(S^1\injects D^2\) has no continuous left inverse.

    [@DF04].

    Interior point and interior

    Let \(X\) be a topological space and \(A\subseteq X\). A point \(p\in A\) is an interior point of \(A\) if there exists a neighborhood \(U\) of \(p\) with \(U\subseteq A\). The interior \(A^\circ\) of \(A\) is the set of interior points of \(A\).

    Isolated point

    Let \(X\) be a topological space and \(A\subseteq X\). A point \(p\in A\) is an isolated point of \(A\) if \(p\) is not a limit point of \(A\).

    A point \(p\in A\) is an isolated point of \(A\) if and only if there is a neighborhood \(U\) of \(p\) with \(U\cap A = \ts{p}\).

    Limit point compact space

    A topological space \(X\) is limit point compact if every infinite subset of \(X\) has a limit point in \(X\).

    What does it mean for a space to be limit point compact?

    Locally connected space

    A topological space \(X\) is locally connected if for every \(x\in X\) and every neighborhood \(U\) of \(x\) there is a connected open set \(V\) with \(x\in V\subseteq U\).

    See [@Mun00].

    Locally compact space

    A topological space \(X\) is locally compact if every \(x\in X\) has a neighborhood \(U\) contained in a compact subset \(K\subseteq X\).

    Every compact space \(X\) is locally compact, with \(U = K = X\). The converse fails: for \(n\geq 1\), \(\RR^n\) is locally compact, since each \(x\) lies in an open ball whose closure is compact, but \(\RR^n\) is not compact.

    The following subspaces of Euclidean space are not locally compact:

    • \(\QQ\subseteq\RR\);

    • \(\ts{\mathbf 0} \union \ts{(x, y)\in\RR^2 \st x>0} \subseteq\RR^2\), in which the origin has no neighborhood contained in a compact subset.

    Some authors also require a locally compact space to be Hausdorff. For Hausdorff \(X\) the condition above is equivalent to every point having a neighborhood whose closure is compact.

    [@Mun00].

    What does it mean for a space to be locally compact?

    Locally finite collection of subsets

    Let \(X\) be a topological space. A collection \(\mcs\) of subsets of \(X\) is locally finite if every \(x\in X\) has a neighborhood \(U\) such that \(U\intersect S\neq\emptyset\) for only finitely many \(S\in\mcs\).

    Local homeomorphism

    Let \(X\) and \(Y\) be topological spaces. A continuous map \(f\colon X\to Y\) is a local homeomorphism if every \(x\in X\) has an open neighborhood \(U_x\) such that \(f(U_x)\) is open in \(Y\) and \(f\vert_{U_x}\colon U_x \to f(U_x)\) is a homeomorphism.

    Every covering space \(p\colon\tilde X\to X\) is a local homeomorphism. For a sheaf of sets \(\mathcal F\) on a space \(X\), the projection from its étale space to \(X\) is a local homeomorphism.

    When is a map $f: X \to Y$ a local homeomorphism?

    Locally homeomorphic spaces

    Let \(X\) and \(Y\) be topological spaces. The space \(X\) is locally homeomorphic to \(Y\) if every \(x\in X\) has an open neighborhood \(U_x\) that is homeomorphic to an open subset of \(Y\).

    If there is a local homeomorphism \(f\colon X\to Y\), then \(X\) is locally homeomorphic to \(Y\); the converse fails.

    The sphere \(S^2\) is locally homeomorphic to \(\RR^2\), by stereographic projection from each of two antipodal points, but there is no local homeomorphism \(f\colon S^2 \to \RR^2\): the image \(f(S^2)\) would be a nonempty subset of \(\RR^2\) that is open, because local homeomorphisms are open maps, and closed, because it is compact, hence equal to the noncompact space \(\RR^2\).

    What does it mean for $X$ to be locally homeomorphic to $Y$?

    Locally path-connected space

    A topological space \(X\) is locally path-connected if for every \(x\in X\) and every neighborhood \(U\) of \(x\) there exists a path-connected open set \(V\) with \(x\in V\subseteq U\).

    Equivalently, \(X\) has a basis consisting of path-connected open sets.

    [@Mun00].

    Neighborhood

    Let \(X\) be a topological space and \(x\in X\). A neighborhood of \(x\) is an open set \(U\subseteq X\) with \(x\in U\).

    See [@Mun00].

    Normal space

    A topological space \(X\) is normal if for all disjoint closed subsets \(A, B\subseteq X\) there exist disjoint open sets \(U, V\subseteq X\) with \(A\subseteq U\) and \(B\subseteq V\).

    What does it mean for a space to be normal?

    Neighborhood basis

    Let \(X\) be a topological space and \(p\in X\). A neighborhood basis at \(p\) is a collection \(\mathcal B_p\) of neighborhoods of \(p\) such that every neighborhood \(N\) of \(p\) contains some \(B\in\mathcal B_p\).

    What is a neighborhood basis at a point $x$?
    What collection of open sets forms a neighborhood basis at $x$?

    Open and closed maps

    Let \(X\) and \(Y\) be topological spaces and \(f\colon X\to Y\) a function. Then \(f\) is an open map if \(f(U)\) is open in \(Y\) for every open \(U\subseteq X\), and \(f\) is a closed map if \(f(C)\) is closed in \(Y\) for every closed \(C\subseteq X\).

    Paracompact space

    A topological space \(X\) is paracompact if every open cover of \(X\) has a locally finite open refinement that covers \(X\).

    The relation \(\mcv\leq\mcu\) if \(\mcv\) refines \(\mcu\) is a preorder on the set of open covers of \(X\).

    See [@Mun00].

    Quotient map

    Let \(X\) and \(Y\) be topological spaces. A map \(q\colon X\to Y\) is a quotient map if

    1. \(q\) is surjective, and

    2. for every \(U\subseteq Y\), \(U\) is open in \(Y\) if and only if \(\inverseof{q}(U)\) is open in \(X\)

    [@Mun00].

    Quotient map

    Let \(X\) and \(Y\) be topological spaces. A map \(q\colon X\to Y\) is a quotient map if \(q\) is surjective and, for every \(U\subseteq Y\), \(U\) is open in \(Y\) if and only if \(\inverseof{q}(U)\) is open in \(X\).

    The implication from \(U\) open to \(\inverseof{q}(U)\) open is continuity of \(q\); a quotient map is a continuous surjection that also satisfies the converse implication.

    A surjective map \(q\colon X\to Y\) is a quotient map if and only if \(q\) is continuous and \(q(V)\) is open in \(Y\) for every open subset \(V\subseteq X\) that is saturated with respect to \(q\). The same holds with “open” replaced by “closed” throughout.

    See [@Mun00].

    Path-connected space

    Let \(X\) be a topological space and \(I = [0,1]\). The space \(X\) is path connected if for every pair of points \(x\neq y\) in \(X\) there exists a path \(f\colon I \to X\) with \(f(0) = x\) and \(f(1) = y\).

    Path components

    Let \(X\) be a topological space. For \(x, y\in X\), put \(x\sim y\) if there exists a path-connected subspace \(U\subseteq X\) with \(x, y\in U\). The equivalence classes of \(\sim\) are the path components of \(X\).

    The relation \(\sim\) is an equivalence relation on \(X\), and \(x\sim y\) if and only if there is a path in \(X\) from \(x\) to \(y\).

    [@Mun00].

    Precompact subset

    Let \(X\) be a topological space. A subset \(A\subseteq X\) is precompact if its closure \(\cl_{X}(A)\) is compact.

    Product topology on $X\times Y$

    Let \((X, \tau_X)\) and \((Y, \tau_Y)\) be topological spaces. The product topology on \(X\times Y\) is the topology generated by the basis \[ \mathcal B\coloneqq\ts{U\times V \st U\in\tau_X,\ V\in\tau_Y}: \] a subset \(W\subseteq X\times Y\) is open if and only if it is a union of elements of \(\mathcal B\).

    See [@Mun00].

    Proper map

    Let \(X\) and \(Y\) be topological spaces. A continuous map \(f\colon X\to Y\) is proper if \(\inverseof{f}(K)\) is compact for every compact \(K\subseteq Y\).

    Let \(f\colon X\to Y\) be a continuous map.

    1. If \(Y\) is Hausdorff and locally compact, then \(f\) is proper if and only if \(f\) is a closed map and \(\inverseof{f}(\ts{y})\) is compact for every \(y\in Y\).

    2. If \(X\) is Hausdorff and \(Y\) is Hausdorff and locally compact, then \(f\) is proper if and only if \(f\) is universally closed: for every topological space \(Z\), the map \(f\times\id_Z\colon X\times Z\to Y\times Z\) is a closed map.

    3. If \(X\) is a metric space and \(Y\) is Hausdorff, then \(f\) is proper if and only if for every sequence \((x_k)_{k\geq1}\) in \(X\) such that each compact subset of \(X\) contains \(x_k\) for only finitely many \(k\), each compact subset of \(Y\) contains \(f(x_k)\) for only finitely many \(k\).

    [@Lee13].

    Quasicompact

    A topological space \(X\), not necessarily Hausdorff, is quasicompact if every open cover of \(X\) has a finite subcover. A quasicompact space is compact if it is also Hausdorff.

    Requiring compact spaces to be Hausdorff is the convention of Bourbaki and of algebraic geometry, where the Zariski topology on the spectrum \(\Spec A\) of a commutative ring \(A\) is always quasicompact and is Hausdorff if and only if every prime ideal of \(A\) is maximal. In [@Mun00] a space is compact under the open-cover condition alone, so compact in that sense means quasicompact in the Bourbaki convention.

    Product and box topologies

    Let \((X_i)_{i\in I}\) be a family of topological spaces, let \(X \coloneqq \prod_{i\in I} X_i\), and for \(j\in I\) let \(p_j\colon X\to X_j\) be the projection. The box topology on \(X\) is the topology generated by the basis \[ \mathcal B_{\Box} \coloneqq \ts{ \prod_{i\in I} U_i \st U_i \text{ open in } X_i \text{ for every } i }. \] The product topology on \(X\) is the topology generated by the basis \[ \mathcal B_{\prod} \coloneqq \ts{ \prod_{i\in I} U_i \st U_i \text{ open in } X_i \text{ for every } i, \text{ and } U_i = X_i \text{ for all but finitely many } i }. \]

    The product topology on \(X\) is the coarsest topology on \(X\) for which every projection \(p_j\) is continuous. It is the topology generated by the subbasis \(\ts{ \inverseof{p_j}(U) \st j\in I,\ U \text{ open in } X_j }\).

    Give \(X\) the product topology. For every topological space \(Y\) and every family of continuous maps \(f_i\colon Y\to X_i\), there is a unique continuous map \(f\colon Y\to X\) with \(p_i\circ f = f_i\) for every \(i\in I\). Hence \((X, (p_i)_{i\in I})\) is the product of the \(X_i\) in \(\Top\):

    Every element of \(\mathcal B_{\prod}\) lies in \(\mathcal B_{\Box}\), so the box topology is finer than the product topology. The two topologies coincide when \(I\) is finite.

    For the box topology, the universal property of the product fails. Let \(X = \prod_{n\geq 1}\RR\) and let \(\Delta\colon\RR\to X\), \(\Delta(t) = (t,t,\ldots)\), whose components are all \(\id_\RR\). Then \(\Delta\) is continuous for the product topology, but not for the box topology: \(\prod_{n\geq1}(-1/n, 1/n)\) is open in the box topology and \(\inverseof{\Delta}\qty{\prod_{n\geq1}(-1/n,1/n)} = \ts{0}\) is not open in \(\RR\).

    See [@Mun00].

    Refinement of a cover

    Let \(X\) be a topological space and let \(\mathcal U\) and \(\mathcal V\) be covers of \(X\). The cover \(\mathcal V\) is a refinement of \(\mathcal U\) if for every \(V\in\mathcal V\) there exists \(U\in\mathcal U\) with \(V\subseteq U\).

    Regular space

    A topological space \(X\) is regular if for every closed set \(F\subseteq X\) and every point \(x\in X\sm F\) there are disjoint open sets \(U, V\subseteq X\) with \(x\in U\) and \(F\subseteq V\).

    Retract

    Let \(X\) be a topological space, \(A\subseteq X\) a subspace, and \(\iota\colon A\injects X\) the inclusion. A retraction of \(X\) onto \(A\) is a continuous map \(r\colon X\to A\) with \(r\circ\iota=\id_A\), that is, \(r(a)=a\) for every \(a\in A\):

    The subspace \(A\) is a retract of \(X\) if a retraction of \(X\) onto \(A\) exists.

    Let \(A\subseteq X\) be a subspace with inclusion \(\iota\colon A\injects X\).

    1. Every retraction \(r\colon X\to A\) is surjective.

    2. Continuous maps \(r\colon X\to A\) with \(r\circ\iota=\id_A\) correspond bijectively to continuous maps \(\rho\colon X\to X\) with \(\rho(X)=A\) and \(\rho(a)=a\) for all \(a\in A\), via \(\rho=\iota\circ r\).

    If \(r\) is a retraction and \(a_0\in A\), then \(r_*\circ\iota_* = \id\) on \(\pi_1(A, a_0)\) and on \(H_n(A)\), so \(\iota_*\colon \pi_1(A, a_0)\to\pi_1(X, a_0)\) and \(\iota_*\colon H_n(A)\to H_n(X)\) are injective. For every point \(x_0\) of a topological space \(X\), the subspace \(\ts{x_0}\) is a retract of \(X\), via the constant map \(X\to\ts{x_0}\).

    What does it mean for a subspace $A \subset X$ to be a retract of $X$?

    If \(r\colon X\to A\) is a retraction and \(\iota\colon A\injects X\) the inclusion, then \(r\circ\iota=\id_A\), so \(\iota_*\) is injective on \(\pi_1\) and on homology. For every point \(x_0\) of a space \(X\), the constant map \(X\to\ts{x_0}\) is a retraction.

    Saturated subset

    Let \(p\colon X\to Y\) be a surjective map. A subset \(U\subseteq X\) is saturated with respect to \(p\) if for every \(y\in Y\) with \(U\cap \inverseof{p}(y)\neq\emptyset\) we have \(\inverseof{p}(y)\subseteq U\).

    A subset \(U\subseteq X\) is saturated with respect to \(p\) if and only if \(U = \inverseof{p}(B)\) for some \(B\subseteq Y\), in which case \(B = p(U)\).

    See [@Mun00].

    Separable space

    A topological space \(X\) is separable if there exists a countable subset \(Q\subseteq X\) that is dense in \(X\).

    What does it mean for a space to be separable?

    Sequentially compact space

    A topological space \(X\) is sequentially compact if every sequence in \(X\) has a subsequence that converges to a point of \(X\).

    What does it mean for a space to be sequentially compact?

    Second-countable space

    A topological space \(X\) is second-countable if there exists a countable basis for its topology.

    Subspace topology

    Let \((X, \tau)\) be a topological space and \(A\subseteq X\) a subset. The subspace topology on \(A\) is \[ \tau_A\coloneqq\ts{A\cap U \st U\in\tau}. \]

    See [@Mun00].

    Surjection and right inverses

    A map of sets \(\pi\colon A\to B\) is a surjection if for every \(b\in B\) there exists \(a\in A\) with \(\pi(a)=b\). A right inverse of \(\pi\) is a map \(f\colon B\to A\) with \(\pi\circ f=\id_B\).

    Assuming the axiom of choice, a map of sets \(\pi\colon A\to B\) is a surjection if and only if it has a right inverse.

    In a category, a morphism with a right inverse is a split epimorphism. A continuous surjection need not have a continuous right inverse: \(t\mapsto e^{2\pi it}\) from \(\RR\) onto \(S^1\) has none, since a continuous \(f\colon S^1\to\RR\) with \(e^{2\pi if(z)}=z\) would make \(\id_{S^1}\) factor through the contractible space \(\RR\) and so be nullhomotopic.

    Separation axioms

    Let \(X\) be a topological space. Two subsets \(S, T\subseteq X\) are separated by neighborhoods if there exist disjoint open sets \(U\supseteq S\) and \(V\supseteq T\).

    • \(X\) is \(T_0\) if for all \(x\neq y\) in \(X\) there exists an open set containing exactly one of \(x\) and \(y\).

    • \(X\) is \(T_1\) if for all \(x\neq y\) in \(X\) there exists an open set containing \(x\) and not \(y\).

    • \(X\) is \(T_2\), or Hausdorff, if for all \(x\neq y\) in \(X\) the sets \(\ts{x}\) and \(\ts{y}\) are separated by neighborhoods.

    • \(X\) is \(T_{2\frac12}\) if for all \(x\neq y\) in \(X\) there exist open sets \(U\ni x\) and \(V\ni y\) with \(\cl_X(U)\intersect\cl_X(V) = \emptyset\).

    • \(X\) is \(T_3\) if it is \(T_0\) and regular: for every \(x\in X\) and every closed \(F\subseteq X\) with \(x\notin F\), the sets \(\ts{x}\) and \(F\) are separated by neighborhoods.

    • \(X\) is \(T_{3\frac12}\) if it is \(T_0\) and completely regular: for every \(x\in X\) and every closed \(F\subseteq X\) with \(x\notin F\) there exists a continuous \(f\colon X\to[0,1]\) with \(f(x) = 0\) and \(f(F)\subseteq\ts{1}\).

    • \(X\) is \(T_4\) if it is \(T_1\) and normal: any two disjoint closed subsets of \(X\) are separated by neighborhoods.

