Jump to content

Stiefel–Whitney class: Difference between revisions

From Wikipedia, the free encyclopedia
Content deleted Content added
No edit summary
m Wiki-Gnoming: added lots of \scriptstyle commeands to make the LaTeX fit better into the text
Line 1: Line 1:
In [[mathematics]], the '''Stiefel–Whitney class''' arises as a type of [[characteristic class]] associated to [[Vector bundle#Definition and first consequences|real vector bundles]] <math>E\rightarrow X</math>. It is denoted by ''w''(''E''), taking values in <math>H^*(X; \Z/2\Z)</math>, the [[cohomology group]]s with [[Modular arithmetic#Some consequences of the mathematical usage|mod]] 2 coefficients. The component of <math>w(E)</math> in <math>H^i(X; \Z/2\Z)</math> is denoted by <math>w_i(E)</math> and called the ''<math>i</math>-th Stiefel-Whitney class of <math>E</math>'', so that <math>w(E) = w_0(E) + w_1(E) + w_2(E) + \cdots</math>. As an example, over the [[circle]], <math>S^1</math>, there is a [[line bundle]] that is topologically non-trivial: that is, the line bundle associated to the [[Möbius strip|Möbius band]], usually thought of as having fibres <math>[0,1]</math>. The cohomology group
In [[mathematics]], the '''Stiefel–Whitney class''' arises as a type of [[characteristic class]] associated to [[Vector bundle#Definition and first consequences|real vector bundles]] <math>\scriptstyle E\rightarrow X</math>. It is denoted by ''w''(''E''), taking values in <math>\scriptstyle H^*(X; \Z/2\Z)</math>, the [[cohomology group]]s with [[Modular arithmetic#Some consequences of the mathematical usage|mod]] 2 coefficients. The component of <math>\scriptstyle w(E)</math> in <math>\scriptstyle H^i(X; \Z/2\Z)</math> is denoted by <math>\scriptstyle w_i(E)</math> and called the ''<math>\scriptstyle i</math>-th Stiefel-Whitney class of <math>\scriptstyle E</math>'', so that <math>\scriptstyle w(E) = w_0(E) + w_1(E) + w_2(E) + \cdots</math>. As an example, over the [[circle]], <math>\scriptstyle S^1</math>, there is a [[line bundle]] that is topologically non-trivial: that is, the line bundle associated to the [[Möbius strip|Möbius band]], usually thought of as having fibres <math>\scriptstyle [0,1]</math>. The cohomology group
:<math>H^1(S^1;\mathbb Z/2\mathbb Z)</math>
:<math>H^1(S^1;\mathbb Z/2\mathbb Z)</math>
has just one element other than 0, this element being the first Stiefel-Whitney class, <math>w_1</math>, of that line bundle.
has just one element other than 0, this element being the first Stiefel-Whitney class, <math>\scriptstyle w_1</math>, of that line bundle.


==Origins==
==Origins==


The Stiefel-Whitney classes <math>w_i(E)</math> get their name because Stiefel and [[Hassler Whitney|Whitney]] discovered them as [[modulo|mod-2]] reductions of the [[obstruction theory|obstruction classes]] to constructing <math>n-i+1</math> everywhere [[linearly independent]] [[Section (fiber bundle)|sections]] of the [[vector bundle]] <math>E</math> restricted to the <math>i</math>-skeleton of <math>X</math>. Here <math>n</math> denotes the dimension of the fibre of the vector bundle <math>F \to E \to X</math>.
The Stiefel-Whitney classes <math>\scriptstyle w_i(E)</math> get their name because Stiefel and [[Hassler Whitney|Whitney]] discovered them as [[modulo|mod-2]] reductions of the [[obstruction theory|obstruction classes]] to constructing <math>\scriptstyle n-i+1</math> everywhere [[linearly independent]] [[Section (fiber bundle)|sections]] of the [[vector bundle]] <math>\scriptstyle E</math> restricted to the <math>\scriptstyle i</math>-skeleton of <math>\scriptstyle X</math>. Here <math>\scriptstyle n</math> denotes the dimension of the fibre of the vector bundle <math>\scriptstyle F \to E \to X</math>.


To be precise, provided <math>X</math> is a [[CW-complex]], Whitney defined classes <math>W_i(E)</math> in the <math>i</math>-th cellular [[cohomology group]] of <math>X</math> with twisted coefficients. The coefficient system being the <math>(i-1)</math>-st [[homotopy group]] of the [[Stiefel manifold]] of <math>(n-i+1)</math> linearly independent vectors in the fibres of <math>E</math>. Whitney proved <math>W_i(E)=0</math> if and only if <math>E</math>, when restricted to the <math>i</math>-skeleton of <math>X</math>, has <math>(n-i+1)</math> linearly-independent sections.
To be precise, provided <math>\scriptstyle X</math> is a [[CW-complex]], Whitney defined classes <math>\scriptstyle W_i(E)</math> in the <math>\scriptstyle i</math>-th cellular [[cohomology group]] of <math>\scriptstyle X</math> with twisted coefficients. The coefficient system being the <math>\scriptstyle (i-1)</math>-st [[homotopy group]] of the [[Stiefel manifold]] of <math>\scriptstyle (n-i+1)</math> linearly independent vectors in the fibres of <math>\scriptstyle E</math>. Whitney proved <math>\scriptstyle W_i(E)=0</math> if and only if <math>\scriptstyle E</math>, when restricted to the <math>\scriptstyle i</math>-skeleton of <math>\scriptstyle X</math>, has <math>\scriptstyle (n-i+1)</math> linearly-independent sections.


