Namespaces
Variants
Actions

Difference between revisions of "Obstruction"

From Encyclopedia of Mathematics
Jump to: navigation, search
(Importing text file)
 
m (tex encoded by computer)
 
Line 1: Line 1:
 +
<!--
 +
o0680701.png
 +
$#A+1 = 139 n = 0
 +
$#C+1 = 139 : ~/encyclopedia/old_files/data/O068/O.0608070 Obstruction
 +
Automatically converted into TeX, above some diagnostics.
 +
Please remove this comment and the {{TEX|auto}} line below,
 +
if TeX found to be correct.
 +
-->
 +
 +
{{TEX|auto}}
 +
{{TEX|done}}
 +
 
A concept in homotopy theory: An invariant that equals zero if a (step in a) corresponding problem is solvable and is non-zero otherwise.
 
A concept in homotopy theory: An invariant that equals zero if a (step in a) corresponding problem is solvable and is non-zero otherwise.
  
Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680701.png" /> be a pair of cellular spaces (cf. [[Cellular space|Cellular space]]) and let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680702.png" /> be a simply-connected (more generally, a homotopy-simple) topological space. Can one extend a given continuous mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680703.png" /> to a continuous mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680704.png" />? The extension can be attempted recursively, over successive skeletons <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680705.png" /> of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680706.png" />. Suppose one has constructed a mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680707.png" /> such that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680708.png" />. For any oriented <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o0680709.png" />-dimensional cell <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807010.png" /> the mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807011.png" /> gives a mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807012.png" /> (where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807013.png" /> is the <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807014.png" />-dimensional unit sphere) and an element <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807015.png" /> (it is here that one uses that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807016.png" /> is homotopy simple, which allows one to ignore the base point). This defines a cochain
+
Let $  ( X, A) $
 +
be a pair of cellular spaces (cf. [[Cellular space|Cellular space]]) and let $  Y $
 +
be a simply-connected (more generally, a homotopy-simple) topological space. Can one extend a given continuous mapping $  g: A \rightarrow Y $
 +
to a continuous mapping $  f: X \rightarrow Y $?  
 +
The extension can be attempted recursively, over successive skeletons $  X  ^ {n} $
 +
of $  X $.  
 +
Suppose one has constructed a mapping $  f: X  ^ {n} \cup A \rightarrow Y $
 +
such that $  f  \mid  _ {A} = g $.  
 +
For any oriented $  ( n + 1) $-
 +
dimensional cell $  e  ^ {n+} 1 \rightarrow Y $
 +
the mapping $  f \mid  _ {\partial  e ^ {n + 1 }  } $
 +
gives a mapping $  S  ^ {n} \rightarrow Y $(
 +
where $  S  ^ {n} $
 +
is the $  n $-
 +
dimensional unit sphere) and an element $  \alpha _ {e} \in \pi _ {n} ( Y) $(
 +
it is here that one uses that $  Y $
 +
is homotopy simple, which allows one to ignore the base point). This defines a cochain
  
<table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807017.png" /></td> </tr></table>
+
$$
 +
c _ {f} ^ {n + 1 }  \in \
 +
C ^ {n + 1 } ( X; \pi _ {n} ( Y)),\ \
 +
c _ {f} ^ {n + 1 }
 +
( e ^ {n + 1 } )  = \
 +
\alpha _ {e} .
 +
$$
  
Since for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807018.png" /> one clearly has <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807019.png" />, it follows that
+
Since for $  e ^ {n + 1 } \subset  A $
 +
one clearly has $  c _ {f} ^ {n + 1 } ( e ^ {n + 1 } ) = 0 $,  
 +
it follows that
  
<table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807020.png" /></td> </tr></table>
+
$$
 +
c _ {f} ^ {n + 1 }  \in \
 +
C ^ {n + 1 } ( X, A; \pi _ {n} ( Y)).
 +
$$
  
Clearly <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807021.png" /> if and only if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807022.png" /> can be extended to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807023.png" />, i.e. <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807024.png" /> is an obstruction to extending <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807025.png" /> to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807026.png" />.
+
Clearly $  c _ {f} ^ {n + 1 } = 0 $
 +
if and only if $  f $
 +
can be extended to $  X ^ {n + 1 } $,  
 +
i.e. $  c _ {f} ^ {n + 1 } $
 +
is an obstruction to extending $  f $
 +
to $  X ^ {n + 1 } $.
  