    A space is \(T_1\) if and only if every singleton is closed.

    [@Mun00].

    (Counterexamples for separation axioms)

    • Not \(T_0\): the space \(\ts{ f\colon\RR\to \CC\st \int_\RR \abs{f}^2 < \infty }\) with the topology of the seminorm \(\norm{f}_2\), since two functions that agree almost everywhere have the same neighborhoods.

    • \(T_0\) but not \(T_1\): \(\spec R\) with the Zariski topology, for a commutative ring \(R\) with a prime ideal that is not maximal. The closure of a point \(\mathfrak p\) is \(V(\mathfrak p)\), so the points of \(\spec R \sm \mspec R\) are not closed.

    Topology via open sets and closed sets

    Let \(X\) be a set. A topology on \(X\) is a collection \(\tau\) of subsets of \(X\) such that

    • \(\emptyset \in \tau\) and \(X\in \tau\);

    • \(\Union_{i\in I} U_i \in \tau\) for every family \((U_i)_{i\in I}\) of elements of \(\tau\);

    • \(U_1\intersect \cdots \intersect U_n \in \tau\) for all \(n\geq 1\) and \(U_1, \ldots, U_n\in\tau\).

    The pair \((X, \tau)\) is a topological space, the elements of \(\tau\) are its open sets, and a subset \(F\subseteq X\) is closed if \(X\sm F\) is open.

    By De Morgan’s laws, a collection \(\mathcal F\) of subsets of \(X\) is the collection of closed sets of a topology on \(X\) if and only if \(\emptyset, X\in\mathcal F\), \(\mathcal F\) is closed under arbitrary intersections, and \(\mathcal F\) is closed under finite unions; the topology is then \(\ts{X\sm F \st F\in\mathcal F}\).

    An infinite intersection of open sets need not be open: in \(\RR\), \(\bigcap_{n\geq 1} (-1/n, 1/n) = \ts{0}\), which is closed and not open.

    Separation axioms $T_0$ through $T_4$

    Let \(X\) be a topological space. Two subsets \(S, T\subseteq X\) are separated by neighborhoods if there exist disjoint open sets \(U\supseteq S\) and \(V\supseteq T\).

    • \(X\) is \(T_0\) if for all \(x_1\neq x_2\) in \(X\) some open set contains exactly one of \(x_1,x_2\).

    • \(X\) is \(T_1\) if for all \(x_1\neq x_2\) in \(X\) there is an open set containing \(x_1\) but not \(x_2\), and an open set containing \(x_2\) but not \(x_1\).

    • \(X\) is \(T_2\) if any two distinct points are separated by neighborhoods; this is the Hausdorff condition.

    • \(X\) is \(T_{2.5}\) if any two distinct points \(x_1,x_2\) have open neighborhoods \(U_1\ni x_1\) and \(U_2\ni x_2\) with \(\overline{U_1}\cap\overline{U_2}=\emptyset\).

    • \(X\) is \(T_3\) if it is \(T_0\) and regular: for every \(x\in X\) and every closed \(F\subseteq X\) with \(x\notin F\), the sets \(\ts{x}\) and \(F\) are separated by neighborhoods.

    • \(X\) is \(T_{3.5}\) if it is \(T_0\) and completely regular: for every \(x\in X\) and every closed \(F\subseteq X\) with \(x\notin F\) there is a continuous \(f\colon X\to[0,1]\) with \(f(x)=0\) and \(f(F)\subseteq\ts{1}\).

    • \(X\) is \(T_4\) if it is \(T_1\) and normal: any two disjoint closed subsets of \(X\) are separated by neighborhoods.

    A space is \(T_1\) if and only if every one-point subset is closed.

    A space that is not \(T_0\): let \(\mathcal L^2(\RR)\) be the set of measurable \(f\colon\RR\to\CC\) with \(\int_\RR\abs{f}^2<\infty\), topologized by the pseudometric \(d(f,g)=\qty{\int_\RR\abs{f-g}^2}^{1/2}\). If \(f\neq g\) agree almost everywhere, then \(d(f,g)=0\), so every open set containing one of them contains the other.

    A space that is \(T_0\) but not \(T_1\): for a commutative ring \(R\) with a prime ideal that is not maximal, \(\Spec R\) with the Zariski topology. The closure of the point \(\mathfrak p\) is \(V(\mathfrak p)=\ts{\mathfrak q\in\Spec R \st \mathfrak q\supseteq\mathfrak p}\), so a prime that is not maximal is not a closed point; distinct primes have distinct closures, which gives \(T_0\).

    Topological embedding

    Let \(X\) and \(Y\) be topological spaces. A continuous map \(f\colon X\to Y\) is a topological embedding if it is injective and the map \(X\to f(X)\), \(x\mapsto f(x)\), is a homeomorphism, where \(f(X)\) carries the subspace topology.

    Uniformly continuous map

    Let \((X, d_X)\) and \((Y, d_Y)\) be metric spaces. A map \(f\colon X\to Y\) is uniformly continuous if for every \(\varepsilon > 0\) there exists \(\delta > 0\) such that for all \(x_1, x_2\in X\), \[ d_X(x_1, x_2) < \delta \implies d_Y(f(x_1), f(x_2)) < \varepsilon. \]

    Analysis and metric spaces

    Diameter of a subset of a metric space

    Let \((X, d)\) be a metric space and \(A\subseteq X\) a nonempty subset. The diameter of \(A\) is \[ \diam(A) \coloneqq \sup_{p, q\in A}d(p, q) \in [0, \infty] .\]

    Distance from a point to a subset

    Let \((X, d)\) be a metric space, \(S\subseteq X\) a nonempty subset, and \(x\in X\). The distance from \(x\) to \(S\) is \[ \dist(x, S)\coloneqq\inf_{s\in S} d(x, s). \]

    Lebesgue number of an open cover

    Let \((X,d)\) be a metric space and \(\mathcal U\) an open cover of \(X\). A real number \(\delta>0\) is a Lebesgue number for \(\mathcal U\) if every subset \(A\subseteq X\) with diameter \(\diam(A)<\delta\) is contained in some \(U\in\mathcal U\).

    (Lebesgue number lemma)

    Every open cover of a compact metric space has a Lebesgue number [@Mun00].

    Totally bounded metric space

    A metric space \((M, d)\) is totally bounded if for every \(\eps > 0\) there are finitely many open balls of radius \(\eps\) whose union is \(M\) [@Mun00].

    What does it mean for a metric space to be totally bounded?

    Algebraic topology

    Alexander duality

    The reduced homology of a complement in $S^n$ is the shifted reduced cohomology of the subset.Suggest a slogan

    Let \(n\geq 1\) and let \(K\subsetneq S^n\) be a nonempty, compact, locally contractible subspace. Then for all \(i\), \[ \tilde H_i(S^n \sm K; \ZZ) \cong \tilde H^{n-i-1}(K;\ZZ) .\]

    It follows, with \(M = S^n\) and the long exact sequence of the pair \((S^n, S^n\sm K)\), from the duality \(H_i(M, M\sm K;\ZZ)\cong H^{n-i}(K;\ZZ)\) for a compact, locally contractible subspace \(K\) of a closed, orientable \(n\)-manifold \(M\) [@Hat02].

    [@Hat02].

    Basis of a module

    Let \(R\) be a ring and \(M\) a left $R$-module. A subset \(B\subseteq M\) is a basis of \(M\) if \(B\) is linearly independent and a generating set for \(M\).

    An \(R\)-module is free exactly when it has a basis.

    Boundary of a manifold

    Let \(n\geq 1\) and let \(M\) be a topological $n$-manifold with boundary. The boundary of \(M\) is the subspace \[ \del M \coloneqq \theset{x\in M \st H_n(M, M\setminus\theset{x}; \ZZ) = 0} \] [@Hat02].

    A point \(x\in M\) lies in \(\del M\) if and only if a neighborhood of \(x\) is carried homeomorphically onto an open subset of \(\theset{y\in\RR^n \st y_n\geq 0}\) with \(x\) going to a point with \(y_n = 0\); for every other point, \(H_n(M, M\setminus\theset{x};\ZZ)\cong\ZZ\). The boundary \(\del M\) is an \((n-1)\)-manifold with empty boundary [@Hat02].

    Cap product

    Let \(X\) be a topological space, \(R\) a commutative ring, and \(0\leq q\leq p\). For a singular \(p\)-simplex \(\sigma\colon\Delta^p=[v_0,\ldots,v_p]\to X\) and \(0\leq i\leq j\leq p\), write \(\sigma|_{[v_i,\ldots,v_j]}\) for the singular \((j-i)\)-simplex obtained by restricting \(\sigma\) to the face \([v_i,\ldots,v_j]\), identified with \(\Delta^{j-i}\) preserving the order of the vertices. The cap product on chains and cochains is the \(R\)-bilinear map \[ \frown\colon C_p(X;R)\times C^q(X;R)\to C_{p-q}(X;R),\qquad \sigma\frown\psi\coloneqq\psi\qty{\sigma|_{[v_0,\ldots,v_q]}}\,\sigma|_{[v_q,\ldots,v_p]} \] on singular \(p\)-simplices \(\sigma\) and singular $q$-cochains \(\psi\), extended linearly in \(\sigma\) [@Hat02].

    For \(\sigma\in C_p(X;R)\) and \(\psi\in C^q(X;R)\), \[ \del(\sigma\frown\psi)=(-1)^q\qty{\del\sigma\frown\psi-\sigma\frown\delta\psi}, \] so the cap product of a cycle and a cocycle is a cycle, and it induces an \(R\)-bilinear map \[ \frown\colon H_p(X;R)\times H^q(X;R)\to H_{p-q}(X;R) \] [@Hat02].

    Cellular homology

    Let \(X\) be a CW complex with \(n\)-skeleta \(X^n\), and \(X^{-1}\coloneqq\emptyset\). The cellular chain complex of \(X\) has \(C_n^{\cell}(X) \coloneqq H_n(X^n, X^{n-1})\) and differential \(d_n\colon C_n^{\cell}(X)\to C_{n-1}^{\cell}(X)\) the composite \[ H_n(X^n, X^{n-1}) \mapsvia{\del_n} H_{n-1}(X^{n-1}) \mapsvia{j_{n-1}} H_{n-1}(X^{n-1}, X^{n-2}) \] of the boundary map of the long exact sequence of the pair \((X^n, X^{n-1})\) and the map induced by inclusion. The cellular homology \(H_n^{\cell}(X)\) is the homology of \((C_*^{\cell}(X), d_*)\).

    The group \(C_n^{\cell}(X)\) is free abelian with a basis in bijection with the \(n\)-cells of \(X\), and \(d_{n-1}\circ d_n = 0\). By the cellular boundary formula, for \(n\geq 2\) the coefficient of the \((n-1)\)-cell \(e^{n-1}_\beta\) in \(d_n(e^n_\alpha)\) is the degree of the composite \(S^{n-1}_\alpha \to X^{n-1}\to X^{n-1}/\qty{X^{n-1}\sm e^{n-1}_\beta} \cong S^{n-1}_\beta\) of the attaching map of \(e^n_\alpha\) and the quotient map. There is an isomorphism \(H_n^{\cell}(X)\cong H_n(X)\) for every \(n\) [@Hat02].

    [@Hat02].

    Cellular map

    Let \(X\) and \(Y\) be CW complexes with \(n\)-skeleta \(X^n\) and \(Y^n\). A continuous map \(f\colon X\to Y\) is cellular if \(f(X^n) \subseteq Y^n\) for every \(n\geq 0\).

    Closed manifold

    A closed manifold is a compact manifold without boundary [@Hat02].

    Closed surface

    A closed surface is a compact \(2\)-dimensional manifold without boundary.

    What is a closed surface?

    Constant map

    Let \(X\) and \(Y\) be sets. A map \(f\colon X\to Y\) is constant if there exists \(y_0\in Y\) such that \(f(x) = y_0\) for every \(x\in X\).

    Colimit of a directed system

    Let \(\mathcal C\) be a category and \((I,\leq)\) a directed set. A directed system in \(\mathcal C\) consists of objects \(X_i\) for \(i\in I\) and morphisms \(f_{ij}\colon X_i\to X_j\) for \(i\leq j\), with \(f_{ii}=\id_{X_i}\) and \(f_{jk}\circ f_{ij}=f_{ik}\) for \(i\leq j\leq k\). A colimit of the system is an object \(X\) with morphisms \(\iota_i\colon X_i\to X\) satisfying \(\iota_j\circ f_{ij}=\iota_i\) for \(i\leq j\), such that for every object \(Y\) and morphisms \(\psi_i\colon X_i\to Y\) with \(\psi_j\circ f_{ij}=\psi_i\) for \(i\leq j\) there is a unique morphism \(h\colon X\to Y\) with \(h\circ\iota_i=\psi_i\) for all \(i\).

    The morphisms \(\iota_i\) go from the system to \(X\); reversing every morphism gives the limit, whose structure morphisms go from the limit to the system. This is the special case of the colimit of a diagram in which the indexing category is the directed set \(I\).

    In sets, and for abelian groups with the group structure of the direct limit, the colimit is \[ \varinjlim_iX_i=\qty{\Disjoint_{i\in I}X_i}/\sim, \] where \(\sim\) is the equivalence relation generated by \(x\sim f_{ij}(x)\) for \(x\in X_i\) and \(i\leq j\), with \(\iota_i\) sending \(x\) to its class [@Hat02].

    (Limits and colimits)

    • Coproducts, pushouts, and direct limits are colimits; for example, \(\ZZ[1/p]\) is the direct limit of \(\ZZ\xrightarrow{\times p}\ZZ\xrightarrow{\times p}\cdots\).

    • Products, pullbacks, and inverse limits are limits; for example, the \(p\)-adic integers \(\ZZ_{p}\) are the inverse limit of \(\cdots\to\ZZ/p^2\to\ZZ/p\).

    Colimit

    Let \(\mathcal C\) be a category, \(I\) a small category, and \(F\colon I\to\mathcal C\) a functor. A cocone on \(F\) is an object \(Y\) of \(\mathcal C\) with morphisms \(\psi_i\colon F(i)\to Y\) for \(i\in I\) such that \(\psi_j\circ F(u) = \psi_i\) for every morphism \(u\colon i\to j\) in \(I\). A colimit of \(F\) is a cocone \((X, (\iota_i)_{i\in I})\) such that for every cocone \((Y, (\psi_i)_{i\in I})\) there exists a unique morphism \(\varphi\colon X\to Y\) with \(\varphi\circ\iota_i = \psi_i\) for all \(i\in I\).

    A colimit is unique up to unique isomorphism compatible with the maps \(\iota_i\), and is written \(\colim_I F\). Reversing all arrows gives the dual notion, a limit.

    The following constructions are colimits.

    • A coproduct \(\Disjoint_{i\in I} X_i\) is the colimit of a functor on a category \(I\) with no non-identity morphisms.

    • The pushout of \(X \leftarrow Z \to Y\) is the colimit of that diagram.

    • The direct limit of a directed system \((A_\alpha, f_{\alpha\beta})\) is its colimit.

    • In abelian groups, \(\ZZ[1/p] \cong \colim\qty{\ZZ \mapsvia{p} \ZZ \mapsvia{p} \ZZ \mapsvia{p} \cdots}\), with the \(k\)th copy of \(\ZZ\), for \(k\geq 0\), mapped to \(\ZZ[1/p]\) by \(1\mapsto p^{-k}\).

    [@Wei94].

    Contractible space

    A topological space \(X\) is contractible if \(\id_X\) is nullhomotopic: there is a point \(x_0\in X\) such that \(\id_X\) is homotopic to the constant map \(c\colon X\to X\), \(c(x) = x_0\).

    A topological space \(X\) is contractible if and only if \(X\) is homotopy equivalent to a one-point space \(\ts{x_0}\), that is, there are continuous maps \(f\colon X\to\ts{x_0}\) and \(g\colon\ts{x_0}\to X\) with \(f\circ g = \id_{\ts{x_0}}\) and \(g\circ f\simeq\id_X\).

    See [@Hat02].

    Coproduct

    Let \(\mathcal C\) be a category and \((X_\alpha)_{\alpha\in A}\) a family of objects of \(\mathcal C\). A coproduct of \((X_\alpha)_{\alpha\in A}\) is an object \(\coprod_{\alpha\in A} X_\alpha\) together with morphisms \(\iota_\beta\colon X_\beta\to\coprod_{\alpha\in A} X_\alpha\) for \(\beta\in A\) such that for every object \(Y\) and every family of morphisms \(f_\beta\colon X_\beta\to Y\) there is a unique morphism \(f\colon\coprod_{\alpha\in A} X_\alpha\to Y\) with \(f\circ\iota_\beta = f_\beta\) for every \(\beta\in A\).

    The coproduct in \(\mathcal C\) is the product in the opposite category \(\mathcal C^{\mathrm{op}}\). By the uniqueness in the definition, two coproducts of the same family are isomorphic by a unique isomorphism compatible with the morphisms \(\iota_\beta\).

    In \(\Top\) the coproduct is the disjoint union \(\coprod_\alpha X_\alpha\), a subset \(U\) being open if \(U\cap X_\alpha\) is open in \(X_\alpha\) for every \(\alpha\). In the category of based spaces it is the wedge sum \(\bigvee_\alpha X_\alpha\). In \(\Ab\) it is the direct sum \(\bigoplus_\alpha X_\alpha\), and in \(\Grp\) it is the free product \(\ast_\alpha X_\alpha\).