Since [[Stiefel manifold|<math>\pi_{i-1} V_{n-i+1}(F)</math>]] is either infinite-[[Cyclic group|cyclic]] or [[isomorphic]] to <math>\Bbb Z_2</math>, there is a [[canonical#Mathematics|canonical]] reduction of the <math>W_i(E)</math> classes to classes <math>w_i(E) \in H^i(X;\Bbb Z_2)</math> which are the Stiefel-Whitney classes. Moreover, whenever <math>\pi_{i-1} V_{n-i+1}(F) = \Bbb Z_2</math>, the two classes are identical. Thus, <math>w_1(E) = 0</math> if and only if the bundle <math>E \to X</math> is [[orientable]].
Since [[Stiefel manifold|<math>\scriptstyle \pi_{i-1} V_{n-i+1}(F)</math>]] is either infinite-[[Cyclic group|cyclic]] or [[isomorphic]] to <math>\scriptstyle \Bbb Z_2</math>, there is a [[canonical#Mathematics|canonical]] reduction of the <math>\scriptstyle W_i(E)</math> classes to classes <math>\scriptstyle w_i(E) \in H^i(X;\Bbb Z_2)</math> which are the Stiefel-Whitney classes. Moreover, whenever <math>\scriptstyle \pi_{i-1} V_{n-i+1}(F) = \Bbb Z_2</math>, the two classes are identical. Thus, <math>\scriptstyle w_1(E) = 0</math> if and only if the bundle <math>\scriptstyle E \to X</math> is [[orientable]].


The <math>w_0(E)</math> class is exceptional and has no meaning a priori. Its creation by Whitney was an act of creative notation, allowing the [[Whitney sum]] Formula <math>w(E_1 \oplus E_2) = w(E_1) w(E_2)</math> to be true. However, for generalizations of manifolds (namely certain [[homology manifold]]s), one can have <math>w_0(M) \neq 1</math> – it only needs to equal 1 mod 8.
The <math>\scriptstyle w_0(E)</math> class is exceptional and has no meaning a priori. Its creation by Whitney was an act of creative notation, allowing the [[Whitney sum]] Formula <math>\scriptstyle w(E_1 \oplus E_2) = w(E_1) w(E_2)</math> to be true. However, for generalizations of manifolds (namely certain [[homology manifold]]s), one can have <math>\scriptstyle w_0(M) \neq 1</math> – it only needs to equal 1 mod 8.


==Axioms==
==Axioms==


Throughout, <math>H^i(X;G)</math> denotes [[singular cohomology]] of a space <math>X</math> with coefficients in the [[group (mathematics)|group]] <math>G</math>.
Throughout, <math>\scriptstyle H^i(X;G)</math> denotes [[singular cohomology]] of a space <math>\scriptstyle X</math> with coefficients in the [[group (mathematics)|group]] <math>\scriptstyle G</math>.


# Naturality: <math>w(f^* E) = f^* w(E)</math> for any bundle <math>E \to X</math> and map <math>f:X' \to X</math>, where <math>f^*E</math> denotes the [[Pullback bundle|induced bundle]].
# Naturality: <math>\scriptstyle w(f^* E) = f^* w(E)</math> for any bundle <math>\scriptstyle E \to X</math> and map <math>\scriptstyle f:X' \to X</math>, where <math>\scriptstyle f^*E</math> denotes the [[Pullback bundle|induced bundle]].
# <math>w_0(E)=1</math> in <math>H^0(X;\mathbb Z/2\mathbb Z)</math>.
# <math>\scriptstyle w_0(E)=1</math> in <math>\scriptstyle H^0(X;\mathbb Z/2\mathbb Z)</math>.
# <math>w_1(\gamma^1)</math> is the generator of <math>H^1(\mathbb RP^1;\mathbb
# <math>\scriptstyle w_1(\gamma^1)</math> is the generator of <math>\scriptstyle H^1(\mathbb RP^1;\mathbb
Z/2\mathbb Z)\cong\mathbb Z/2\mathbb Z</math> (normalization condition). Here, <math>\gamma^n</math> is the [[canonical line bundle]].
Z/2\mathbb Z)\cong\mathbb Z/2\mathbb Z</math> (normalization condition). Here, <math>\scriptstyle \gamma^n</math> is the [[canonical line bundle]].
# <math>w(E\oplus F)= w(E) \smallsmile w(F)</math> (Whitney product formula).
# <math>\scriptstyle w(E\oplus F)= w(E) \smallsmile w(F)</math> (Whitney product formula).