The cochain <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807027.png" /> is a cocycle. The fact that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807028.png" /> does not, in general, imply that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807029.png" /> cannot be extended to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807030.png" />: It is possible that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807031.png" /> cannot be extended to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807032.png" /> because of an unsuccessful choice of an extension of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807033.png" /> to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807034.png" />. It may turn out that, e.g., the mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807035.png" /> can be extended to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807036.png" />, i.e. that extension is possible by skipping back one step. It can be shown that the cohomology class
+
The cochain $  c _ {f} ^ {n + 1 } \in C ^ {n + 1 } ( X, A;  \pi _ {n} ( Y)) $
 +
is a cocycle. The fact that $  c _ {f} ^ {n + 1 } \neq 0 $
 +
does not, in general, imply that $  g $
 +
cannot be extended to $  X $:  
 +
It is possible that $  f $
 +
cannot be extended to $  X ^ {n + 1 } $
 +
because of an unsuccessful choice of an extension of $  g $
 +
to $  X  ^ {n} $.  
 +
It may turn out that, e.g., the mapping $  f \mid  _ {X ^ {n - 1 }  \cup A } $
 +
can be extended to $  X ^ {n + 1 } $,  
 +
i.e. that extension is possible by skipping back one step. It can be shown that the cohomology class
  
<table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807037.png" /></td> </tr></table>
+
$$
 +
[ c _ {f} ^ {n + 1 } ]  \in \
 +
H ^ {n + 1 } ( X, A; \pi _ {n} ( Y))
 +
$$
  
is an obstruction to this, i.e. <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807038.png" /> if and only if there is a mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807039.png" /> such that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807040.png" /> (in particular, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807041.png" />). The construction of difference chains and cochains is used in the proof of this statement (cf. [[Difference cochain and chain|Difference cochain and chain]]).
+
is an obstruction to this, i.e. $  [ c _ {f} ^ {n + 1 } ] = 0 $
 +
if and only if there is a mapping $  \widetilde{f}  : X ^ {n + 1 } \cup A \rightarrow Y $
 +
such that $  \widetilde{f}  \mid  _ {X ^ {n - 1 }  \cup A } = f \mid  _ {X ^ {n - 1 }  \cup A } $(
 +
in particular, $  \widetilde{f}  \mid  _ {A} = g $).  
 +
The construction of difference chains and cochains is used in the proof of this statement (cf. [[Difference cochain and chain|Difference cochain and chain]]).
  
Since the problem of homotopy classification of mappings <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807042.png" /> can be interpreted as an extension problem, obstruction theory is applicable also to the description of the set <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807043.png" /> of homotopy classes of mappings from <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807044.png" /> into <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807045.png" />. Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807046.png" /> and let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807047.png" /> be a subspace of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807048.png" />. Then a pair of mappings <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807049.png" /> is interpreted as a mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807050.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807051.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807052.png" />, and the presence of a homotopy between <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807053.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807054.png" /> means the presence of a mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807055.png" /> extending <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807056.png" />. If the homotopy <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807057.png" /> has been constructed on the <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807058.png" />-dimensional skeleton of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807059.png" />, then the obstruction to its extension to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807060.png" /> is the difference cochain
+
Since the problem of homotopy classification of mappings $  X \rightarrow Y $
 +
can be interpreted as an extension problem, obstruction theory is applicable also to the description of the set $  [ X, Y] $
 +
of homotopy classes of mappings from $  X $
 +
into $  Y $.  
 +
Let $  I = [ 0, 1] $
 +
and let $  A = X \times \{ 0, 1 \} $
 +
be a subspace of $  X \times I $.  
 +
Then a pair of mappings $  f _ {0} , f _ {1} : X \rightarrow Y $
 +
is interpreted as a mapping $  G: A \rightarrow Y $,
 +
$  G ( x, i) = f _ {i} ( x) $,
 +
$  i = 0, 1 $,  
 +
and the presence of a homotopy between $  f _ {0} $
 +
and $  f _ {1} $
 +
means the presence of a mapping $  F: X \times I \rightarrow Y $
 +
extending $  G $.  
 +
If the homotopy $  F $
 +
has been constructed on the $  n $-
 +
dimensional skeleton of $  X $,  
 +
then the obstruction to its extension to $  X $
 +
is the difference cochain
  
<table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807061.png" /></td> </tr></table>
+
$$
 +
d  ^ {n} ( f _ {0} , f _ {1} )  \in \
 +
C  ^ {n} ( X; \pi _ {n} ( Y)).
 +
$$
  