    See [@Hat02].

    Covering space

    Let \(X\) be a topological space. A covering space of \(X\) is a topological space \(\tilde X\) together with a continuous map \(p\colon \tilde X\to X\) such that every \(x\in X\) has an open neighborhood \(U\) for which \(\inverseof{p}(U)\) is a union of pairwise disjoint open sets \(V_i\subseteq\tilde X\), each mapped homeomorphically onto \(U\) by \(p\). The sets \(V_i\) are the sheets of \(\tilde X\) over \(U\).

    Let \(p_1\colon\tilde X_1\to X\) and \(p_2\colon\tilde X_2\to X\) be covering spaces of \(X\). An isomorphism of covering spaces is a homeomorphism \(f\colon\tilde X_1\to\tilde X_2\) with \(p_2\circ f = p_1\):

    [@Hat02].

    Cup product

    Let \(X\) be a topological space, \(R\) a ring, and \(p, q\geq 0\). For singular cochains \(\varphi\in C^p(X;R)\) and \(\psi\in C^q(X;R)\), the cup product \(\varphi\smile\psi\in C^{p+q}(X;R)\) is the cochain whose value on a singular \((p+q)\)-simplex \(\sigma\colon[v_0, \ldots, v_{p+q}]\to X\) is \[ (\varphi\smile\psi)(\sigma) \coloneqq \varphi\qty{\ro{\sigma}{[v_0, \ldots, v_p]}}\,\psi\qty{\ro{\sigma}{[v_p, \ldots, v_{p+q}]}} ,\] where each face is identified with the standard simplex of its dimension by the order-preserving linear homeomorphism.

    The coboundary satisfies \(\delta(\varphi\smile\psi) = \delta\varphi\smile\psi + (-1)^p\varphi\smile\delta\psi\), so the cup product of two cocycles is a cocycle and the cup product of a cocycle with a coboundary is a coboundary. Hence it induces a bilinear map \[ \smile\colon H^p(X;R)\cross H^q(X;R)\to H^{p+q}(X;R), \qquad [\varphi]\smile[\psi]\coloneqq[\varphi\smile\psi] .\]

    [@Hat02].

    (Applications of the cup product)

    On a closed oriented manifold, the cup product of the Poincaré duals of transversely intersecting closed oriented submanifolds is the Poincaré dual of their intersection. \(T^2\) and \(S^2 \vee S^1 \vee S^1\) have isomorphic cohomology groups, and they are not homotopy equivalent: the cup product of the two degree-\(1\) generators is nonzero for \(T^2\) and zero for the wedge.

    CW complex

    A CW complex is a topological space \(X\) together with subspaces \(X^0\subseteq X^1\subseteq\cdots\) constructed as follows.

    1. \(X^0\) is a discrete space; its points are the \(0\)-cells.

    2. For \(n\geq 1\), the \(n\)-skeleton \(X^n\) is obtained from \(X^{n-1}\) by attaching a family of \(n\)-cells \(e^n_\alpha\) along continuous maps \(\varphi_\alpha\colon S^{n-1}\to X^{n-1}\): \[ X^n = \qty{X^{n-1} \disjoint \Disjoint_\alpha D^n_\alpha} / \qty{x \sim \varphi_\alpha(x) \text{ for } x \in \del D^n_\alpha}, \] with the quotient topology.

    3. Either \(X = X^n\) for some \(n\), or \(X = \Union_{n\geq 0} X^n\) with the weak topology: a subset \(A\subseteq X\) is closed if and only if \(A \intersect X^n\) is closed in \(X^n\) for every \(n\).

    [@Hat02].

    Attaching an $n$-cell

    Let \(X\) be a topological space and \(n\geq 0\). Let \(D^n\subseteq\RR^n\) be the closed unit ball and \(S^{n-1}=\del D^n\) its boundary sphere, with \(S^{-1}=\emptyset\). For a continuous map \(f\colon S^{n-1}\to X\), the space obtained by attaching an \(n\)-cell to \(X\) along \(f\) is the quotient space \[ X\cup_f D^n \coloneqq \qty{X\disjoint D^n}/\qty{s\sim f(s) \text{ for } s\in S^{n-1}}, \] the case \(X_0=X\), \(X_1=D^n\), \(A=S^{n-1}\) of attaching a space along a map. The map \(f\) is the attaching map. The composite \(\Phi\colon D^n\to X\disjoint D^n\to X\cup_f D^n\) of the inclusion and the quotient map is the characteristic map, its image \(\Phi(D^n)\) is the closed \(n\)-cell, and \(e^n\coloneqq\Phi(D^n\setminus S^{n-1})\) is the open \(n\)-cell.

    In low dimensions:

    • \(n=0\): \(D^0\) is a point and \(S^{-1}=\emptyset\), so \(X\cup_f D^0=X\disjoint\ts{\pt}\).

    • \(n=1\): \(D^1=[-1,1]\subseteq\RR\) and \(f\) is a map \(S^0=\ts{-1,1}\to X\); the closed \(1\)-cell is the image of a path from \(f(-1)\) to \(f(1)\).

    • \(n=2\): \(D^2\subseteq\RR^2\) is the closed disk, attached along a map \(S^1\to X\).

    • \(n=3\): \(D^3\subseteq\RR^3\) is the closed ball, attached along a map \(S^2\to X\).

    Deck transformation

    Let \(p\colon\tilde X\to X\) be a covering space. A deck transformation of \(p\) is a homeomorphism \(\psi\colon\tilde X\to\tilde X\) with \(p\circ\psi=p\). The deck transformations form a group \(\Deck(\tilde X\to X)\) under composition.

    In the slice category \(\mathsf{Top}_{/X}\), whose objects are continuous maps to \(X\) and whose morphisms are maps over \(X\), the group \(\Deck(\tilde X\to X)\) is the automorphism group \(\Aut_{\mathsf{Top}_{/X}}(p)\) of the object \(p\colon\tilde X\to X\). A deck transformation permutes each fiber \(\inverseof{p}(x)\). If \(\tilde X\) is connected, a deck transformation is determined by the image of one point, so a deck transformation with a fixed point is the identity [@Hat02].

    Deformation and deformation retraction

    Let \(X\) be a topological space and \(I=[0,1]\). A deformation of \(X\) is a family of maps \(f_t\colon X\to X\), \(t\in I\), such that \(f_0=\id_X\) and \((x,t)\mapsto f_t(x)\) is a continuous map \(X\times I\to X\). For a subspace \(A\subseteq X\), a deformation retraction of \(X\) onto \(A\) is a deformation \((f_t)_{t\in I}\) of \(X\) with \(f_1(X)=A\) and \(f_t(a)=a\) for all \(a\in A\) and \(t\in I\) [@Hat02].

    If \(X\) deformation retracts onto \(A\), then the inclusion \(A\injects X\) is a homotopy equivalence [@Hat02].

    Deformation retraction

    Let \(X\) be a topological space, \(A\subseteq X\) a subspace, and \(\iota\colon A\injects X\) the inclusion. A deformation retraction of \(X\) onto \(A\) is a homotopy \(F\colon X\cross I\to X\), written \(F_t \coloneqq F(\wait, t)\), such that for all \(t\in I\): \[ F_0 = \id_X, \qquad F_1(X) \subseteq A, \qquad F_t(a) = a \text{ for all } a\in A .\] The subspace \(A\) is a deformation retract of \(X\) if there exists a deformation retraction of \(X\) onto \(A\).

    If \(F\) is a deformation retraction of \(X\) onto \(A\), then \(r\coloneqq F_1\colon X\to A\) is a retraction, \(r\circ\iota = \id_A\), and \(F\) is a homotopy from \(\id_X\) to \(\iota\circ r\). Hence \(\iota\) is a homotopy equivalence with homotopy inverse \(r\).

    Two spaces \(X\) and \(Y\) are homotopy equivalent if and only if there exists a space \(Z\) containing both \(X\) and \(Y\) as deformation retracts; for a homotopy equivalence \(f\colon X\to Y\), the mapping cylinder of \(f\) is such a \(Z\).

    [@Hat02].

    What does it mean for $A \subset X$ to be a deformation retract of $X$?
    What map exhibits $A \subset X$ as a deformation retract of $X$?

    If spaces \(X\) and \(Y\) both deformation retract onto a common subspace \(A\), then \(X \homotopic A \homotopic Y\).

    Degree of a map of spheres

    Let \(n\geq 1\) and let \(f\colon S^n \to S^n\) be a continuous map. Since \(H^n(S^n;\ZZ) \cong \ZZ\), the induced homomorphism \(f^*\colon H^n(S^n;\ZZ) \to H^n(S^n;\ZZ)\) is multiplication by a unique integer \(d\), that is, \(f^*(u) = d\,u\) for every \(u\in H^n(S^n;\ZZ)\). The degree of \(f\) is \(\deg f \coloneqq d\).

    Let \(n\geq 1\) and \(f, g\colon S^n\to S^n\) be continuous.

    1. If \(f\) and \(g\) are homotopic, then \(\deg f = \deg g\).

    2. \(\deg f\) equals the integer \(d'\) with \(f_*(v) = d'\,v\) for every \(v\in H_n(S^n;\ZZ)\).

    [@Hat02].

    Dimension of a manifold

    Let \(M\) be an \(n\)-manifold and \(x\in M\). Then \(H_i(M, M\sm\ts{x};\ZZ)\cong\ZZ\) for \(i = n\) and \(H_i(M, M\sm\ts{x};\ZZ) = 0\) for \(i\neq n\). In particular, a nonempty space that is both an \(m\)-manifold and an \(n\)-manifold has \(m = n\).

    Let \(M\) be a nonempty space that is an \(n\)-manifold for some \(n\geq 0\). The dimension of \(M\) is this integer \(n\), equivalently the unique \(i\) with \(H_i(M, M\sm\ts{x};\ZZ)\neq 0\) for any \(x\in M\).

    See [@Hat02].

    Euler characteristic of a finite CW complex

    Let \(X\) be a finite CW complex with \(c_n\) cells of dimension \(n\). The Euler characteristic of \(X\) is \[ \chi(X)\coloneqq\sum_{n\geq0}(-1)^nc_n. \]

    For a finite CW complex \(X\), \[ \chi(X)=\sum_{n\geq0}(-1)^n\rank H_n(X;\ZZ), \] where \(H_n(X;\ZZ)\) is singular homology [@Hat02]. Consequently \(\chi(X)\) does not depend on the CW structure, and finite CW complexes that are homotopy equivalent have the same Euler characteristic.

    Left exact, right exact, and exact functors

    Let \(T\colon\mathcal A\to\mathcal B\) be an additive covariant functor between abelian categories, such as categories of modules. The functor \(T\) is right exact if for every short exact sequence \(0\to A\mapsvia{f}B\mapsvia{g}C\to0\) in \(\mathcal A\) the sequence \[ T(A)\mapsvia{T(f)}T(B)\mapsvia{T(g)}T(C)\to0 \] is exact, and left exact if for every such sequence \[ 0\to T(A)\mapsvia{T(f)}T(B)\mapsvia{T(g)}T(C) \] is exact. It is exact if it is both left and right exact, that is, if it takes every short exact sequence to a short exact sequence.

    For a commutative ring \(R\) and an \(R\)-module \(N\), the functor \(-\otimes_RN\) on \(R\)-modules is right exact [@DF04], and so is \(N\otimes_R-\).

    (A right exact functor)

    For a commutative ring \(R\) and an \(R\)-module \(M\), \(\wait \tensor_{R} M\) is right exact, and it is exact if and only if \(M\) is flat.

    Excision axiom

    Let \(h_*\) be a sequence of functors \(h_n\) from pairs of topological spaces to abelian groups. The functors \(h_*\) satisfy excision if for every pair \((X, A)\) and every subspace \(Z\subseteq A\) with \(\cl_X(Z) \subseteq A^\circ\), the inclusion \((X\sm Z, A\sm Z)\injects (X, A)\) induces isomorphisms \[ h_n(X\sm Z, A\sm Z) \mapsvia{\sim} h_n(X, A) \quad\text{for all } n .\]

    Excision is one of the Eilenberg–Steenrod axioms for a homology theory. Singular homology satisfies it [@Hat02]; relative homotopy groups do not.

    [@Hat02].

    Free and properly discontinuous group actions

    Let \(G\) be a group acting on a topological space \(X\) by homeomorphisms, \(G\actson X\).

    • The action is free if \(g(x) = x\) for some \(x\in X\) implies \(g = e\).

    • The action is properly discontinuous if every \(x\in X\) has a neighborhood \(U\) such that for all \(g_1, g_2\in G\), \(g_1(U) \intersect g_2(U) \neq \emptyset\) implies \(g_1 = g_2\).

    A properly discontinuous action in this sense is free. Other sources call an action properly discontinuous under the weaker condition that every \(x\in X\) has a neighborhood \(U\) with \(U \intersect g(U) \neq \emptyset\) for only finitely many \(g\in G\). An action satisfying the condition in the definition is called a covering space action in [@Hat02].

    Free product of groups

    Let \((G_\alpha)_{\alpha\in A}\) be a family of groups. A reduced word is a finite sequence \(g_1g_2\cdots g_k\), \(k\geq 0\), in which each letter \(g_i\) lies in \(G_{\alpha_i}\sm\ts{1}\) for some \(\alpha_i\in A\) and \(\alpha_i\neq\alpha_{i+1}\) for \(1\leq i<k\). The free product \(\ast_{\alpha\in A} G_\alpha\) is the set of reduced words, with product given by concatenating two words and then reducing: adjacent letters from the same \(G_\alpha\) are replaced by their product in \(G_\alpha\), and letters equal to \(1\) are deleted, until the word is reduced.

    The free product \(\ast_\alpha G_\alpha\) is a group whose identity is the empty word, and each \(G_\beta\) is a subgroup via one-letter words. For every group \(H\) and homomorphisms \(\varphi_\alpha\colon G_\alpha\to H\) there is a unique homomorphism \(\varphi\colon\ast_\alpha G_\alpha\to H\) restricting to \(\varphi_\alpha\) on each \(G_\alpha\), so \(\ast_\alpha G_\alpha\) is the coproduct of the \(G_\alpha\) in \(\Grp\).

    For a set \(S\), the free group on \(S\) is isomorphic to \(\ast_{s\in S}\ZZ\).

    See [@Hat02].

    Free product with amalgamation

    Let \(A\), \(B\), and \(C\) be groups, and let \(i\colon C\to A\) and \(j\colon C\to B\) be homomorphisms. The free product with amalgamation of \(A\) and \(B\) over \(C\) is \[ A \ast_C B \coloneqq \qty{A \ast B} / N, \] where \(A\ast B\) is the free product and \(N\) is the normal subgroup of \(A\ast B\) generated by \(\ts{ i(c) \inverseof{j(c)} \st c\in C }\).

    Together with the homomorphisms \(A\to A\ast_C B\) and \(B\to A\ast_C B\) induced by the inclusions of \(A\) and \(B\) into \(A\ast B\), the group \(A\ast_C B\) is the pushout of \(A\xleftarrow{i} C\xrightarrow{j} B\) in \(\Grp\).

    By van Kampen's theorem, if \(X = U_1\union U_2\) with \(U_1\), \(U_2\) and \(U_1\intersect U_2\) open and path-connected, then \(\pi_1(X)\) is the free product of \(\pi_1(U_1)\) and \(\pi_1(U_2)\) amalgamated over \(\pi_1(U_1\intersect U_2)\) [@Hat02].

    [@Hat02].

    Fundamental class

    Let \(M\) be a closed connected \(n\)-manifold with an orientation \(x\mapsto\mu_x\in H_n(M,M\setminus\ts{x};\ZZ)\). A fundamental class of \(M\) is a class \([M]\in H_n(M;\ZZ)\) whose image under \(H_n(M;\ZZ)\to H_n(M,M\setminus\ts{x};\ZZ)\) is \(\mu_x\) for every \(x\in M\).

    Let \(M\) be a closed connected \(n\)-manifold with an orientation \(x\mapsto\mu_x\). Then \(H_n(M;\ZZ)\cong\ZZ\), the fundamental class \([M]\) exists and is unique, and it generates \(H_n(M;\ZZ)\) [@Hat02].

    Fundamental group

    Let \(X\) be a topological space and \(x_0\in X\), and let \(I = [0,1]\). The fundamental group \(\pi_1(X, x_0)\) is the set of classes of loops \(\gamma\colon I\to X\) based at \(x_0\) under homotopy rel endpoints, with multiplication \[ [\gamma][\eta]\coloneqq[\gamma\cdot\eta], \] where \(\gamma\cdot\eta\) is the concatenation of \(\gamma\) and \(\eta\).

    The multiplication on \(\pi_1(X, x_0)\) is well defined and makes \(\pi_1(X, x_0)\) a group. Its identity is the class of the constant loop at \(x_0\), and \(\inverseof{[\gamma]} = [\bar\gamma]\), where \(\bar\gamma(s)\coloneqq\gamma(1-s)\).

    Let \(h\colon I\to X\) be a path from \(x_0\) to \(x_1\), with reverse \(\bar h\). Then \(\beta_h\colon\pi_1(X, x_1)\to\pi_1(X, x_0)\), \(\beta_h[\gamma]\coloneqq[h\cdot\gamma\cdot\bar h]\), is a group isomorphism. It depends only on the homotopy class of \(h\) rel endpoints, and for a second path \(h'\) from \(x_0\) to \(x_1\), \(\beta_{h'} = c\circ\beta_h\), where \(c\) is conjugation by \([h'\cdot\bar h]\in\pi_1(X, x_0)\). In particular, if \(X\) is path connected, the groups \(\pi_1(X, x)\) for \(x\in X\) are all isomorphic.