Some work is required to show that such classes do indeed exist and are unique (at least for [[paracompact space]]s ''X''); see section 17.2 and 17.3 in Husemoller or section 8 in Milnor and Stasheff.
Some work is required to show that such classes do indeed exist and are unique (at least for [[paracompact space]]s ''X''); see section 17.2 and 17.3 in Husemoller or section 8 in Milnor and Stasheff.
Line 27: Line 27:
==Line bundles==
==Line bundles==


Let <math>X</math> be a paracompact space, and let <math>Vect_n(X)</math> denote the set of real vector bundles over X of dimension n for some fixed positive integer <math>n</math>. For any vector space V, let <math>Gr_n(V)</math> denote the [[Grassmannian]] <math>Gr_n(V) = \{W\subset V:\, \dim W = n\}</math>. Set <math>Gr_n = Gr_n(\R^\infty)</math>. Define the [[tautological bundle]] <math>\gamma^n \to Gr_n</math> by <math>\gamma^n = \{(W, x):\, W\in Gr_n, x\in W\}</math>; this is a real bundle of dimension n, with projection <math>\gamma^n \to Gr_n</math> given by <math>(W, x) \to W</math>. For any map <math>f:X \to Gr_n</math>, the induced bundle <math>f^*\gamma^n \in Vect_n(X)</math>. Since any two homotopic maps <math>f, g: X \to Gr_n</math> have <math>f^*\gamma^n</math> and <math>g^*\gamma^n</math> isomorphic, the map <math>\alpha:[X; Gr_n] \to Vect_n(X)</math> given by <math>f \to f^* \gamma^n</math> is well-defined, where <math>[X; Gr_n]</math> denotes the set of homotopy equivalence classes of maps <math>X \to Gr_n</math>. It's not difficult to prove that this map <math>\alpha</math> is actually an isomorphism (see Sections 3.5 and 3.6 in Husemoller, for example). As a result, <math>Gr_n</math> is called the [[classifying space]] of real ''n''-bundles.
Let <math>\scriptstyle X</math> be a paracompact space, and let <math>\scriptstyle Vect_n(X)</math> denote the set of real vector bundles over X of dimension n for some fixed positive integer <math>\scriptstyle n</math>. For any vector space V, let <math>\scriptstyle Gr_n(V)</math> denote the [[Grassmannian]] <math>\scriptstyle Gr_n(V) = \{W\subset V:\, \dim W = n\}</math>. Set <math>\scriptstyle Gr_n = Gr_n(\R^\infty)</math>. Define the [[tautological bundle]] <math>\scriptstyle \gamma^n \to Gr_n</math> by <math>\scriptstyle \gamma^n = \{(W, x):\, W\in Gr_n, x\in W\}</math>; this is a real bundle of dimension n, with projection <math>\scriptstyle \gamma^n \to Gr_n</math> given by <math>\scriptstyle (W, x) \to W</math>. For any map <math>\scriptstyle f:X \to Gr_n</math>, the induced bundle <math>\scriptstyle f^*\gamma^n \in Vect_n(X)</math>. Since any two homotopic maps <math>\scriptstyle f, g: X \to Gr_n</math> have <math>\scriptstyle f^*\gamma^n</math> and <math>\scriptstyle g^*\gamma^n</math> isomorphic, the map <math>\scriptstyle \alpha:[X; Gr_n] \to Vect_n(X)</math> given by <math>\scriptstyle f \to f^* \gamma^n</math> is well-defined, where <math>\scriptstyle [X; Gr_n]</math> denotes the set of homotopy equivalence classes of maps <math>\scriptstyle X \to Gr_n</math>. It's not difficult to prove that this map <math>\scriptstyle \alpha</math> is actually an isomorphism (see Sections 3.5 and 3.6 in Husemoller, for example). As a result, <math>\scriptstyle Gr_n</math> is called the [[classifying space]] of real ''n''-bundles.