    See [@Hat02].

    Attaching a space along a map

    Let \(X_0\) and \(X_1\) be topological spaces, \(A\subseteq X_1\) a subspace, and \(f\colon A\to X_0\) a continuous map. The space obtained by attaching \(X_1\) to \(X_0\) along \(f\) is the quotient space \[ X_0\sqcup_f X_1\coloneqq\qty{X_0\sqcup X_1}/\qty{a\sim f(a) \text{ for } a\in A}, \] where \(\sim\) is the equivalence relation generated by \(a\sim f(a)\).

    Attaching an \(n\)-cell to \(X_0\) along \(\varphi\colon S^{n-1}\to X_0\) is the case \(X_1 = D^n\), \(A = S^{n-1}\), \(f = \varphi\). For a continuous map \(g\colon X\to Y\), the mapping cylinder of \(g\) is the case \(X_0 = Y\), \(X_1 = X\times[0,1]\), \(A = X\times\ts{1}\), \(f(x, 1) = g(x)\).

    See [@Hat02].

    Homotopic maps

    Let \(X\) and \(Y\) be topological spaces, \(I = [0,1]\), and \(f, g\colon X\to Y\) continuous maps. The maps \(f\) and \(g\) are homotopic, written \(f\homotopic g\), if there is a continuous map \(H\colon X\cross I \to Y\) with \(H(x, 0) = f(x)\) and \(H(x, 1) = g(x)\) for all \(x\in X\); such a map \(H\) is a homotopy from \(f\) to \(g\). Two paths \(\gamma, \eta\colon I\to X\) with \(\gamma(0) = \eta(0)\) and \(\gamma(1) = \eta(1)\) are homotopic rel endpoints if there is a continuous map \(H\colon I\cross I\to X\) with \(H(s, 0) = \gamma(s)\), \(H(s, 1) = \eta(s)\), \(H(0,t) = \gamma(0)\), and \(H(1,t) = \gamma(1)\) for all \(s, t\in I\).

    Homotopy is an equivalence relation on the set of continuous maps \(X\to Y\), and homotopy rel endpoints is an equivalence relation on the set of paths in \(X\) with given endpoints. The set of homotopy classes of continuous maps \(X\to Y\) is denoted \([X,Y]\).

    The fundamental group \(\pi_1(X, x_0)\) is the set of classes of loops at \(x_0\) under homotopy rel endpoints.

    [@Hat02]; [@Mun00].

    Homotopy class

    Let \(X\) and \(Y\) be topological spaces. The homotopy class \([f]\) of a continuous map \(f\colon X\to Y\) is its equivalence class under the relation \(\simeq\) of being homotopic. The set of homotopy classes of continuous maps \(X\to Y\) is \[ [X, Y]\coloneqq\Hom_{\Top}(X, Y)/\simeq. \]

    If \(f, g\colon X\to Y\) are homotopic, then \(f_* = g_*\colon H_n(X)\to H_n(Y)\) for every \(n\).

    For a based space \((Y, y_0)\) and \(n\geq 1\), the homotopy group \(\pi_n(Y, y_0)\) is the set \([(S^n, s_0), (Y, y_0)]\) of homotopy classes of based maps under homotopies fixing the basepoint.

    See [@Hat02].

    Homotopy equivalence

    Let \(X\) and \(Y\) be topological spaces. A continuous map \(f\colon X\to Y\) is a homotopy equivalence if there is a continuous map \(g\colon Y\to X\) with \(f\circ g\simeq\id_Y\) and \(g\circ f\simeq\id_X\), where \(\simeq\) denotes homotopy. Such a map \(g\) is a homotopy inverse of \(f\). The spaces \(X\) and \(Y\) are homotopy equivalent, or have the same homotopy type, written \(X\simeq Y\), if there is a homotopy equivalence \(X\to Y\).

    Every homeomorphism is a homotopy equivalence, and the converse fails: for \(n\geq 1\) the inclusion \(\ts{0}\hookrightarrow\RR^n\) is a homotopy equivalence, with homotopy inverse the constant map \(\RR^n\to\ts{0}\) and homotopy \(H(x, t) = tx\) from the constant map to \(\id_{\RR^n}\), but \(\ts{0}\) and \(\RR^n\) are not homeomorphic.

    Homotopy extension property

    Let \(X\) be a topological space, \(A\subseteq X\) a subspace, and \(I = [0,1]\). The pair \((X, A)\) has the homotopy extension property if for every space \(Y\), every continuous map \(f_0\colon X\to Y\), and every homotopy \(g_t\colon A\to Y\) with \(g_0 = f_0|_A\), there is a homotopy \(f_t\colon X\to Y\) starting at \(f_0\) with \(f_t|_A = g_t\) for all \(t\in I\). Equivalently, every pair of continuous maps \(X\times\ts{0}\to Y\) and \(A\times I\to Y\) that agree on \(A\times\ts{0}\) extends to a continuous map \(X\times I\to Y\).

    If \(A\) is closed in \(X\), then \((X, A)\) has the homotopy extension property if and only if \(X\times\ts{0}\cup A\times I\) is a retract of \(X\times I\).

    Every CW pair \((X, A)\) has the homotopy extension property.

    See [@Hat02].

    Homotopy groups

    Let \(X\) be a topological space, \(x_0\in X\), and \(n\geq 1\), and let \(I^n = [0,1]^n\) with boundary \(\del I^n\). The \(n\)th homotopy group \(\pi_n(X, x_0)\) is the set of homotopy classes of maps \(f\colon(I^n, \del I^n)\to(X, x_0)\), where the homotopies \(f_t\) are required to satisfy \(f_t(\del I^n) = \ts{x_0}\) for all \(t\), with the operation \([f]+[g]\coloneqq[f+g]\) given by \[ (f+g)(s_1, s_2, \ldots, s_n)\coloneqq \begin{cases} f(2s_1, s_2, \ldots, s_n) & 0\leq s_1\leq 1/2, \\ g(2s_1-1, s_2, \ldots, s_n) & 1/2\leq s_1\leq 1. \end{cases} \]

    For \(n\geq 1\) the operation on \(\pi_n(X, x_0)\) is well defined and makes \(\pi_n(X, x_0)\) a group, and \(\pi_1(X, x_0)\) is the fundamental group. For \(n\geq 2\) the group \(\pi_n(X, x_0)\) is abelian. Since \(I^n/\del I^n\cong S^n\), the set \(\pi_n(X, x_0)\) is in bijection with the set \([(S^n, s_0), (X, x_0)]\) of based homotopy classes of based maps \((S^n, s_0)\to(X, x_0)\).

    See [@Hat02].

    Homotopy lifting property

    Let \(p\colon E\to B\) be a continuous map and \(Y\) a topological space. The map \(p\) has the homotopy lifting property with respect to \(Y\) if for every homotopy \(g_t\colon Y\to B\), \(t\in[0,1]\), and every continuous \(\tilde g_0\colon Y\to E\) with \(p\circ\tilde g_0 = g_0\), there is a homotopy \(\tilde g_t\colon Y\to E\) starting at \(\tilde g_0\) with \(p\circ\tilde g_t = g_t\) for all \(t\). The map \(p\) is a fibration if it has the homotopy lifting property with respect to every space \(Y\).

    A covering space \(p\colon\tilde X\to X\) has the homotopy lifting property with respect to every space \(Y\), and the lifted homotopy \(\tilde g_t\) starting at a given \(\tilde g_0\) is unique.

    See [@Hat02].

    Intersection product and intersection pairing

    Let \(M\) be a closed connected oriented \(n\)-manifold with fundamental class \([M]\in H_n(M;\ZZ)\), and let \(D\colon H^k(M;\ZZ)\to H_{n-k}(M;\ZZ)\), \(D(\varphi)=[M]\frown\varphi\), be the Poincaré duality isomorphism. The intersection product is the bilinear map \[ H_i(M;\ZZ)\times H_j(M;\ZZ)\to H_{i+j-n}(M;\ZZ),\qquad \alpha\cdot\beta\coloneqq[M]\frown\qty{\inverseof{D}(\alpha)\smile \inverseof{D}(\beta)}, \] using the cup product and the cap product. For \(j=n-i\), composing with \(H_0(M;\ZZ)\cong\ZZ\) gives the intersection pairing \(H_i(M;\ZZ)\times H_{n-i}(M;\ZZ)\to\ZZ\).

    The intersection pairing vanishes when either argument is a torsion class, and the induced pairing \[ H_i(M;\ZZ)/\mathrm{torsion}\times H_{n-i}(M;\ZZ)/\mathrm{torsion}\to\ZZ \] is nondegenerate: its adjoint is an isomorphism onto \(\Hom(H_{n-i}(M;\ZZ)/\mathrm{torsion},\ZZ)\). This is the cup product pairing of [@Hat02] transported by \(D\).

    If \(A,B\subseteq M\) are closed oriented submanifolds of dimensions \(i\) and \(j\) that intersect transversely, with fundamental classes representing \(\alpha\) and \(\beta\), then \(\alpha\cdot\beta\) is represented by the submanifold \(A\cap B\) of dimension \(i+j-n\) with its induced orientation. When \(i+j=n\), the intersection \(A\cap B\) is a finite set of points and \(\alpha\cdot\beta\in\ZZ\) is the number of intersection points counted with signs given by the orientations.

    Intersection form of a manifold

    Let \(M\) be a closed, connected, oriented \(2k\)-manifold with fundamental class \([M]\in H_{2k}(M;\ZZ)\). The intersection form of \(M\) is the bilinear form \[ I\colon H^k(M;\ZZ)\cross H^k(M;\ZZ) \to \ZZ, \qquad I(a, b)\coloneqq \inner{a\smile b}{[M]} ,\] where \(\smile\) is the cup product and \(\inner{\wait}{\wait}\) is the Kronecker pairing.

    Graded commutativity of the cup product gives \(I(a, b) = (-1)^k I(b, a)\), so \(I\) is symmetric when \(k\) is even, that is, when \(\dim M\equiv 0 \pmod 4\), and skew-symmetric when \(k\) is odd. By Poincaré duality, \(I\) vanishes on torsion classes and induces a nondegenerate form on \(H^k(M;\ZZ)/\text{torsion}\) whose Gram matrix has determinant \(\pm1\).

    [@Hat02].

    Kronecker pairing

    Let \(X\) be a topological space, \(R\) a commutative ring, and \(n\geq 0\). For a singular cochain \(\varphi\in C^n(X;R) = \Hom_R(C_n(X;R), R)\) and a singular chain \(\alpha\in C_n(X;R)\), put \(\inner{\varphi}{\alpha} \coloneqq \varphi(\alpha)\). The Kronecker pairing, also called the Kronecker product, is the \(R\)-bilinear map \[ \inner{\wait}{\wait}\colon H^n(X; R) \cross H_n(X; R) \to R, \qquad \inner{[\varphi]}{[\alpha]} \coloneqq \varphi(\alpha), \] for a cocycle \(\varphi\) and a cycle \(\alpha\). Its adjoint is the evaluation homomorphism \[ h\colon H^n(X; R) \to \Hom_R(H_n(X;R), R), \qquad h([\varphi])([\alpha]) \coloneqq \varphi(\alpha). \]

    Let \(X\) be a topological space, \(R\) a commutative ring, and \(n\geq 0\).

    1. For \(\psi\in C^{n-1}(X;R)\) and \(\alpha\in C_n(X;R)\), \(\inner{\delta\psi}{\alpha} = \inner{\psi}{\del\alpha}\); hence the Kronecker pairing is well defined on classes.

    2. For a continuous map \(f\colon X\to Y\), \(\beta\in H^n(Y;R)\), and \(x \in H_n(X;R)\), \(\inner{f^*\beta}{x} = \inner{\beta}{f_*x}\).

    3. If \(R\) is a principal ideal domain, there is a split short exact sequence \[ 0 \to \Ext_R(H_{n-1}(X;R), R) \to H^n(X;R) \xrightarrow{h} \Hom_R(H_n(X;R), R) \to 0, \] so \(h\) is surjective, and \(h\) is an isomorphism if and only if \(\Ext_R(H_{n-1}(X;R), R) = 0\).

    The evaluation homomorphism \(h\) need not be injective. For \(X = \RR P^2\), \(R = \ZZ\), and \(n = 2\), one has \(H_1(\RR P^2;\ZZ)\cong\ZZ/2\) and \(H_2(\RR P^2;\ZZ) = 0\), so \(H^2(\RR P^2;\ZZ)\cong\Ext_\ZZ(\ZZ/2,\ZZ)\cong\ZZ/2\) while \(\Hom_\ZZ(H_2(\RR P^2;\ZZ),\ZZ) = 0\).

    [@Hat02].

    Lefschetz duality

    For manifolds with boundary, cap product exchanges relative cohomology on one boundary piece with homology relative to the other.Suggest a slogan

    Let \(M\) be a compact \(R\dash\)orientable \(n\dash\)manifold whose boundary is decomposed as \(\del M = A \union B\) with \(A, B\) compact \((n-1)\dash\)manifolds meeting in \(\del A = \del B = A \intersect B\). Then cap product with a fundamental class \([M]\in H_n(M, \del M; R)\) gives isomorphisms \[ D_M: H^k(M, A; R) \mapsvia{\sim} H_{n-k}(M, B; R) \] for all \(k\). The cases \(A = \del M\), \(B = \emptyset\) and \(A = \emptyset\), \(B = \del M\) give the isomorphisms \[ H^k(M, \del M; R) \cong H_{n-k}(M; R), \qquad H^k(M; R)\cong H_{n-k}(M, \del M; R) \] [@Hat02].

    For \(\del M = \emptyset\) this is Poincaré duality.

    Lefschetz number

    Let \(X\) be a finite simplicial complex, or a retract of one, and let \(f\colon X\to X\) be a continuous map. The Lefschetz number of \(f\) is \[ \tau(f) \coloneqq \sum_{n\geq 0} (-1)^n \tr\qty{ f_*\colon H_n(X;\QQ) \to H_n(X;\QQ) } .\]

    Since \(\tr\qty{\id_{H_n(X;\QQ)}} = \dim_\QQ H_n(X;\QQ)\), the identity has \(\tau(\id_X) = \chi(X)\), the Euler characteristic. By the Lefschetz fixed point theorem, if \(\tau(f)\neq 0\), then \(f\) has a fixed point.

    [@Hat02].

    Lens space

    Let \(m\geq 2\), \(n \geq 1\) and let \(\ell_1, \ldots, \ell_n\) be integers coprime to \(m\). Let \(\ZZ/m\) act on the unit sphere \(S^{2n-1}\subseteq \CC^n\) with the generator acting by \[ (z_1, \ldots, z_n) \mapsto \qty{ e^{2\pi i \ell_1/m} z_1, \ldots, e^{2\pi i \ell_n /m} z_n } .\] The lens space is the orbit space \[ L_m(\ell_1, \ldots, \ell_n) \coloneqq S^{2n-1}/\qty{\ZZ/m} \] [@Hat02].

    The action is free, so \(S^{2n-1}\to L_m(\ell_1,\ldots,\ell_n)\) is a covering space [@Hat02], and \(L_m(\ell_1,\ldots,\ell_n)\) is a closed \((2n-1)\)-manifold. For \(n\geq 2\) the sphere \(S^{2n-1}\) is simply connected, so it is the universal cover and \(\pi_1(L_m(\ell_1,\ldots,\ell_n)) \cong \ZZ/m\) [@Hat02]. For \(m=2\) the lens space is \(\RP^{2n-1}\).

    Local degree

    Let \(n\geq 1\), let \(f\colon S^n\to S^n\) be a continuous map, and let \(y\in S^n\) be a point whose preimage \(\inverseof{f}(y) = \ts{x_1, \ldots, x_m}\) is finite. Homology is taken with coefficients in \(\ZZ\). Choose an open neighborhood \(V\) of \(y\) and disjoint open neighborhoods \(U_i\) of \(x_i\) with \(f(U_i)\subseteq V\), and fix a generator of \(H_n(S^n)\). Then \(f\) restricts to maps of pairs \(f_i\colon (U_i, U_i\sm\ts{x_i})\to (V, V\sm\ts{y})\), and under the isomorphisms \[ H_n(U_i, U_i\sm\ts{x_i})\cong H_n(S^n, S^n\sm\ts{x_i})\cong H_n(S^n)\cong\ZZ, \qquad H_n(V, V\sm\ts{y})\cong H_n(S^n, S^n\sm\ts{y})\cong H_n(S^n)\cong\ZZ \] given by excision and the long exact sequence of the pair, the induced homomorphism \(f_{i*}\) is multiplication by an integer. This integer is the local degree \(\deg f\vert_{x_i}\) of \(f\) at \(x_i\).

    Let \(n\geq 1\), let \(f\colon S^n\to S^n\) be a continuous map, and let \(y\in S^n\) be a point with finite preimage \(\inverseof{f}(y) = \ts{x_1, \ldots, x_m}\). Then \(\deg f = \sum_{i=1}^m \deg f\vert_{x_i}\), where \(\deg f\) is the degree of \(f\).

    [@Hat02].

    Local orientation and orientation of a manifold

    Let \(M\) be an \(n\)-manifold, and for \(A\subseteq M\) write \(H_n(M\mid A)\coloneqq H_n(M, M\sm A;\ZZ)\). For \(x\in M\), excision gives \(H_n(M\mid x)\cong H_n(\RR^n, \RR^n\sm\ts{0};\ZZ)\cong\ZZ\).

    • A local orientation of \(M\) at \(x\) is a generator \(\mu_x\) of \(H_n(M\mid x)\).

    • An orientation of \(M\) is a function \(x\mapsto\mu_x\) assigning to each \(x\in M\) a local orientation at \(x\), such that each \(x\in M\) has an open neighborhood \(U\subseteq M\) with a homeomorphism \(U\cong\RR^n\) and an open ball \(B\subseteq U\) of finite radius containing \(x\), together with a generator \(\mu_B\) of \(H_n(M\mid B)\cong\ZZ\) whose image under the map \(H_n(M\mid B)\to H_n(M\mid y)\) induced by inclusion is \(\mu_y\) for every \(y\in B\).

    • \(M\) is orientable if an orientation of \(M\) exists.

    Each \(H_n(M\mid x)\) has exactly two generators, so \(M\) has exactly two local orientations at each point. The local orientations form the orientation double cover \(\tilde M = \ts{\mu_x \st x\in M,\ \mu_x \text{ a local orientation of } M \text{ at } x}\), with \(\mu_x\mapsto x\). The manifold \(\tilde M\) is orientable, and a connected manifold \(M\) is orientable if and only if \(\tilde M\) has two components [@Hat02].

    Let \(M\) be a connected \(n\)-manifold with orientation double cover \(p\colon\tilde M\to M\), and let \(x\in M\). The following are equivalent:

    1. \(M\) is orientable.

    2. \(\tilde M\) is disconnected.

    3. \(\tilde M\) has two components, each mapped homeomorphically onto \(M\) by \(p\).

    4. \(p\) has a continuous section \(s\colon M\to\tilde M\), \(p\circ s = \id_M\).

    5. The homomorphism \(\pi_1(M, x)\to\ZZ/2\) given by the action of \(\pi_1(M,x)\) on the two-point fiber \(\inverseof{p}(x)\) is trivial.

    [@Hat02].

    Local homology

    Let \(X\) be a topological space, \(A\subseteq X\) a subset, and \(n\geq 0\). The local homology of \(X\) at \(A\) in degree \(n\) is the relative homology group \[ H_n(X\mid A) \coloneqq H_n(X, X\sm A; \ZZ) .\] For a point \(x\in X\), write \(H_n(X\mid x)\coloneqq H_n(X\mid\ts{x})\).

    [@Hat02].

    Local orientation of a manifold

    Let \(M\) be an \(n\)-manifold and \(x\in M\). Then \(H_n(M, M\sm\ts{x};\ZZ)\cong H_n(\RR^n, \RR^n\sm\ts{0};\ZZ)\cong\ZZ\).

    Let \(M\) be an \(n\)-manifold and \(x\in M\). A local orientation of \(M\) at \(x\) is a choice of generator \(\mu_x\) of the infinite cyclic group \(H_n(M, M\sm\ts{x};\ZZ)\).

    See [@Hat02].

    Manifold

    Let \(n\geq0\). An \(n\)-manifold is a Hausdorff space \(M\) in which every point has an open neighborhood homeomorphic to \(\RR^n\) [@Hat02].

    The definition of a topological manifold in [@Lee13] also requires \(M\) to be second-countable. The open long line is a Hausdorff space in which every point has an open neighborhood homeomorphic to \(\RR\), and it is not second-countable.

    Manifold with boundary

    Let \(n\geq 1\) and let \(\mathbb H^n\coloneqq\ts{x\in\RR^n \st x_n\geq 0}\) be the closed upper half-space, with the subspace topology. A topological \(n\)-manifold with boundary is a Hausdorff space \(M\) in which every point has an open neighborhood homeomorphic to \(\RR^n\) or to \(\mathbb H^n\).

    Mayer--Vietoris sequence

    Let \(X\) be a topological space and \(A, B\subseteq X\) subspaces with \(X = A^\circ\cup B^\circ\). Let \(i\colon A\cap B\hookrightarrow A\), \(j\colon A\cap B\hookrightarrow B\), \(k\colon A\hookrightarrow X\), and \(l\colon B\hookrightarrow X\) be the inclusions, and let \(H_n\) denote singular homology with integer coefficients. Define \[ \Phi\colon H_n(A\cap B)\to H_n(A)\oplus H_n(B),\quad \Phi(x) = (i_*x, -j_*x), \qquad \Psi\colon H_n(A)\oplus H_n(B)\to H_n(X),\quad \Psi(a, b) = k_*a + l_*b, \] and \(\del\colon H_n(X)\to H_{n-1}(A\cap B)\) as follows: every \(\alpha\in H_n(X)\) is represented by a cycle \(z = x + y\) with \(x\) an \(n\)-chain in \(A\) and \(y\) an \(n\)-chain in \(B\), and \(\del\alpha\coloneqq[\del x]\). The Mayer–Vietoris sequence of \((A, B)\) is \[ \cdots\to H_n(A\cap B)\xrightarrow{\Phi} H_n(A)\oplus H_n(B)\xrightarrow{\Psi} H_n(X)\xrightarrow{\del} H_{n-1}(A\cap B)\to\cdots\to H_0(X)\to 0. \]

    Let \(X\) be a topological space and \(A, B\subseteq X\) subspaces with \(X = A^\circ\cup B^\circ\). Then \(\del\colon H_n(X)\to H_{n-1}(A\cap B)\) is well defined and the Mayer–Vietoris sequence of \((A, B)\) is exact [@Hat02].

    Let \(C_n(A+B)\subseteq C_n(X)\) be the subgroup of sums of an \(n\)-chain in \(A\) and an \(n\)-chain in \(B\). Since \(X = A^\circ\cup B^\circ\), the inclusion \(C_\bullet(A+B)\hookrightarrow C_\bullet(X)\) induces isomorphisms on homology. The Mayer–Vietoris sequence is the long exact homology sequence of the short exact sequence of chain complexes \[ 0\to C_n(A\cap B)\xrightarrow{x\mapsto(x, -x)} C_n(A)\oplus C_n(B)\xrightarrow{(x, y)\mapsto x+y} C_n(A+B)\to 0, \] and \(\del\) is its connecting homomorphism; since \(\del z = 0\), the chain \(\del x = -\del y\) lies in \(A\cap B\) and is a cycle there.

    When \(A\cap B\neq\emptyset\) there is an exact sequence of the same form in reduced homology, and there is an exact sequence for singular cohomology with the arrows reversed.

    Write the Mayer-Vietoris long exact sequence for $X = A \cup B$.

    Monodromy action on a fiber

    Let \(p\colon \tilde X \to X\) be a covering space and \(x_0 \in X\). For \(\tilde x\in \inverseof{p}(x_0)\) and a loop \(\gamma\) in \(X\) based at \(x_0\), let \(\tilde\gamma\) be the unique lift of \(\gamma\) with \(\tilde\gamma(0) = \tilde x\). The monodromy action is the right action of \(\pi_1(X, x_0)\) on \(\inverseof{p}(x_0)\) given by \[ \tilde x \cdot [\gamma] \coloneqq \tilde\gamma(1) .\]

    The endpoint \(\tilde\gamma(1)\) depends only on the homotopy class \([\gamma]\), by the homotopy lifting property of covering spaces. The stabilizer of \(\tilde x\) is \(p_*\pi_1(\tilde X, \tilde x)\). If \(X\) is path-connected, the action is transitive if and only if \(\tilde X\) is path-connected, and then \(\tilde x\cdot[\gamma]\mapsto p_*\pi_1(\tilde X, \tilde x)\,[\gamma]\) is a bijection from \(\inverseof{p}(x_0)\) onto the set of right cosets of \(p_*\pi_1(\tilde X, \tilde x)\) in \(\pi_1(X, x_0)\).

    [@Hat02].

    Cells of a CW complex

    Let \(X\) be a CW complex whose \(n\)-skeleton \(X^n\) is obtained from \(X^{n-1}\) by attaching \(n\)-disks \(D^n_\alpha\) along attaching maps \(\varphi_\alpha\colon S^{n-1}\to X^{n-1}\). The characteristic map of \(D^n_\alpha\) is the composite \(\Phi_\alpha\colon D^n_\alpha\hookrightarrow X^{n-1}\sqcup\coprod_\beta D^n_\beta\to X^n\hookrightarrow X\), whose restriction to \(S^{n-1}\) is \(\varphi_\alpha\). The \(n\)-cell \(e^n_\alpha\) is the image \(\Phi_\alpha(D^n\sm S^{n-1})\).

    Each \(n\)-cell \(e^n_\alpha\) is homeomorphic to the open disk \(D^n\sm S^{n-1}\cong\RR^n\), and the underlying set of \(X\) is the disjoint union of the cells \(e^n_\alpha\) over all \(n\) and \(\alpha\).

    See [@Hat02].

    $n$-connected spaces and pairs

    Let \(n\geq 0\). A nonempty topological space \(X\) is \(n\)-connected if \(X\) is path-connected and \(\pi_i(X, x_0) = 0\) for all \(1\leq i\leq n\) and all \(x_0\in X\). A pair \((X, A)\) with \(A\subseteq X\) is \(n\)-connected if every path component of \(X\) meets \(A\) and the relative homotopy groups satisfy \(\pi_i(X, A, x_0) = 0\) for all \(1\leq i\leq n\) and all \(x_0\in A\).

    A space is \(0\)-connected if and only if it is path-connected, and \(1\)-connected if and only if it is simply connected.

    [@Hat02].

    Galois, normal, and regular covering spaces

    Let \(p\colon\tilde X\to X\) be a covering space, let \(x_0\in X\), and let \(\Deck(\tilde X)\) be its group of deck transformations. The covering \(p\) is Galois, also called normal or regular, over \(x_0\) if \(\Deck(\tilde X)\) acts transitively on the fiber \(\inverseof{p}(x_0)\): for all \(\tilde x_1,\tilde x_2\in \inverseof{p}(x_0)\) there is \(\psi\in\Deck(\tilde X)\) with \(\psi(\tilde x_1)=\tilde x_2\).

    Nullhomotopic map

    Let \(X\) and \(Y\) be topological spaces. A continuous map \(f\colon X\to Y\) is nullhomotopic if it is homotopic to a constant map: there are \(y_0\in Y\) and a continuous map \(F\colon X\times[0,1]\to Y\) with \(F(x, 0) = f(x)\) and \(F(x, 1) = y_0\) for all \(x\in X\).

    If \(f\colon X\to Y\) is nullhomotopic, then \(f_*\colon\tilde H_n(X)\to\tilde H_n(Y)\) is zero for every \(n\).

    See [@Hat02].

    Orbit space

    Let \(G\) be a group acting on a topological space \(X\). For \(x,y\in X\) write \(x\sim y\) if there exists \(g\in G\) with \(g\cdot x=y\); this is an equivalence relation whose classes are the orbits of the action. Let \(X/G\coloneqq X/{\sim}\) be the set of orbits and \(q\colon X\to X/G\) the map sending \(x\) to its orbit. The orbit space is \(X/G\) with the quotient topology: \(U\subseteq X/G\) is open if and only if \(\inverseof{q}(U)\) is open in \(X\), so that \(q\) is a quotient map.

    Orientable manifold

    An \(n\)-manifold \(M\) is orientable if there exists an orientation of \(M\).

    Orientation cover

    Let \(M\) be an \(n\)-manifold, and for \(A\subseteq M\) write \(H_n(M\mid A) \coloneqq H_n(M, M\sm A;\ZZ)\). The orientation double cover of \(M\) is the set \(\tilde M\) of all local orientations \(\mu_x\) of \(M\) at points \(x\in M\), with the map \(p\colon\tilde M\to M\), \(\mu_x\mapsto x\), and the topology with basis the sets \[ U(\mu_B) \coloneqq \ts{ \mu_x \st x\in B,\ \mu_x \text{ the image of } \mu_B \text{ under } H_n(M\mid B)\to H_n(M\mid x) }, \] where \(B\) ranges over the open balls of finite radius in charts \(\RR^n\subseteq M\) and \(\mu_B\) over the generators of \(H_n(M\mid B)\cong\ZZ\).

    Let \(M\) be an \(n\)-manifold with orientation double cover \(p\colon\tilde M\to M\). Then \(p\) is a two-sheeted covering space, and \(\tilde M\) is an orientable \(n\)-manifold. If \(M\) is connected, then \(M\) is orientable if and only if \(\tilde M\) has two components.

    [@Hat02].

    Oriented manifold

    Let \(M\) be an \(n\)-manifold. An orientation of \(M\) is a function \(x\mapsto\mu_x\) assigning to each \(x\in M\) a local orientation \(\mu_x\), a generator of \(H_n(M, M\sm\ts{x};\ZZ)\cong\ZZ\), such that every \(x\in M\) has an open ball neighborhood \(B\) of finite radius in a chart \(\RR^n\subseteq M\) and a class \(\mu_B\in H_n(M, M\sm B;\ZZ)\) whose image under the map \(H_n(M, M\sm B;\ZZ)\to H_n(M, M\sm\ts{y};\ZZ)\) induced by inclusion is \(\mu_y\) for every \(y\in B\). An oriented manifold is a pair consisting of a manifold and an orientation of it. The manifold \(M\) is orientable if an orientation of \(M\) exists.

    A nonempty connected orientable \(n\)-manifold has exactly two orientations.

    Let \(M\) be a closed connected oriented \(n\)-manifold with orientation \(\mu\). There is a unique class \([M]\in H_n(M;\ZZ)\) whose image in \(H_n(M, M\sm\ts{x};\ZZ)\) is \(\mu_x\) for every \(x\in M\), and \([M]\) generates \(H_n(M;\ZZ)\cong\ZZ\).

    The class \([M]\) is the fundamental class of \(M\).

    See [@Hat02].

    Path

    Let \(X\) be a topological space and \(I\coloneqq[0,1]\). A path in \(X\) is a continuous map \(\gamma\colon I\to X\); it is a path from \(\gamma(0)\) to \(\gamma(1)\). A path \(\gamma\) is a loop if \(\gamma(0) = \gamma(1)\), and then it is a loop based at \(\gamma(0)\). For paths \(\gamma, \eta\colon I\to X\) with \(\gamma(1) = \eta(0)\), the concatenation \(\gamma\cdot\eta\colon I\to X\) is the path \[ (\gamma\cdot\eta)(s)\coloneqq \begin{cases} \gamma(2s) & 0\leq s\leq 1/2, \\ \eta(2s-1) & 1/2\leq s\leq 1. \end{cases} \]

    See [@Hat02] and [@Mun00].

    Path lifting property

    Let \(p\colon E\to B\) be a continuous map and \(I=[0,1]\). The map \(p\) has the path lifting property if for every path \(\gamma\colon I\to B\) and every \(e_0\in \inverseof{p}(\gamma(0))\) there is a path \(\tilde\gamma\colon I\to E\) with \(p\circ\tilde\gamma=\gamma\) and \(\tilde\gamma(0)=e_0\).

    Identifying \(\pt\times I\) with \(I\), this is the homotopy lifting property of \(p\) with respect to the one-point space \(Y=\pt\).

    Let \(p\colon\tilde X\to X\) be a covering space. For every path \(\gamma\colon I\to X\) and every \(\tilde x_0\in \inverseof{p}(\gamma(0))\) there is a unique path \(\tilde\gamma\colon I\to\tilde X\) with \(p\circ\tilde\gamma=\gamma\) and \(\tilde\gamma(0)=\tilde x_0\) [@Hat02].

    Poincaré duality

    Let \(R\) be a commutative ring and \(M\) a closed $R$-orientable \(n\)-manifold with fundamental class \([M]\in H_n(M;R)\). The Poincaré duality map is \[ D\colon H^k(M;R)\to H_{n-k}(M;R),\qquad D(\alpha)\coloneqq[M]\frown\alpha, \] given by the cap product with \([M]\).

    Let \(R\) be a commutative ring and \(M\) a closed \(R\)-orientable \(n\)-manifold with fundamental class \([M]\in H_n(M;R)\). Then the Poincaré duality map \(D\colon H^k(M;R)\to H_{n-k}(M;R)\), \(\alpha\mapsto[M]\frown\alpha\), is an isomorphism for every \(k\) [@Hat02].

    $R$-orientability

    Let \(R\) be a commutative ring with identity and \(M\) an \(n\)-manifold, and for \(A\subseteq M\) write \(H_n(M\mid A;R)\coloneqq H_n(M, M\sm A; R)\), so that \(H_n(M\mid x;R)\cong R\) for \(x\in M\). An \(R\)-orientation of \(M\) is a function \(x\mapsto\mu_x\) assigning to each \(x\in M\) a generator \(\mu_x\) of the \(R\)-module \(H_n(M\mid x;R)\) that is locally consistent: each \(x\in M\) has a neighborhood \(\RR^n\subseteq M\) containing an open ball \(B\) of finite radius about \(x\) and an element \(\mu_B\in H_n(M\mid B;R)\) whose image under the map \(H_n(M\mid B;R)\to H_n(M\mid y;R)\) induced by inclusion is \(\mu_y\) for every \(y\in B\). The manifold \(M\) is \(R\)-orientable if it has an \(R\)-orientation.

    Let \(M\) be an \(n\)-manifold.