Now consider the space <math>Vect_1(X)</math> of line bundles over <math>X</math>. For <math>n = 1</math>, the Grassmannian <math>Gr_1</math> is just <math>\R P^\infty = \R^\infty/\R^* = S^\infty/(\Z/2\Z)</math>, where the nonzero element of <math>\Z/2\Z</math> acts by <math>x \to -x</math>. The quotient map <math>S^\infty \to S^\infty/(\Z/2\Z) = \R P^\infty</math> is therefore a double cover. Since <math>S^\infty</math> is contractible, we have <math>\pi_i(\R P^\infty) = \pi_i(S^\infty) = 0</math> for <math>i > 1</math> and <math>\#\pi_1(\R P^\infty) = 2</math>; that is, <math>\pi_1(\R P^\infty) = \Z/2\Z</math>. Hence <math>\R P^\infty</math> is the Eilenberg-Maclane space <math>K(\Z/2\Z, 1)</math>. Hence <math>[X; Gr_1] = H^1(X; \Z/2\Z)</math> for any <math>X</math>, with the isomorphism given by <math>f \to f^* \eta</math>, where <math>\eta</math> is the generator <math>H^1(\R P^\infty; \Z/2\Z) = \Z/2\Z</math>. Since <math>\alpha:[X, Gr_1] \to Vect_1(X)</math> is also a bijection, we have another bijection <math>w_1:Vect_1 \to H^1(X; \Z/2\Z)</math>. This map <math>w_1</math> is precisely the Stiefel-Whitney class <math>w_1</math> for a line bundle. (Since the corresponding classifying space <math>C P^\infty</math> for complex bundles is a <math>K(\Z, 2)</math>, the same argument shows that the Chern class defines a bijection between complex line bundles over <math>X</math> and <math>H^2(X; \Z)</math>.) For example, since <math>H^1(S^1; \Z/2\Z) = \Z/2\Z</math>, there are only two line bundles over the circle up to bundle isomorphism: the trivial one, and the open Möbius strip (i.e., the Möbius strip with its boundary deleted). If <math>Vect_1(X)</math> is considered as a group under the operation of tensor product, then <math>\alpha</math> is an isomorphism: <math>w_1(\lambda \otimes \mu) = w_1(\lambda) + w_1(\mu)</math> for all line bundles <math>\lambda, \mu \to X</math>.
Now consider the space <math>\scriptstyle Vect_1(X)</math> of line bundles over <math>\scriptstyle X</math>. For <math>\scriptstyle n = 1</math>, the Grassmannian <math>\scriptstyle Gr_1</math> is just <math>\scriptstyle \R P^\infty = \R^\infty/\R^* = S^\infty/(\Z/2\Z)</math>, where the nonzero element of <math>\scriptstyle \Z/2\Z</math> acts by <math>\scriptstyle x \to -x</math>. The quotient map <math>\scriptstyle S^\infty \to S^\infty/(\Z/2\Z) = \R P^\infty</math> is therefore a double cover. Since <math>\scriptstyle S^\infty</math> is contractible, we have <math>\scriptstyle \pi_i(\R P^\infty) = \pi_i(S^\infty) = 0</math> for <math>\scriptstyle i > 1</math> and <math>\scriptstyle \#\pi_1(\R P^\infty) = 2</math>; that is, <math>\scriptstyle \pi_1(\R P^\infty) = \Z/2\Z</math>. Hence <math>\scriptstyle \R P^\infty</math> is the Eilenberg-Maclane space <math>\scriptstyle K(\Z/2\Z, 1)</math>. Hence <math>\scriptstyle [X; Gr_1] = H^1(X; \Z/2\Z)</math> for any <math>\scriptstyle X</math>, with the isomorphism given by <math>\scriptstyle f \to f^* \eta</math>, where <math>\scriptstyle \eta</math> is the generator <math>\scriptstyle H^1(\R P^\infty; \Z/2\Z) = \Z/2\Z</math>. Since <math>\scriptstyle \alpha:[X, Gr_1] \to Vect_1(X)</math> is also a bijection, we have another bijection <math>\scriptstyle w_1:Vect_1 \to H^1(X; \Z/2\Z)</math>. This map <math>\scriptstyle w_1</math> is precisely the Stiefel-Whitney class <math>\scriptstyle w_1</math> for a line bundle. (Since the corresponding classifying space <math>\scriptstyle C P^\infty</math> for complex bundles is a <math>\scriptstyle K(\Z, 2)</math>, the same argument shows that the Chern class defines a bijection between complex line bundles over <math>\scriptstyle X</math> and <math>\scriptstyle H^2(X; \Z)</math>.) For example, since <math>\scriptstyle H^1(S^1; \Z/2\Z) = \Z/2\Z</math>, there are only two line bundles over the circle up to bundle isomorphism: the trivial one, and the open Möbius strip (i.e., the Möbius strip with its boundary deleted). If <math>\scriptstyle Vect_1(X)</math> is considered as a group under the operation of tensor product, then <math>\scriptstyle \alpha</math> is an isomorphism: <math>\scriptstyle w_1(\lambda \otimes \mu) = w_1(\lambda) + w_1(\mu)</math> for all line bundles <math>\scriptstyle \lambda, \mu \to X</math>.