    1. \(M\) is \(\ZZ/2\)-orientable.

    2. If \(M\) is orientable, then \(M\) is \(R\)-orientable for every commutative ring \(R\) with identity.

    3. If \(R\) is a commutative ring with identity in which \(2\neq 0\), then \(M\) is \(R\)-orientable if and only if \(M\) is orientable.

    [@Hat02].

    Relative boundaries

    Let \(X\) be a topological space, \(A\subseteq X\) a subspace, and \(n\geq 0\), and regard the singular chain group \(C_n(A)\) as a subgroup of \(C_n(X)\) via the inclusion. A relative boundary is a chain \(\alpha\in C_n(X)\) of the form \(\alpha = \del\beta + \gamma\) with \(\beta \in C_{n+1}(X)\) and \(\gamma \in C_n(A)\).

    A chain \(\alpha\in C_n(X)\) is a relative boundary if and only if its image in \(C_n(X,A) = C_n(X)/C_n(A)\) is a boundary for the induced boundary map. The relative homology group \(H_n(X,A)\) is isomorphic to the group of relative cycles in \(C_n(X)\) modulo the subgroup of relative boundaries.

    [@Hat02].

    Relative cycle

    Let \(X\) be a topological space and \(A\subseteq X\) a subspace, and regard the singular chain group \(C_n(A)\) as a subgroup of \(C_n(X)\) via the inclusion. The group of relative chains is \(C_n(X,A) \coloneqq C_n(X)/C_n(A)\). A relative cycle is a chain \(\alpha \in C_n(X)\) with \(\del_n \alpha \in C_{n-1}(A)\).

    A chain \(\alpha\in C_n(X)\) is a relative cycle if and only if its image in \(C_n(X,A)\) is a cycle for the boundary map induced by \(\del_n\).

    [@Hat02].

    Relative homotopy groups

    Let \(X\) be a topological space, \(A\subseteq X\) a subspace, \(x_0 \in A\), and \(n\geq 1\). Let \(D^n\) be the closed unit ball, \(S^{n-1} = \del D^n\), and \(s_0\in S^{n-1}\) a basepoint. A map of triples \(f\colon (D^n, S^{n-1}, s_0) \to (X, A, x_0)\) is a continuous map \(f\colon D^n\to X\) with \(f(S^{n-1})\subseteq A\) and \(f(s_0) = x_0\). The relative homotopy group is \[ \pi_n(X, A, x_0) \coloneqq \ts{ f\colon (D^n, S^{n-1}, s_0) \to (X, A, x_0) } / \homotopic, \] the set of homotopy classes of maps of triples, where the homotopies are through maps of triples.

    Let \(X\) be a topological space, \(A\subseteq X\) a subspace, and \(x_0\in A\).

    1. \(\pi_n(X, A, x_0)\) is a group for \(n\geq 2\) and an abelian group for \(n\geq 3\); \(\pi_1(X, A, x_0)\) is a pointed set.

    2. The sequence \[ \cdots \to \pi_n(A, x_0) \to \pi_n(X, x_0) \to \pi_n(X, A, x_0) \xrightarrow{\del} \pi_{n-1}(A, x_0)\to \cdots \to \pi_0(X, x_0), \] whose unlabelled maps are induced by the inclusions \((A, x_0)\injects(X, x_0)\) and \((X, x_0, x_0)\injects (X, A, x_0)\) and where \(\del\) restricts a map of triples to \(S^{n-1}\), is exact.

    [@Hat02].

    Semilocally simply connected space

    A topological space \(X\) is semilocally simply connected if every \(x\in X\) has a neighborhood \(U\) such that the homomorphism \(\pi_1(U, x)\to\pi_1(X, x)\) of fundamental groups induced by the inclusion \(U\hookrightarrow X\) is trivial.

    [@Hat02].

    What does it mean for a space to be semilocally simply connected?

    Short exact sequence

    Let \(R\) be a ring. A short exact sequence of \(R\)-modules is an exact sequence \[ 0\to A\mapsvia{i}B\mapsvia{j}C\to0 \] of \(R\)-modules and \(R\)-linear maps; equivalently, \(i\) is injective, \(j\) is surjective, and \(\im i=\ker j\) [@Hat02]. It splits if there is an \(R\)-linear map \(s\colon C\to B\) with \(j\circ s=\id_C\).

    In a short exact sequence, \(j\) induces an isomorphism \(B/i(A)\cong C\).

    (Splitting lemma)

    For a short exact sequence \(0\to A\mapsvia{i}B\mapsvia{j}C\to0\) of \(R\)-modules, the following are equivalent:

    1. there is an \(R\)-linear \(s\colon C\to B\) with \(j\circ s=\id_C\);

    2. there is an \(R\)-linear \(r\colon B\to A\) with \(r\circ i=\id_A\);

    3. there is an isomorphism \(B\cong A\oplus C\) under which \(i\) becomes \(a\mapsto(a,0)\) and \(j\) becomes \((a,c)\mapsto c\) [@Hat02].

    Simplicial complex

    Let \(N\geq0\) and let \(v_0,\ldots,v_k\in\RR^N\) be affinely independent, meaning that \(v_1-v_0,\ldots,v_k-v_0\) are linearly independent. The \(k\)-simplex \([v_0,\ldots,v_k]\subseteq\RR^N\) is the convex hull of \(v_0,\ldots,v_k\), and these points are its vertices. A face of \([v_0,\ldots,v_k]\) is a simplex spanned by a nonempty subset of \(\ts{v_0,\ldots,v_k}\); for \(0\leq i\leq k\) and \(k\geq1\), the \(i\)-th facet is \[ \del_i[v_0,\ldots,v_k]\coloneqq[v_0,\ldots,\hat v_i,\ldots,v_k]. \] A simplicial complex in \(\RR^N\) is a set \(K\) of simplices in \(\RR^N\) such that:

    1. every face of a simplex in \(K\) belongs to \(K\);

    2. for \(\sigma,\tau\in K\), the intersection \(\sigma\cap\tau\) is either empty or a face of both \(\sigma\) and \(\tau\);

    3. every point of \(\RR^N\) has a neighborhood that meets only finitely many simplices of \(K\).

    Simplicial map

    Let \(K\) and \(L\) be simplicial complexes with vertex sets \(V(K)\) and \(V(L)\). A simplicial map \(f\colon K\to L\) is a map \(f\colon V(K)\to V(L)\) such that whenever \(v_0,\ldots,v_k\) are the vertices of a simplex of \(K\), the points \(f(v_0),\ldots,f(v_k)\), not necessarily distinct, are the vertices of a simplex of \(L\).

    A simplicial map extends to a continuous map \(\abs{K}\to\abs{L}\) of underlying spaces that is affine on each simplex: \(\sum_it_iv_i\mapsto\sum_it_if(v_i)\) for \(t_i\geq0\) with \(\sum_it_i=1\) [@Mun84].

    Simply connected space

    A topological space \(X\) is simply connected if \(X\) is path connected and the fundamental group \(\pi_1(X, x_0)\) is trivial for every \(x_0\in X\).

    For a path connected space \(X\), the following are equivalent:

    1. \(X\) is simply connected;

    2. every continuous map \(\gamma\colon S^1\to X\) is homotopic to a constant map;

    3. every continuous map \(\gamma\colon S^1\to X\) extends to a continuous map \(\hat\gamma\colon D^2\to X\), that is, \(\hat\gamma|_{S^1} = \gamma\);

    4. any two paths \(p_1, p_2\colon[0,1]\to X\) with \(p_1(0) = p_2(0)\) and \(p_1(1) = p_2(1)\) are homotopic rel endpoints.

    See [@Hat02].

    Singular chain

    Let \(X\) be a topological space, \(R\) a commutative ring, and \(p\geq 0\). A singular \(p\)-chain in \(X\) with coefficients in \(R\) is an element of the free \(R\)-module \(C_p(X;R)\) on the singular \(p\)-simplices \(\sigma\colon\Delta^p\to X\), that is, a finite sum \(c = \sum_{i=1}^k r_i\sigma_i\) with \(r_i\in R\) and \(\sigma_i\colon\Delta^p\to X\) continuous.

    See [@Hat02].

    Singular cochain

    Let \(X\) be a topological space, \(G\) an abelian group, and \(n\geq0\). Let \(C_n(X)\) be the free abelian group on the singular \(n\)-simplices \(\sigma\colon\Delta^n\to X\), as in singular homology. A singular \(n\)-cochain of \(X\) with coefficients in \(G\) is a homomorphism \(\varphi\colon C_n(X)\to G\), and \[ C^n(X;G)\coloneqq\Hom(C_n(X),G) \] is the group of singular \(n\)-cochains [@Hat02].

    Since \(C_n(X)\) is free on the singular \(n\)-simplices, a singular \(n\)-cochain is the same as an arbitrary function assigning to each singular \(n\)-simplex \(\sigma\) an element \(\varphi(\sigma)\in G\), extended linearly: \(\varphi\qty{\sum_in_i\sigma_i}=\sum_in_i\varphi(\sigma_i)\).

    Singular homology

    Let \(X\) be a topological space and \(n\geq 0\), and let \(\Delta^n = [v_0, \ldots, v_n]\) be the standard \(n\)-simplex. A singular \(n\)-simplex in \(X\) is a continuous map \(\sigma\colon \Delta^n \to X\). The singular chain group \(C_n(X)\) is the free abelian group on the singular \(n\)-simplices in \(X\), and \(C_n(X)\coloneqq 0\) for \(n<0\). The boundary map \(\del_n\colon C_n(X)\to C_{n-1}(X)\) is the homomorphism with \[ \del_n \sigma \coloneqq \sum_{i=0}^n (-1)^i \ro{\sigma}{[v_0,\ldots, \hat v_i, \ldots, v_n]} \] on singular simplices, each face \([v_0,\ldots, \hat v_i, \ldots, v_n]\) being identified with \(\Delta^{n-1}\) by the order-preserving linear homeomorphism. The singular homology of \(X\) is \[ H_n(X) \coloneqq \ker \del_n / \im \del_{n+1} .\]

    The quotient is defined because \(\del_n \circ \del_{n+1} = 0\).

    [@Hat02].

    Universal cover

    Let \(X\) be a topological space. A universal cover of \(X\) is a covering space \(p\colon \tilde X\to X\) with \(\tilde X\) simply connected [@Hat02].

    Let \(X\) be path-connected and locally path-connected. Then \(X\) has a universal cover if and only if \(X\) is semilocally simply connected [@Hat02], and a universal cover is unique up to isomorphism of covering spaces [@Hat02]. A universal cover covers every path-connected covering space of \(X\) [@Hat02], and its deck transformation group is isomorphic to \(\pi_1(X)\) [@Hat02].

    Weak homotopy equivalence

    A continuous map \(f\colon X\to Y\) of topological spaces is a weak homotopy equivalence if \(f_*\colon\pi_0(X)\to\pi_0(Y)\) is a bijection and, for every \(x_0\in X\) and every \(n\geq1\), the induced map on homotopy groups \[ f_*\colon\pi_n(X,x_0)\to\pi_n(Y,f(x_0)) \] is an isomorphism [@Hat02].

    Let \(f\colon X\to Y\) be a weak homotopy equivalence.

    1. For every \(n\) and every abelian group \(G\), \(f_*\colon H_n(X;G)\to H_n(Y;G)\) and \(f^*\colon H^n(Y;G)\to H^n(X;G)\) are isomorphisms [@Hat02].

    2. (Whitehead) If \(X\) and \(Y\) are CW complexes, then \(f\) is a homotopy equivalence [@Hat02].

    Weak topology determined by a family of subspaces

    Let \(X\) be a topological space and let \((X_\alpha)_{\alpha}\) be a family of subspaces with \(X = \Union_\alpha X_\alpha\). The weak topology determined by \((X_\alpha)_\alpha\) is the topology on \(X\) in which a subset \(A\subseteq X\) is closed if and only if \(A\intersect X_\alpha\) is closed in \(X_\alpha\) for every \(\alpha\).

    The weak topology is the finest topology on the set \(X\) for which every inclusion \(X_\alpha\injects X\) is continuous. A CW complex \(X\) carries the weak topology determined by its skeleta \(X^n\), equivalently by its closed cells.

    [@Hat02].

    Wedge sum

    Let \((X_\alpha, x_\alpha)_{\alpha\in A}\) be a family of based topological spaces. The wedge sum \(\bigvee_{\alpha\in A} X_\alpha\) is the quotient space of the disjoint union \(\coprod_{\alpha\in A} X_\alpha\) obtained by identifying all the basepoints \(x_\alpha\) to a single point, which is its basepoint. For two based spaces, \(X\vee Y\) is the quotient of \(X\sqcup Y\) identifying \(x_0\) with \(y_0\).

    Let \(X\) be a CW complex with \(n\)-skeleton \(X^n\). Then \(X^n/X^{n-1}\) is homeomorphic to \(\bigvee_\alpha S^n_\alpha\), a wedge of \(n\)-spheres with one sphere for each $n$-cell of \(X\).

    If each \((X_\alpha, x_\alpha)\) is a good pair, then the inclusions \(X_\alpha\hookrightarrow\bigvee_\alpha X_\alpha\) induce an isomorphism \[ \bigoplus_\alpha\tilde H_n(X_\alpha)\to\tilde H_n\qty{\bigvee_\alpha X_\alpha} \] for every \(n\).

    See [@Hat02].

    Homotopy

    Cone on a space

    Let \(X\) be a topological space and \(I = [0,1]\). The cone on \(X\) is the quotient space \[ CX\coloneqq(X\times I)/(X\times\ts{0}), \] in which \(X\times\ts{0}\) is collapsed to a point.

    If \(X\) is nonempty, then \(CX\) is contractible, and \(x\mapsto[(x, 1)]\) is an embedding of \(X\) into \(CX\).

    The cone \(CS^1\) on the circle is homeomorphic to the closed disk \(D^2\), by \([(z, t)]\mapsto tz\) for \(z\in S^1\subseteq\CC\) and \(t\in I\).

    See [@Hat02].

    Suspension and reduced suspension

    Let \(X\) be a topological space and \(I=[0,1]\). The suspension \(SX\) is the quotient of \(X\times I\) obtained by collapsing \(X\times\ts{0}\) to one point and \(X\times\ts{1}\) to another point [@Hat02]. For a based space \((X,x_0)\), the reduced suspension is \[ \Sigma X\coloneqq\frac{X\times I}{(X\times\ts{0})\cup(X\times\ts{1})\cup(\ts{x_0}\times I)}, \] the quotient of \(SX\) obtained by collapsing the segment \(\ts{x_0}\times I\) to a point [@Hat02].

    \(SX\) is the union of two copies of the cone \(CX=(X\times I)/(X\times\ts{0})\) glued along their copies of \(X\times\ts{1}\cong X\).

    Smash product

    Let \((X,x_0)\) and \((Y,y_0)\) be based spaces. Identify the wedge sum \(X\vee Y\) with the subspace \(X\times\ts{y_0}\cup\ts{x_0}\times Y\) of \(X\times Y\). The smash product is the quotient space \[ X\wedge Y\coloneqq(X\times Y)/(X\vee Y), \] based at the image of \(X\vee Y\) [@Hat02].

    For \(m,n\geq0\) there is a homeomorphism \(S^m\wedge S^n\cong S^{m+n}\), and for a based space \(X\) the reduced suspension satisfies \(\Sigma X\cong S^1\wedge X\) [@Hat02].

    Moore space

    Let \(G\) be an abelian group and \(n\geq 1\). A Moore space \(M(G, n)\) is a CW complex \(X\) with \[ \tilde H_i(X;\ZZ)\cong \begin{cases} G & i = n, \\ 0 & i\neq n. \end{cases} \]

    For every abelian group \(G\) and \(n\geq 1\) a Moore space \(M(G, n)\) exists. Choose a free resolution \(0\to F_1\mapsvia{\varphi} F_0\to G\to 0\) with a basis \((a_\alpha)_{\alpha}\) of the free abelian group \(F_0\) and a basis \((b_\beta)_{\beta}\) of the free abelian group \(F_1\); such a resolution exists because every subgroup of a free abelian group is free. Let \(X^n\coloneqq\bigvee_\alpha S^n_\alpha\) be a wedge of \(n\)-spheres, one for each \(a_\alpha\), and identify \(H_n(X^n;\ZZ)\) with \(F_0\) by sending the class of \(S^n_\alpha\) to \(a_\alpha\). For each \(\beta\), attach an \((n+1)\)-cell \(e^{n+1}_\beta\) along a map \(f_\beta\colon S^n\to X^n\) with \((f_\beta)_*[S^n] = \varphi(b_\beta)\), which exists because the Hurewicz map \(\pi_n(X^n)\to H_n(X^n;\ZZ)\) is surjective; composed with the collapse \(X^n\to S^n_\alpha\), the map \(f_\beta\) has degree equal to the \(a_\alpha\)-coordinate of \(\varphi(b_\beta)\). The resulting CW complex \(X\) has cellular chain complex \(0\to F_1\mapsvia{\varphi}F_0\to 0\) in degrees \(n+1\) and \(n\), so \(\tilde H_n(X;\ZZ)\cong\coker\varphi\cong G\) and \(\tilde H_{n+1}(X;\ZZ)\cong\ker\varphi = 0\) [@Hat02].

    The sphere \(S^n\) is an \(M(\ZZ, n)\). For \(m\geq 2\), the space \(S^n\cup_m e^{n+1}\) obtained by attaching an \((n+1)\)-cell to \(S^n\) along a map \(S^n\to S^n\) of degree \(m\) is an \(M(\ZZ/m, n)\).

    See [@Hat02].

    Mapping cone

    Let \(f\colon X\to Y\) be a continuous map with mapping cylinder \(M_f=\qty{(X\times I)\disjoint Y}/\qty{(x,1)\sim f(x)}\). The mapping cone of \(f\) is \[ C_f\coloneqq M_f/(X\times\ts{0})=\qty{CX\disjoint Y}/\qty{(x,1)\sim f(x)\text{ for }x\in X}, \] where \(CX=(X\times I)/(X\times\ts{0})\) is the cone on \(X\); that is, \(C_f\) is \(Y\) with \(CX\) attached along \(f\) [@Hat02].

    If \((X,A)\) is a CW pair and \(\iota\colon A\injects X\) is the inclusion, then the mapping cone \(C_\iota=X\cup CA\) is homotopy equivalent to \(X/A\) [@Hat02].

    Mapping cylinder

    Let \(f\colon X\to Y\) be a continuous map and \(I=[0,1]\). The mapping cylinder of \(f\) is the quotient space \[ M_f\coloneqq\qty{(X\times I)\disjoint Y}/\qty{(x,1)\sim f(x)\text{ for } x\in X} \] [@Hat02].

    Let \(\iota\colon X\to M_f\) be the map \(x\mapsto[(x,0)]\) and \(r\colon M_f\to Y\) the map with \(r([(x,t)])=f(x)\) and \(r([y])=y\). Then \(\iota\) is an embedding, \(f=r\circ\iota\), and \(M_f\) deformation retracts onto \(Y\) by sliding each segment \(\ts{x}\times I\) to its endpoint \(f(x)\); in particular \(r\) is a homotopy equivalence. Thus every continuous map factors as an embedding followed by a homotopy equivalence [@Hat02].

    Mapping path space

    Let \(f\colon X\to Y\) be a continuous map, \(I=[0,1]\), and \(Y^I\) the space of paths \(I\to Y\) with the compact-open topology. The mapping path space of \(f\) is the subspace \[ E_f\coloneqq\ts{(x,\gamma)\in X\times Y^I \st \gamma(0)=f(x)} \] of \(X\times Y^I\) [@Hat02].

    Let \(\iota\colon X\to E_f\) be \(x\mapsto(x,c_{f(x)})\), where \(c_y\) is the constant path at \(y\), and let \(p\colon E_f\to Y\) be \((x,\gamma)\mapsto\gamma(1)\). Then \(f=p\circ\iota\), the space \(E_f\) deformation retracts onto \(\iota(X)\), so \(\iota\) is a homotopy equivalence, and \(p\) is a fibration. Thus every continuous map factors as a homotopy equivalence followed by a fibration [@Hat02].

    The mapping path space is the Eckmann–Hilton dual of the mapping cylinder, which factors a map as an embedding followed by a homotopy equivalence.

    Loop space

    Let \((X,x_0)\) be a based space and \(s_0\in S^1\) a basepoint. The loop space of \(X\) is the set \[ \Omega X\coloneqq\ts{\gamma\colon S^1\to X \st \gamma\text{ continuous},\ \gamma(s_0)=x_0} \] with the compact-open topology, based at the constant loop at \(x_0\) [@Hat02].

    Let \((X,x_0)\) and \((Y,y_0)\) be based spaces, and write \([-,-]_*\) for based homotopy classes of based maps and \(\Sigma\) for reduced suspension.

    1. \(\pi_n(\Omega X)\cong\pi_{n+1}(X,x_0)\) for all \(n\geq0\).

    2. There is a bijection \([\Sigma X,Y]_*\cong[X,\Omega Y]_*\), natural in \(X\) and \(Y\).

    [@Hat02]

    Eilenberg--MacLane space

    Let \(n\geq 1\) and let \(G\) be a group, abelian if \(n\geq 2\). An Eilenberg–MacLane space \(K(G, n)\) is a path connected CW complex \(X\) with homotopy groups \[ \pi_i(X)\cong \begin{cases} G & i = n, \\ 0 & i\neq n. \end{cases} \]

    Let \(n\geq 1\) and \(G\) a group, abelian if \(n\geq 2\). A CW complex \(K(G, n)\) exists, and any two are homotopy equivalent. If \(G\) is abelian, then for every CW complex \(X\) there is a natural bijection \[ H^n(X; G)\cong\langle X, K(G, n)\rangle \] with the set of basepoint-preserving homotopy classes of basepoint-preserving maps \(X\to K(G, n)\).

    The circle \(S^1\) is a \(K(\ZZ, 1)\), \(\CP^\infty\) is a \(K(\ZZ, 2)\), and \(\RP^\infty\) is a \(K(\ZZ/2, 1)\).

    [@Hat02].

    Homological algebra

    Tor functors

    Let \(R\) be a commutative ring, let \(B\) be an \(R\)-module, and let \(n\geq 0\). The functor \(\Tor_n^R(\wait, B)\) is the \(n\)th left derived functor of \(\wait\tensor_R B\): \[ \Tor_n^R(A, B) \coloneqq L_n(\wait\tensor_R B)(A) = H_n(P_\bullet\tensor_R B) ,\] where \(P_\bullet \to A\) is a projective resolution of the \(R\)-module \(A\).

    [@DF04].

    Inverse limit

    Let \((I,\leq)\) be a directed set. An inverse system of abelian groups indexed by \(I\) consists of abelian groups \(A_\alpha\) for \(\alpha\in I\) and homomorphisms \(f_{\beta\alpha}\colon A_\beta\to A_\alpha\) for \(\alpha\leq\beta\), with \(f_{\alpha\alpha}=\id_{A_\alpha}\) and \(f_{\beta\alpha}\circ f_{\gamma\beta}=f_{\gamma\alpha}\) for \(\alpha\leq\beta\leq\gamma\). Its inverse limit is the subgroup of compatible families \[ \varprojlim_\alpha A_\alpha\coloneqq\ts{(a_\alpha)_{\alpha\in I}\in\prod_{\alpha\in I}A_\alpha \st f_{\beta\alpha}(a_\beta)=a_\alpha\text{ whenever }\alpha\leq\beta}, \] with the projections \(\pi_\alpha\colon\varprojlim A_\alpha\to A_\alpha\).

    Inverse limits are left exact: if \(0\to A_\alpha\mapsvia{u_\alpha}B_\alpha\mapsvia{v_\alpha}C_\alpha\) is exact for every \(\alpha\in I\) and the \(u_\alpha\) and \(v_\alpha\) commute with the maps of the inverse systems, then the induced sequence \(0\to\varprojlim A_\alpha\mapsvia{u}\varprojlim B_\alpha\mapsvia{v}\varprojlim C_\alpha\) is exact.

    For \(I=\NN\), write \(f_n\colon A_{n+1}\to A_n\) for the bonding maps and define \(\delta\colon\prod_nA_n\to\prod_nA_n\) by \(\delta((a_n)_n)\coloneqq(a_n-f_n(a_{n+1}))_n\). Then \(\ker\delta=\varprojlim A_n\), and \(\varprojlim^1A_n\coloneqq\coker\delta\). A short exact sequence \(0\to A_n\to B_n\to C_n\to0\) of inverse systems indexed by \(\NN\) gives an exact sequence \[ 0\to\varprojlim A_n\to\varprojlim B_n\to\varprojlim C_n\to\varprojlim{}^1A_n\to\varprojlim{}^1B_n\to\varprojlim{}^1C_n\to0, \] so \(\varprojlim^1\) measures the failure of \(\varprojlim B_n\to\varprojlim C_n\) to be surjective [@Hat02].

    Generating set of a module

    Let \(R\) be a ring and \(M\) a left $R$-module. A subset \(S\subseteq M\) is a generating set for \(M\) if every \(x\in M\) is a finite \(R\)-linear combination of elements of \(S\): there exist \(n\geq 0\), \(s_1,\ldots,s_n\in S\), and \(r_1,\ldots,r_n\in R\) with \(x = \sum_{i=1}^n r_i s_i\).

    Acyclic spaces and chain complexes

    A topological space \(X\) is acyclic if its reduced singular homology vanishes, \(\tilde H_n(X;\ZZ) = 0\) for all \(n\). A chain complex \((C_*, \del)\) is acyclic if \(H_n(C_*) = 0\) for all \(n\).

    A nonempty space is acyclic if and only if the map to a point induces isomorphisms on all homology groups. A chain complex is acyclic if and only if it is exact.

    [@Hat02].

    Chain homotopy

    Let \(R\) be a ring, let \((C_*, \del^C)\) and \((D_*, \del^D)\) be chain complexes of \(R\)-modules, and let \(f, g\colon C_*\to D_*\) be chain maps. A chain homotopy from \(f\) to \(g\) is a family of \(R\)-linear maps \(h_i\colon C_i\to D_{i+1}\), \(i\in\ZZ\), such that \[ f_i - g_i = \del^D_{i+1}\circ h_i + h_{i-1}\circ \del^C_i \quad\text{for all } i\in\ZZ .\] The chain maps \(f\) and \(g\) are chain homotopic if a chain homotopy from \(f\) to \(g\) exists.

    [@Hat02].

    Chain map

    Let \(R\) be a ring and let \((C_*, \del^C)\) and \((D_*, \del^D)\) be chain complexes of \(R\)-modules, with \(\del^C_i\colon C_i\to C_{i-1}\) and \(\del^D_i\colon D_i\to D_{i-1}\). A chain map \(f\colon C_*\to D_*\) is a family of \(R\)-linear maps \(f_i\colon C_i\to D_i\), \(i\in\ZZ\), such that \[ f_{i-1}\circ\del^C_i = \del^D_i \circ f_i \quad\text{for all } i\in\ZZ .\]

    Coboundary

    Let \(X\) be a topological space and \(G\) an abelian group, and let \(C^n(X;G)\coloneqq\Hom(C_n(X), G)\) be the group of \(n\)-cochains, where \(C_n(X)\) is the group of singular $n$-chains with boundary map \(\del_n\colon C_n(X)\to C_{n-1}(X)\). The coboundary map \(\delta^n\colon C^n(X;G)\to C^{n+1}(X;G)\) is \(\delta^n(\varphi)\coloneqq\varphi\circ\del_{n+1}\). An \(n\)-cochain \(\varphi\) is a coboundary if \(\varphi\in B^n(X;G)\coloneqq\im\delta^{n-1}\), with \(B^0(X;G)\coloneqq 0\).

    For every \(n\geq 0\), \(\delta^{n+1}\circ\delta^n = 0\); hence \(B^{n+1}(X;G)\subseteq\ker\delta^{n+1}\).

    See [@Hat02].

    Cochain

    Let \(X\) be a topological space, \(R\) a commutative ring, \(p\geq 0\), and \(C_p(X;R)\) the free \(R\)-module on the singular \(p\)-simplices \(\sigma\colon\Delta^p\to X\). A \(p\)-cochain on \(X\) with coefficients in \(R\) is an \(R\)-linear map \(c\colon C_p(X;R)\to R\). The \(p\)-cochains form the \(R\)-module \(C^p(X;R)\coloneqq\Hom_R(C_p(X;R), R)\).

    Since \(C_p(X;R)\) is free on the singular \(p\)-simplices, restriction to them identifies \(C^p(X;R)\) with the \(R\)-module of all functions from the set of singular \(p\)-simplices of \(X\) to \(R\).

    Cocycle

    Let \(X\) be a topological space and \(G\) an abelian group. Let \(C^n(X;G)\coloneqq\Hom(C_n(X),G)\) be the group of singular $n$-cochains, where \(C_n(X)\) is the group of singular $n$-chains with boundary map \(\del_{n+1}\colon C_{n+1}(X)\to C_n(X)\), and let \(\delta^n\colon C^n(X;G)\to C^{n+1}(X;G)\) be the coboundary map \(\delta^n\psi\coloneqq\psi\circ\del_{n+1}\). An \(n\)-cochain \(\psi\in C^n(X;G)\) is an \(n\)-cocycle if \(\delta^n\psi=0\) [@Hat02]. The \(n\)-cocycles form the subgroup \(Z^n(X;G)\coloneqq\ker\delta^n\).

    An \(n\)-cochain \(\psi\in C^n(X;G)\) is a cocycle if and only if \(\psi(b)=0\) for every boundary \(b\in\im\del_{n+1}\).

    With \(B^n(X;G)\coloneqq\im\delta^{n-1}\) the group of coboundaries, the relation \(\delta^n\circ\delta^{n-1}=0\) gives \(B^n(X;G)\subseteq Z^n(X;G)\), and the singular cohomology group is \(H^n(X;G)=Z^n(X;G)/B^n(X;G)\).

    Cycle

    Let \(X\) be a topological space, let \(C_n(X)\) be the group of singular $n$-chains of \(X\), and let \(\del_n\colon C_n(X)\to C_{n-1}(X)\) be the boundary map. An \(n\)-chain \(\alpha\in C_n(X)\) is an \(n\)-cycle if \(\del_n\alpha=0\) [@Hat02]. The \(n\)-cycles form the subgroup \(Z_n(X)\coloneqq\ker\del_n\).

    With \(B_n(X)\coloneqq\im\del_{n+1}\) the group of \(n\)-boundaries, the relation \(\del_n\circ\del_{n+1}=0\) gives \(B_n(X)\subseteq Z_n(X)\), and the singular homology group is \(H_n(X)=Z_n(X)/B_n(X)\).

    Deck transformation

    Let \(p\colon \tilde X\to X\) be a covering space. A deck transformation of \(p\) is a homeomorphism \(f\colon\tilde X\to\tilde X\) with \(p\circ f = p\). The deck transformations of \(p\) form a group \(\Deck(p)\) under composition.

    Left derived functor

    Let \(R\) be a ring and let \(T\) be an additive covariant functor from \(R\)-modules to abelian groups. For an \(R\)-module \(A\), choose a projective resolution \(\cdots\to P_1\to P_0\to A\to 0\) and let \(P_\bullet\) be the complex \(\cdots \to P_1\to P_0\to 0\) with \(A\) deleted. The left derived functors of \(T\) are \[ L_n T(A) \coloneqq H_n\qty{T(P_\bullet)}, \qquad n\geq 0 .\]

    Any two projective resolutions of \(A\) are chain homotopy equivalent, so \(L_nT(A)\) is well defined up to natural isomorphism, and a morphism \(A\to A'\) induces \(L_nT(A)\to L_nT(A')\). If \(T\) is right exact, then \(L_0T\cong T\).

    [@Wei94].

    Direct limit

    Let \((I,\leq)\) be a directed set. A directed system of abelian groups indexed by \(I\) consists of abelian groups \(A_\alpha\) for \(\alpha\in I\) and homomorphisms \(f_{\alpha\beta}\colon A_\alpha\to A_\beta\) for \(\alpha\leq\beta\), with \(f_{\alpha\alpha}=\id_{A_\alpha}\) and \(f_{\beta\gamma}\circ f_{\alpha\beta}=f_{\alpha\gamma}\) for \(\alpha\leq\beta\leq\gamma\). Its direct limit is the set \[ \varinjlim_\alpha A_\alpha\coloneqq\qty{\Disjoint_{\alpha\in I}A_\alpha}/\sim, \] where \(\sim\) is the equivalence relation generated by \(a\sim f_{\alpha\beta}(a)\) for \(a\in A_\alpha\) and \(\alpha\leq\beta\), with the group operation \([a]+[b]\coloneqq[f_{\alpha\gamma}(a)+f_{\beta\gamma}(b)]\) for \(a\in A_\alpha\), \(b\in A_\beta\), and any \(\gamma\geq\alpha,\beta\) [@Hat02]. For each \(\alpha\), the map \(\iota_\alpha\colon A_\alpha\to\varinjlim A_\alpha\), \(a\mapsto[a]\), is a homomorphism with \(\iota_\beta\circ f_{\alpha\beta}=\iota_\alpha\).

    Because \(I\) is directed, \(a\in A_\alpha\) and \(b\in A_\beta\) satisfy \(a\sim b\) if and only if \(f_{\alpha\gamma}(a)=f_{\beta\gamma}(b)\) for some \(\gamma\geq\alpha,\beta\); this makes the group operation well defined. With the maps \(\iota_\alpha\), the direct limit is the colimit of the system in the category of abelian groups.

    Direct limits over a directed set are exact: if \(A_\alpha\mapsvia{u_\alpha}B_\alpha\mapsvia{v_\alpha}C_\alpha\) is an exact sequence for each \(\alpha\in I\), and the \(u_\alpha\) and \(v_\alpha\) commute with the maps of directed systems \((A_\alpha)\), \((B_\alpha)\), \((C_\alpha)\), then the induced sequence \(\varinjlim A_\alpha\mapsvia{u}\varinjlim B_\alpha\mapsvia{v}\varinjlim C_\alpha\) is exact.

    If a CW complex \(X\) is the union of a family of subcomplexes \(X_\alpha\) directed by inclusion, then the inclusions induce an isomorphism \(\varinjlim_\alpha H_n(X_\alpha;G)\cong H_n(X;G)\) for every \(n\) and every abelian group \(G\) [@Hat02].