==Higher dimensions==
==Higher dimensions==


The bijection above for line bundles implies that any functor <math>\theta</math> satisfying the four axioms above is equal to ''w''. Let <math>\xi \to X</math> be an n-bundle. Then <math>\xi</math> admits a [[splitting map]], a map <math>f:X' \to X</math> for some space <math>X'</math> such that <math>f^*:H^*(X; \Z/2\Z) \to H^*(X'; \Z/2\Z)</math> is injective and <math>f^*\xi = \lambda_1 \oplus \cdots \oplus \lambda_n</math> for some line bundles <math>\lambda_i \to X'</math>. Any line bundle over ''X'' is of the form <math>g^* \gamma^1</math> for some map ''g'', and <math>\theta(g^*\gamma^1) = g^* \theta(\gamma^1) = 1 + w_1(g^*\gamma_1)</math> by naturality. Thus <math>\theta = w</math> on <math>Vect_1(X)</math>. It follows from the fourth axiom above that
The bijection above for line bundles implies that any functor <math>\scriptstyle \scriptstyle \theta</math> satisfying the four axioms above is equal to ''w''. Let <math>\scriptstyle \scriptstyle \xi \to X</math> be an n-bundle. Then <math>\scriptstyle \scriptstyle \xi</math> admits a [[splitting map]], a map <math>\scriptstyle \scriptstyle f:X' \to X</math> for some space <math>\scriptstyle \scriptstyle X'</math> such that <math>\scriptstyle \scriptstyle f^*:H^*(X; \Z/2\Z) \to H^*(X'; \Z/2\Z)</math> is injective and <math>\scriptstyle \scriptstyle f^*\xi = \lambda_1 \oplus \cdots \oplus \lambda_n</math> for some line bundles <math>\scriptstyle \scriptstyle \lambda_i \to X'</math>. Any line bundle over ''X'' is of the form <math>\scriptstyle \scriptstyle g^* \gamma^1</math> for some map ''g'', and <math>\scriptstyle \scriptstyle \theta(g^*\gamma^1) = g^* \theta(\gamma^1) = 1 + w_1(g^*\gamma_1)</math> by naturality. Thus <math>\scriptstyle \scriptstyle \theta = w</math> on <math>\scriptstyle \scriptstyle Vect_1(X)</math>. It follows from the fourth axiom above that
:<math>\scriptstyle f^*\theta(\xi) = \theta(f^*\xi) = \theta(\lambda_1 \oplus \cdots \oplus \lambda_n) = \theta(\lambda_1) \cdots \theta(\lambda_n) = (1 + w_1(\lambda_1)) \cdots (1 + w_1(\lambda_n)) = w(\lambda_1) \cdots w(\lambda_n) = w(f^*\xi) = f^* w(\xi).</math>
:<math> \scriptstyle f^*\theta(\xi) = \theta(f^*\xi) = \theta(\lambda_1 \oplus \cdots \oplus \lambda_n) = \theta(\lambda_1) \cdots \theta(\lambda_n) = (1 + w_1(\lambda_1)) \cdots (1 + w_1(\lambda_n)) = w(\lambda_1) \cdots w(\lambda_n) = w(f^*\xi) = f^* w(\xi).</math>
Since <math>f^*</math> is injective, <math>\theta = w</math> Thus the Stiefel-Whitney class is the unique functor satisfying the four axioms above.
Since <math>\scriptstyle \scriptstyle f^*</math> is injective, <math>\scriptstyle \scriptstyle \theta = w</math> Thus the Stiefel-Whitney class is the unique functor satisfying the four axioms above.


Although the map <math>w_1:Vect_1(X) \to H^1(X; \Z/2\Z)</math> is a bijection, the corresponding map is not necessarily injective in higher dimensions. For example, consider the tangent bundle <math>TS^n</math> for <math>n</math> even. With the canonical embedding of <math>S^n</math> in <math>\R^{n+1}</math>, the normal bundle <math>\nu</math> to <math>S^n</math> is a line bundle. Since <math>S^n</math> is orientable, <math>\nu</math> is trivial. The sum <math>TS^n \oplus \nu</math> is just the restriction of <math>T\R^{n+1}</math> to <math>S^n</math>, which is trivial since <math>\R^{n+1}</math> is contractible. Hence <math>w(TS^n) = w(TS^n)w(\nu) = w(TS^n \oplus \nu) = 1</math>. But <math>TS^n \to S^n</math> is not trivial; its [[Euler class]] <math>e(TS^n) = \chi(TS^n)[S^n] = 2[S^n] \not =0</math>, where <math>[S^n]</math> denotes a [[fundamental class]] of <math>S^n</math> and <math>\chi</math> the [[Euler characteristic]].
Although the map <math>\scriptstyle \scriptstyle w_1:Vect_1(X) \to H^1(X; \Z/2\Z)</math> is a bijection, the corresponding map is not necessarily injective in higher dimensions. For example, consider the tangent bundle <math>\scriptstyle \scriptstyle TS^n</math> for <math>\scriptstyle \scriptstyle n</math> even. With the canonical embedding of <math>\scriptstyle \scriptstyle S^n</math> in <math>\scriptstyle \scriptstyle \R^{n+1}</math>, the normal bundle <math>\scriptstyle \scriptstyle \nu</math> to <math>\scriptstyle \scriptstyle S^n</math> is a line bundle. Since <math>\scriptstyle \scriptstyle S^n</math> is orientable, <math>\scriptstyle \scriptstyle \nu</math> is trivial. The sum <math>\scriptstyle \scriptstyle TS^n \oplus \nu</math> is just the restriction of <math>\scriptstyle \scriptstyle T\R^{n+1}</math> to <math>\scriptstyle \scriptstyle S^n</math>, which is trivial since <math>\scriptstyle \scriptstyle \R^{n+1}</math> is contractible. Hence <math>\scriptstyle \scriptstyle w(TS^n) = w(TS^n)w(\nu) = w(TS^n \oplus \nu) = 1</math>. But <math>\scriptstyle \scriptstyle TS^n \to S^n</math> is not trivial; its [[Euler class]] <math>\scriptstyle \scriptstyle e(TS^n) = \chi(TS^n)[S^n] = 2[S^n] \not =0</math>, where <math>\scriptstyle \scriptstyle [S^n]</math> denotes a [[fundamental class]] of <math>\scriptstyle \scriptstyle S^n</math> and <math>\scriptstyle \scriptstyle \chi</math> the [[Euler characteristic]].


==Stiefel–Whitney numbers==
==Stiefel–Whitney numbers==


If we work on a manifold of dimension ''n'', then any product of Stiefel-Whitney classes of total degree ''n''
If we work on a manifold of dimension ''n'', then any product of Stiefel-Whitney classes of total degree ''n''
can be paired with the <math>\scriptstyle \mathbb{Z}_2</math>-[[fundamental class]] of the manifold to give an element of
can be paired with the <math>\scriptstyle \scriptstyle \scriptstyle \mathbb{Z}_2</math>-[[fundamental class]] of the manifold to give an element of
<math>\scriptstyle \mathbb{Z}_2</math>, a '''Stiefel-Whitney number''' of the vector bundle. For example, if the manifold has dimension 3, there are three linearly independent Stiefel-Whitney numbers, given by <math>\scriptstyle w_1^3, w_1 w_2, w_3</math>.
<math>\scriptstyle \scriptstyle \scriptstyle \mathbb{Z}_2</math>, a '''Stiefel-Whitney number''' of the vector bundle. For example, if the manifold has dimension 3, there are three linearly independent Stiefel-Whitney numbers, given by <math>\scriptstyle \scriptstyle \scriptstyle w_1^3, w_1 w_2, w_3</math>.
In general, if the manifold has dimension ''n'', the number of possible independent Stiefel-Whitney numbers is the number of [[integer partition|partition]]s of ''n''.
In general, if the manifold has dimension ''n'', the number of possible independent Stiefel-Whitney numbers is the number of [[integer partition|partition]]s of ''n''.


The Stiefel–Whitney numbers of the tangent bundle of a smooth manifold are called the Stiefel–Whitney numbers of the manifold. They are known to be [[cobordism]] invariants.
The Stiefel–Whitney numbers of the tangent bundle of a smooth manifold are called the Stiefel–Whitney numbers of the manifold. They are known to be [[cobordism]] invariants.


One Stiefel–Whitney number of important in [[surgery theory]] is the ''[[de Rham invariant]]'' of a (4''k''+1)-dimensional manifold, <math>w_2w_{4k-1}.</math>
One Stiefel–Whitney number of important in [[surgery theory]] is the ''[[de Rham invariant]]'' of a (4''k''+1)-dimensional manifold, <math>\scriptstyle w_2w_{4k-1}.</math>


==Wu classes==
==Wu classes==
The Stiefel–Whitney classes <math>w_k</math> are the [[Steenrod square]]s of the '''Wu classes''' <math>v_k,</math> defined by [[Wu Wenjun]] in {{Harv|Wu|1955}}. Most simply, the total Stiefel–Whitney class is the total Steenrod square of the total Wu class: <math>Sq(v)=w.</math> Wu classes are most often defined implicitly in terms of Steenrod squares, as the cohomology class representing the Steenrod squares: <math>v_k \cup x = Sq^k(x),</math> or more narrowly <math>\langle v_k \cup x, \mu\rangle = \langle Sq^k(x), \mu \rangle;</math> {{Harv |Milnor |Stasheff |1974 | loc = pp. 131-133}}.
The Stiefel–Whitney classes <math>\scriptstyle w_k</math> are the [[Steenrod square]]s of the '''Wu classes''' <math>\scriptstyle v_k,</math> defined by [[Wu Wenjun]] in {{Harv|Wu|1955}}. Most simply, the total Stiefel–Whitney class is the total Steenrod square of the total Wu class: <math>\scriptstyle Sq(v)=w.</math> Wu classes are most often defined implicitly in terms of Steenrod squares, as the cohomology class representing the Steenrod squares: <math>\scriptstyle v_k \cup x = Sq^k(x),</math> or more narrowly <math>\scriptstyle \langle v_k \cup x, \mu\rangle = \langle Sq^k(x), \mu \rangle;</math> {{Harv |Milnor |Stasheff |1974 | loc = pp. 131-133}}.


==Properties==
==Properties==
# If <math>E^k</math> has <math>s_1,\ldots,s_{\ell}</math> [[fiber bundle#Sections|sections]] which are everywhere [[linearly independent]] then <math>w_{k-\ell+1}=\cdots=w_k=0</math>.
# If <math>\scriptstyle E^k</math> has <math>\scriptstyle s_1,\ldots,s_{\ell}</math> [[fiber bundle#Sections|sections]] which are everywhere [[linearly independent]] then <math>\scriptstyle w_{k-\ell+1}=\cdots=w_k=0</math>.
# <math>w_ i(E)=0</math> whenever <math>i>\mathrm{rank}(E)</math>.
# <math>\scriptstyle w_ i(E)=0</math> whenever <math>\scriptstyle i>\mathrm{rank}(E)</math>.
#The first Stiefel-Whitney class is zero if and only if the bundle is [[orientability|orientable]]. In particular, a manifold ''M'' is orientable if and only if <math>w_1(TM) = 0</math>.
#The first Stiefel-Whitney class is zero if and only if the bundle is [[orientability|orientable]]. In particular, a manifold ''M'' is orientable if and only if <math>\scriptstyle w_1(TM) = 0</math>.
#The bundle admits a [[spin structure]] if and only if both the first and second Stiefel-Whitney classes are zero.
#The bundle admits a [[spin structure]] if and only if both the first and second Stiefel-Whitney classes are zero.
#For an orientable bundle, the second Stiefel-Whitney class is in the image of the natural map <math>\scriptstyle H^2(M, \Z) \rightarrow H^2(M,\Z/2\Z)</math> (equivalently, the so-called third '''integral''' Stiefel-Whitney class is zero) if and only if the bundle admits a [[spin structure|spin<sup>c</sup> structure]].
#For an orientable bundle, the second Stiefel-Whitney class is in the image of the natural map <math>\scriptstyle \scriptstyle H^2(M, \Z) \rightarrow H^2(M,\Z/2\Z)</math> (equivalently, the so-called third '''integral''' Stiefel-Whitney class is zero) if and only if the bundle admits a [[spin structure|spin<sup>c</sup> structure]].
#All the Stiefel-Whitney numbers of a smooth compact manifold ''X'' vanish if and only if the manifold is a boundary (unoriented) of a smooth compact manifold.
#All the Stiefel-Whitney numbers of a smooth compact manifold ''X'' vanish if and only if the manifold is a boundary (unoriented) of a smooth compact manifold.