    Direct product of modules

    Let \(R\) be a ring and \((M_\alpha)_{\alpha\in A}\) a family of left \(R\)-modules. The direct product of the family is the set \[ \prod_{\alpha\in A} M_\alpha \coloneqq \ts{ (m_\alpha)_{\alpha\in A} \st m_\alpha \in M_\alpha \text{ for all } \alpha\in A } \] with componentwise addition and scalar multiplication, together with the projections \(\pi_\beta\colon\prod_{\alpha\in A} M_\alpha\to M_\beta\), \((m_\alpha)_{\alpha\in A}\mapsto m_\beta\), for \(\beta\in A\).

    For every left \(R\)-module \(N\) and every family of \(R\)-linear maps \(f_\beta\colon N\to M_\beta\), \(\beta\in A\), there is a unique \(R\)-linear map \(f\colon N\to \prod_{\alpha\in A} M_\alpha\) with \(\pi_\beta\circ f = f_\beta\) for all \(\beta\in A\), namely \(f(n) = (f_\alpha(n))_{\alpha\in A}\). Thus \(\prod_{\alpha\in A} M_\alpha\) is the categorical product of the family in the category of left \(R\)-modules.

    In \(\Top\), the categorical product of a family of spaces \((X_\alpha)_{\alpha\in A}\) is the set \(\prod_{\alpha\in A} X_\alpha\) with the product topology, the coarsest topology for which every projection \(\pi_\beta\) is continuous.

    [@DF04]; [@Hat02].

    Direct sum of modules

    Let \(R\) be a ring and \((M_\alpha)_{\alpha\in A}\) a family of \(R\)-modules with direct product \(\prod_{\alpha}M_\alpha\). The direct sum of the family is the submodule \[ \bigoplus_{\alpha\in A}M_\alpha\coloneqq\ts{(m_\alpha)_{\alpha\in A}\in\prod_{\alpha\in A}M_\alpha \st m_\alpha=0\text{ for all but finitely many }\alpha}, \] with the inclusions \(\iota_\beta\colon M_\beta\injects\bigoplus_\alpha M_\alpha\) sending \(m\) to the family with \(m\) in position \(\beta\) and \(0\) elsewhere [@DF04].

    The direct sum is the coproduct in the category of \(R\)-modules: for every \(R\)-module \(N\) and every family of \(R\)-linear maps \(\varphi_\alpha\colon M_\alpha\to N\) there is a unique \(R\)-linear map \(\varphi\colon\bigoplus_\alpha M_\alpha\to N\) with \(\varphi\circ\iota_\alpha=\varphi_\alpha\) for all \(\alpha\) [@DF04]. If \(A\) is finite, then \(\bigoplus_\alpha M_\alpha=\prod_\alpha M_\alpha\).

    Exact sequence

    Let \(R\) be a ring and let \[ \cdots\to A_{n+1}\mapsvia{d_{n+1}}A_n\mapsvia{d_n}A_{n-1}\to\cdots \] be a sequence of \(R\)-modules and \(R\)-linear maps. The sequence is exact at \(A_n\) if \(\im d_{n+1}=\ker d_n\), and exact if it is exact at every term that has both an incoming and an outgoing map [@Hat02].

    The sequence is a chain complex when \(\im d_{n+1}\subseteq\ker d_n\) for all \(n\). A chain complex is exact at \(A_n\) if and only if its homology \(\ker d_n/\im d_{n+1}\) at \(A_n\) is zero.

    Ext group

    Let \(A\) and \(G\) be abelian groups, and let \[ \cdots \to F_2 \to F_1 \to F_0 \to A \to 0 \] be a free resolution of \(A\). Applying \(\Hom(\wait, G)\) to \(\cdots\to F_1\to F_0\) gives a cochain complex \(0\to\Hom(F_0,G)\to\Hom(F_1,G)\to\cdots\), and the Ext group \(\Ext^n(A, G)\) is its \(n\)th cohomology group; it is independent of the choice of free resolution [@Hat02].

    One has \(\Ext^0(A,G) \cong \Hom(A,G)\) and \(\Ext^n(A,G) = 0\) for \(n\geq 2\). For \(m\geq 1\), \[ \Ext^1(\ZZ, G) = 0, \qquad \Ext^1(\ZZ/m, G) \cong G/mG \] [@Hat02].

    Flat module

    Let \(R\) be a commutative ring. An \(R\)-module \(M\) is flat if the functor \(M\otimes_R -\) from \(R\)-modules to \(R\)-modules is exact.

    Free module

    Let \(R\) be a ring. An $R$-module \(M\) is free if it has a basis: a family \((\beta_i)_{i\in I}\) in \(M\) such that every \(m\in M\) can be written as \(m=\sum_{i\in I} r_i\beta_i\) with \(r_i\in R\), all but finitely many zero, and \(\sum_{i\in I} r_i\beta_i = 0\) implies \(r_i = 0\) for all \(i\).

    Let \(R\) be a ring and \(M\) an \(R\)-module. The following are equivalent:

    1. \(M\) is free, with basis \((\beta_i)_{i\in I}\);

    2. there are a set \(\mathcal{B}\) and a map of sets \(\iota\colon\mathcal{B}\to M\) such that for every \(R\)-module \(N\) and every map of sets \(f\colon\mathcal{B}\to N\) there is a unique \(R\)-linear map \(\tilde f\colon M\to N\) with \(\tilde f\circ\iota = f\):

    1. there is a family \((\beta_i)_{i\in I}\) in \(M\) such that \(M = \bigoplus_{i\in I} R\beta_i\) is the internal direct sum of the cyclic submodules \(R\beta_i\), and \(r\mapsto r\beta_i\) is injective for each \(i\), so that \(R\beta_i\cong R\).

    Given a basis \((\beta_i)_{i\in I}\), take \(\mathcal B=I\) and \(\iota(i)=\beta_i\); the unique extension of \(f\) is \(\tilde f(\sum_i r_i\beta_i)=\sum_i r_if(i)\), which is well defined because the coefficients \(r_i\) are unique. Conversely, if \((\mathcal B,\iota)\) has the universal property, compare \(M\) with the module \(R^{(\mathcal B)}\) of finitely supported functions \(\mathcal B\to R\) and its standard basis \((e_b)_{b\in\mathcal B}\): the linear maps \(M\to R^{(\mathcal B)}\) extending \(b\mapsto e_b\) and \(R^{(\mathcal B)}\to M\) extending \(e_b\mapsto\iota(b)\) compose in both orders to maps that restrict to the identity on the generators, so by uniqueness they are inverse isomorphisms, and \((\iota(b))_{b\in\mathcal B}\) is a basis of \(M\). Finally, the spanning condition says that \(M=\sum_i R\beta_i\), and linear independence says exactly that this sum is direct and each \(r\mapsto r\beta_i\) is injective.

    Let \(R\) be a ring, \(M\) a left \(R\)-module, \(S\subseteq M\) a subset, and \(R^{(S)} \coloneqq \bigoplus_{s\in S} R\) with standard basis \((e_s)_{s\in S}\). There is a unique \(R\)-linear map \(\varphi\colon R^{(S)}\to M\) with \(\varphi(e_s) = s\) for all \(s\in S\).

    1. \(\varphi\) is surjective if and only if \(S\) is a generating set of \(M\); in that case \(M\cong R^{(S)}/\ker\varphi\).

    2. \(\varphi\) is an isomorphism if and only if \(S\) is a basis of \(M\). In particular, \(M\) is free if and only if \(M\cong R^{(T)}\) for some set \(T\) [@DF04].

    Group ring

    Let \(R\) be a commutative ring and \(G = \ts{g_1, \ldots, g_n}\) a finite group. The group ring \(RG\) is the free \(R\)-module with basis the elements of \(G\), \[ RG\coloneqq\ts{ \sum_{i=1}^n a_i g_i \st a_i\in R }, \] with the multiplication that extends the group law \(R\)-bilinearly: \((ag)(bh)\coloneqq(ab)(gh)\) for \(a, b\in R\) and \(g, h\in G\).

    The group ring \(RG\) is a ring with identity \(1_R 1_G\). If \(R\neq 0\), then \(RG\) is commutative if and only if \(G\) is abelian.

    See [@DF04].

    Homologous cycles

    Let \(X\) be a topological space and \(n\geq 0\). Two cycles \(\alpha, \beta \in Z_n(X)\) are homologous if there exists \(\gamma\in C_{n+1}(X)\) with \(\alpha - \beta = \del_{n+1}\gamma\).

    Two cycles are homologous if and only if they represent the same class in the singular homology group \(H_n(X)\).

    [@Hat02].

    Limit of a diagram

    Let \(J\) be a small category, \(\mathcal C\) a category, and \(F\colon J\to\mathcal C\) a functor. A cone over \(F\) is an object \(Y\) of \(\mathcal C\) with morphisms \(\psi_j\colon Y\to F(j)\) for \(j\in J\) such that \(F(u)\circ\psi_j=\psi_k\) for every morphism \(u\colon j\to k\) in \(J\). A limit of \(F\) is a cone \((L,(\pi_j)_{j\in J})\) over \(F\) such that for every cone \((Y,(\psi_j)_{j\in J})\) over \(F\) there is a unique morphism \(h\colon Y\to L\) with \(\pi_j\circ h=\psi_j\) for all \(j\in J\).

    Products are limits over a category \(J\) with no nonidentity morphisms, pullbacks are limits over \(\bullet\to\bullet\leftarrow\bullet\), and inverse limits are limits over a directed set, viewed as a category with a morphism \(\beta\to\alpha\) whenever \(\alpha\leq\beta\).

    The dual notion, obtained by reversing every morphism, is the colimit. If every functor \(J\to\mathcal C\) has a limit, then \(\lim\colon\mathcal C^J\to\mathcal C\) is a right adjoint of the diagonal functor \(\Delta\colon\mathcal C\to\mathcal C^J\), which sends an object to the constant functor at it.

    Linearly independent subset of a module

    Let \(R\) be a ring and \(M\) an \(R\)-module. A subset \(S\subseteq M\) is linearly independent if for all \(n\geq 1\), all pairwise distinct \(s_1, \ldots, s_n\in S\) and all \(r_1, \ldots, r_n\in R\), \[ r_1 s_1 + \cdots + r_n s_n = 0 \implies r_1 = \cdots = r_n = 0 .\]

    [@DF04].

    Long exact sequence

    A long exact sequence is an exact sequence with infinitely many terms, indexed by \(\ZZ\) or by an infinite interval of \(\ZZ\).

    Let \(0\to A_\bullet\mapsvia{i}B_\bullet\mapsvia{j}C_\bullet\to0\) be a short exact sequence of chain complexes of abelian groups. There are homomorphisms \(\del\colon H_n(C_\bullet)\to H_{n-1}(A_\bullet)\), the connecting homomorphisms, such that \[ \cdots\to H_n(A_\bullet)\mapsvia{i_*}H_n(B_\bullet)\mapsvia{j_*}H_n(C_\bullet)\mapsvia{\del}H_{n-1}(A_\bullet)\to\cdots \] is exact [@Hat02].

    For a topological space \(X\) and a subspace \(A\), the short exact sequence \(0\to C_\bullet(A)\to C_\bullet(X)\to C_\bullet(X,A)\to0\) of singular chain complexes gives the long exact sequence of the pair \[ \cdots\to H_n(A)\to H_n(X)\to H_n(X,A)\mapsvia{\del}H_{n-1}(A)\to\cdots\to H_0(X,A)\to0 \] [@Hat02].

    Perfect pairing

    Let \(R\) be a commutative ring and \(M\), \(N\), \(L\) \(R\)-modules. A pairing of \(M\) and \(N\) with values in \(L\) is an \(R\)-bilinear map \(p\colon M\times N\to L\), equivalently an \(R\)-linear map \(M\otimes_R N\to L\) on the tensor product. The pairing \(p\) is perfect if the \(R\)-linear map \[ \phi\colon M\to\Hom_R(N, L),\qquad \phi(m)(n)\coloneqq p(m, n), \] is an isomorphism.

    Pullback of spaces

    Let \(f\colon X\to Z\) and \(g\colon Y\to Z\) be continuous maps of topological spaces. The pullback of \(f\) and \(g\) is the subspace \[ X \cross_Z Y \coloneqq \ts{ (x,y) \in X\cross Y \st f(x) = g(y) } \] of the product space \(X\cross Y\), together with the restrictions \(p_X\colon X\cross_Z Y\to X\) and \(p_Y\colon X\cross_Z Y\to Y\) of the two projections.

    Let \(f\colon X\to Z\) and \(g\colon Y\to Z\) be continuous maps with pullback \((X\cross_Z Y, p_X, p_Y)\). Then \(f\circ p_X = g\circ p_Y\), and for every topological space \(W\) and continuous maps \(a\colon W\to X\) and \(b\colon W\to Y\) with \(f\circ a = g\circ b\) there is a unique continuous map \(u\colon W\to X\cross_Z Y\) with \(p_X\circ u = a\) and \(p_Y\circ u = b\), namely \(u(w) = (a(w), b(w))\). Thus \(X\cross_Z Y\) is the limit of the diagram \(X\xrightarrow{f} Z \xleftarrow{g} Y\) in \(\Top\). If \(Z\) is a point, then \(X\cross_Z Y = X\cross Y\).

    [@Hat02].

    Pushout of topological spaces

    Let \(f\colon Z\to X\) and \(g\colon Z\to Y\) be continuous maps. The pushout of \(X \xleftarrow{f} Z \mapsvia{g} Y\) is the quotient space \[ X \disjoint_Z Y \coloneqq \qty{X \disjoint Y}/\sim ,\] where \(\sim\) is the equivalence relation generated by \(f(z)\sim g(z)\) for \(z\in Z\), together with the maps \(j_X\colon X\to X\disjoint_Z Y\) and \(j_Y\colon Y\to X\disjoint_Z Y\) induced by the inclusions.

    The maps satisfy \(j_X\circ f = j_Y\circ g\), and \((X\disjoint_Z Y, j_X, j_Y)\) is the colimit of \(X \xleftarrow{f} Z \mapsvia{g} Y\) in the category of topological spaces: for continuous maps \(a\colon X\to W\) and \(b\colon Y\to W\) with \(a\circ f = b\circ g\) there is a unique continuous \(\varphi\colon X\disjoint_Z Y\to W\) with \(\varphi\circ j_X = a\) and \(\varphi\circ j_Y = b\). When \(g\) is the inclusion of a subspace \(Z\subseteq Y\), the pushout is the space obtained by attaching $Y$ to $X$ along $f$. The dual construction is the pullback. If \(X = U\union V\) with \(U, V\) open and path-connected, \(U\intersect V\) path-connected and \(x_0\in U\intersect V\), then \(X\) is the pushout of \(U \hookleftarrow U\intersect V \injects V\), and the Seifert–van Kampen theorem gives the corresponding pushout of groups, the amalgamated free product \(\pi_1(X, x_0)\cong \pi_1(U, x_0) \ast_{\pi_1(U\intersect V, x_0)} \pi_1(V, x_0)\) [@Hat02].

    [@Hat02].

    Quasi-isomorphism

    Let \(R\) be a ring and \(C_\bullet\), \(D_\bullet\) chain complexes of \(R\)-modules. A chain map \(f\colon C_\bullet\to D_\bullet\) is a quasi-isomorphism if the induced map \(H_n(f)\colon H_n(C_\bullet)\to H_n(D_\bullet)\) is an isomorphism for every \(n\in\ZZ\).

    Every chain homotopy equivalence is a quasi-isomorphism.

    A quasi-isomorphism need not be a chain homotopy equivalence. Let \(C_\bullet\) be the complex of abelian groups with \(C_1 = \ZZ\), \(C_0 = \ZZ\), differential \(C_1\to C_0\) multiplication by \(2\), and all other terms zero, and let \(D_\bullet\) be \(\ZZ/2\) concentrated in degree \(0\). The reduction map \(C_0 = \ZZ\to\ZZ/2 = D_0\) is a chain map inducing an isomorphism \(H_0(C_\bullet) = \ZZ/2\to H_0(D_\bullet)\), and all other homology groups vanish, so it is a quasi-isomorphism. Every chain map \(D_\bullet\to C_\bullet\) is zero, since \(\Hom(\ZZ/2, \ZZ) = 0\), so it induces the zero map on \(H_0\) and cannot be a homotopy inverse.

    The derived category of \(R\)-modules is obtained from the category of chain complexes by formally inverting the quasi-isomorphisms.

    [@Wei94].

    Projective resolution

    Let \(R\) be a ring and \(M\) an \(R\)-module. A projective resolution of \(M\) is an exact sequence of \(R\)-modules \[ \cdots\to P_2\mapsvia{d_2}P_1\mapsvia{d_1}P_0\mapsvia{\varepsilon}M\to0 \] in which every \(P_i\) is a projective module. It is a free resolution if every \(P_i\) is a free module.

    Let \(R\) be a ring and \(M\) an \(R\)-module.

    1. \(M\) has a free resolution, and hence a projective resolution.

    2. If \(P_\bullet\to M\) and \(Q_\bullet\to M\) are projective resolutions, there is a chain map \(P_\bullet\to Q_\bullet\) over \(\id_M\), and any such chain map is a chain homotopy equivalence [@Hat02].

    Let \(R\) be commutative and \(N\) an \(R\)-module. By (b), the cohomology of the cochain complex \(\Hom_R(P_\bullet,N)\) and the homology of the chain complex \(P_\bullet\otimes_RN\) do not depend on the projective resolution \(P_\bullet\) of \(M\), up to canonical isomorphism; they are \(\Ext^n_R(M,N)\) and \(\Tor_n^R(M,N)\).

    What links to this