==Integral Stiefel-Whitney classes==
==Integral Stiefel-Whitney classes==
The element <math>\beta w_i \in H^{i+1}(X;\mathbb{Z})</math> is called the <math>i+1</math> ''integral'' Stiefel-Whitney class, where β is the [[Bockstein homomorphism]], corresponding to reduction modulo 2, <math>\mathbb{Z} \to \mathbb{Z}/2</math>:
The element <math>\scriptstyle \beta w_i \in H^{i+1}(X;\mathbb{Z})</math> is called the <math>\scriptstyle i+1</math> ''integral'' Stiefel-Whitney class, where β is the [[Bockstein homomorphism]], corresponding to reduction modulo 2, <math>\scriptstyle \mathbb{Z} \to \mathbb{Z}/2</math>:
:<math>\beta\colon H^i(X;\mathbb{Z}/2) \to H^{i+1}(X;\mathbb{Z}).</math>
:<math>\scriptstyle \beta\colon H^i(X;\mathbb{Z}/2) \to H^{i+1}(X;\mathbb{Z}).</math>


For instance, the third integral Stiefel-Whitney class is the obstruction to a [[Spin_structure#SpinC_structures|Spin<sup>c</sup> structure]].
For instance, the third integral Stiefel-Whitney class is the obstruction to a [[Spin_structure#SpinC_structures|Spin<sup>c</sup> structure]].


==Relations over the Steenrod algebra==
==Relations over the Steenrod algebra==
Over the [[Steenrod algebra]], the Stiefel-Whitney classes <math>\{w_{2^i}\}</math> generate all the Stiefel-Whitney classes. In particular, the Stiefel-Whitney classes satisfy the '''{{visible anchor|Wu formula}}''', named for [[Wu Wenjun]]:<ref>{{Harv|May|1999|p=197}}</ref>
Over the [[Steenrod algebra]], the Stiefel-Whitney classes <math>\scriptstyle \{w_{2^i}\}</math> generate all the Stiefel-Whitney classes. In particular, the Stiefel-Whitney classes satisfy the '''{{visible anchor|Wu formula}}''', named for [[Wu Wenjun]]:<ref>{{Harv|May|1999|p=197}}</ref>
:<math>Sq^i(w_j)=\sum_{t=0}^i {j+t-i-1 \choose t} w_{i-t}w_{j+t}.</math>
:<math>\scriptstyle Sq^i(w_j)=\sum_{t=0}^i {j+t-i-1 \choose t} w_{i-t}w_{j+t}.</math>


==References==
==References==

Revision as of 19:49, 9 July 2010

In mathematics, the Stiefel–Whitney class arises as a type of characteristic class associated to real vector bundles . It is denoted by w(E), taking values in , the cohomology groups with mod 2 coefficients. The component of in is denoted by and called the -th Stiefel-Whitney class of , so that . As an example, over the circle, , there is a line bundle that is topologically non-trivial: that is, the line bundle associated to the Möbius band, usually thought of as having fibres . The cohomology group

has just one element other than 0, this element being the first Stiefel-Whitney class, , of that line bundle.

Origins

The Stiefel-Whitney classes get their name because Stiefel and Whitney discovered them as mod-2 reductions of the obstruction classes to constructing everywhere linearly independent sections of the vector bundle restricted to the -skeleton of . Here denotes the dimension of the fibre of the vector bundle .

To be precise, provided is a CW-complex, Whitney defined classes in the -th cellular cohomology group of with twisted coefficients. The coefficient system being the -st homotopy group of the Stiefel manifold of linearly independent vectors in the fibres of . Whitney proved if and only if , when restricted to the -skeleton of , has linearly-independent sections.

Since is either infinite-cyclic or isomorphic to , there is a canonical reduction of the classes to classes which are the Stiefel-Whitney classes. Moreover, whenever , the two classes are identical. Thus, if and only if the bundle is orientable.

The class is exceptional and has no meaning a priori. Its creation by Whitney was an act of creative notation, allowing the Whitney sum Formula to be true. However, for generalizations of manifolds (namely certain homology manifolds), one can have – it only needs to equal 1 mod 8.

Axioms

Throughout, denotes singular cohomology of a space with coefficients in the group .

  1. Naturality: for any bundle and map , where denotes the induced bundle.
  2. in .
  3. is the generator of (normalization condition). Here, is the canonical line bundle.
  4. (Whitney product formula).

Some work is required to show that such classes do indeed exist and are unique (at least for paracompact spaces X); see section 17.2 and 17.3 in Husemoller or section 8 in Milnor and Stasheff.

Line bundles

Let be a paracompact space, and let denote the set of real vector bundles over X of dimension n for some fixed positive integer . For any vector space V, let denote the Grassmannian . Set . Define the tautological bundle by ; this is a real bundle of dimension n, with projection given by . For any map , the induced bundle . Since any two homotopic maps have and isomorphic, the map given by is well-defined, where denotes the set of homotopy equivalence classes of maps . It's not difficult to prove that this map is actually an isomorphism (see Sections 3.5 and 3.6 in Husemoller, for example). As a result, is called the classifying space of real n-bundles.

Now consider the space of line bundles over . For , the Grassmannian is just , where the nonzero element of acts by . The quotient map is therefore a double cover. Since is contractible, we have for and ; that is, . Hence is the Eilenberg-Maclane space . Hence for any , with the isomorphism given by , where is the generator . Since is also a bijection, we have another bijection . This map is precisely the Stiefel-Whitney class for a line bundle. (Since the corresponding classifying space for complex bundles is a , the same argument shows that the Chern class defines a bijection between complex line bundles over and .) For example, since , there are only two line bundles over the circle up to bundle isomorphism: the trivial one, and the open Möbius strip (i.e., the Möbius strip with its boundary deleted). If is considered as a group under the operation of tensor product, then is an isomorphism: for all line bundles .

Higher dimensions

The bijection above for line bundles implies that any functor satisfying the four axioms above is equal to w. Let be an n-bundle. Then admits a splitting map, a map for some space such that is injective and for some line bundles . Any line bundle over X is of the form for some map g, and by naturality. Thus on . It follows from the fourth axiom above that

Since is injective, Thus the Stiefel-Whitney class is the unique functor satisfying the four axioms above.

Although the map is a bijection, the corresponding map is not necessarily injective in higher dimensions. For example, consider the tangent bundle for even. With the canonical embedding of in , the normal bundle to is a line bundle. Since is orientable, is trivial. The sum is just the restriction of to , which is trivial since is contractible. Hence . But is not trivial; its Euler class , where denotes a fundamental class of and the Euler characteristic.

Stiefel–Whitney numbers

If we work on a manifold of dimension n, then any product of Stiefel-Whitney classes of total degree n can be paired with the -fundamental class of the manifold to give an element of , a Stiefel-Whitney number of the vector bundle. For example, if the manifold has dimension 3, there are three linearly independent Stiefel-Whitney numbers, given by . In general, if the manifold has dimension n, the number of possible independent Stiefel-Whitney numbers is the number of partitions of n.

The Stiefel–Whitney numbers of the tangent bundle of a smooth manifold are called the Stiefel–Whitney numbers of the manifold. They are known to be cobordism invariants.

One Stiefel–Whitney number of important in surgery theory is the de Rham invariant of a (4k+1)-dimensional manifold,

Wu classes

The Stiefel–Whitney classes are the Steenrod squares of the Wu classes defined by Wu Wenjun in (Wu 1955). Most simply, the total Stiefel–Whitney class is the total Steenrod square of the total Wu class: Wu classes are most often defined implicitly in terms of Steenrod squares, as the cohomology class representing the Steenrod squares: or more narrowly (Milnor & Stasheff 1974, pp. 131-133).

Properties

  1. If has sections which are everywhere linearly independent then .
  2. whenever .
  3. The first Stiefel-Whitney class is zero if and only if the bundle is orientable. In particular, a manifold M is orientable if and only if .
  4. The bundle admits a spin structure if and only if both the first and second Stiefel-Whitney classes are zero.
  5. For an orientable bundle, the second Stiefel-Whitney class is in the image of the natural map (equivalently, the so-called third integral Stiefel-Whitney class is zero) if and only if the bundle admits a spinc structure.
  6. All the Stiefel-Whitney numbers of a smooth compact manifold X vanish if and only if the manifold is a boundary (unoriented) of a smooth compact manifold.

Integral Stiefel-Whitney classes

The element is called the integral Stiefel-Whitney class, where β is the Bockstein homomorphism, corresponding to reduction modulo 2, :

For instance, the third integral Stiefel-Whitney class is the obstruction to a Spinc structure.

Relations over the Steenrod algebra

Over the Steenrod algebra, the Stiefel-Whitney classes generate all the Stiefel-Whitney classes. In particular, the Stiefel-Whitney classes satisfy the Wu formula, named for Wu Wenjun:[1]

References

  1. ^ (May 1999, p. 197)
  • D. Husemoller, Fibre Bundles, Springer-Verlag, 1994.
  • J. Milnor & J. Stasheff, Characteristic Classes, Princeton, 1974.
  • May, J. P. (1999), A Concise Course in Algebraic Topology (PDF), U. Chicago Press, Chicago, retrieved 2009-08-07