As an application one may consider the description of the set <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807062.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807063.png" />, where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807064.png" /> is the [[Eilenberg–MacLane space|Eilenberg–MacLane space]]: <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807065.png" /> for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807066.png" />; <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807067.png" />. Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807068.png" /> be a constant mapping and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807069.png" /> an arbitrary continuous mapping. Since <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807070.png" /> for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807071.png" />, the mappings <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807072.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807073.png" /> are homotopic on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807074.png" /> and, after having chosen such a homotopy, one can define the difference cochain
+
As an application one may consider the description of the set $  [ X, Y] = [ X, K ( \pi , n)] $,  
 +
$  n > 1 $,  
 +
where $  K ( \pi , n) $
 +
is the [[Eilenberg–MacLane space|Eilenberg–MacLane space]]: $  \pi _ {i} ( K ( \pi , n)) = 0 $
 +
for $  i \neq n $;  
 +
$  \pi _ {n} ( K ( \pi , n)) = \pi $.  
 +
Let $  f _ {0} : X \rightarrow K ( \pi , n) $
 +
be a constant mapping and $  f: X \rightarrow K ( \pi , n) $
 +
an arbitrary continuous mapping. Since $  H  ^ {i} ( X;  \pi _ {i} ( Y)) = 0 $
 +
for $  i < n $,  
 +
the mappings $  f _ {0} $
 +
and $  f $
 +
are homotopic on $  X ^ {n - 1 } $
 +
and, after having chosen such a homotopy, one can define the difference cochain
  
<table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807075.png" /></td> </tr></table>
+
$$
 +
d  ^ {n} ( f, f _ {0} )  \in \
 +
C  ^ {n} ( X; \pi _ {n} ( Y))  = \
 +
C  ^ {n} ( X; \pi ).
 +
$$
  
The cohomology class <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807076.png" /> is well-defined, i.e. does not depend on the choice of a homotopy between <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807077.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807078.png" /> (since <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807079.png" /> for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807080.png" />). Further, if two mappings <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807081.png" /> are such that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807082.png" />, then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807083.png" />, and hence <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807084.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807085.png" /> are homotopic on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807086.png" />. The obstructions to extending this homotopy to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807087.png" /> lie in the groups <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807088.png" /> (since <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807089.png" />), and hence <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807090.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807091.png" /> are homotopic. Thus, the homotopy class of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807092.png" /> is completely determined by the element <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807093.png" />. Finally, for any <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807094.png" /> there is a mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807095.png" /> with <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807096.png" />, hence <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807097.png" />. Similarly, if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807098.png" /> for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o06807099.png" /> and if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070100.png" />, then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070101.png" />.
+
The cohomology class $  [ d  ^ {n} ( f, f _ {0} )] \in H  ^ {n} ( X;  \pi ) $
 +
is well-defined, i.e. does not depend on the choice of a homotopy between $  f _ {0} $
 +
and $  f $(
 +
since $  \pi _ {i} ( Y) = 0 $
 +
for $  i < n $).  
 +
Further, if two mappings $  f, g: X \rightarrow Y $
 +
are such that $  [ d  ^ {n} ( f, f _ {0} )] = [ d  ^ {n} ( g, f _ {0} )] $,  
 +
then $  [ d  ^ {n} ( f, g)] = 0 $,  
 +
and hence $  f $
 +
and $  g $
 +
are homotopic on $  X  ^ {n} $.  
 +
The obstructions to extending this homotopy to $  X $
 +
lie in the groups $  H  ^ {i} ( X;  \pi _ {i} ( Y)) = 0 $(
 +
since $  i > n $),  
 +
and hence $  f $
 +
and $  g $
 +
are homotopic. Thus, the homotopy class of $  f $
 +
is completely determined by the element $  [ d  ^ {n} ( f, f _ {0} )] \in H  ^ {n} ( X;  \pi ) $.  
 +
Finally, for any $  x \in H  ^ {n} ( X;  \pi ) $
 +
there is a mapping $  f $
 +
with $  [ d  ^ {n} ( f, f _ {0} )] = x $,  
 +
hence $  [ X, K ( \pi , n)] = H  ^ {n} ( X;  \pi ) $.  
 +
Similarly, if $  \pi _ {i} ( Y) = i $
 +
for $  i < n $
 +
and if $  \mathop{\rm dim}  X \leq  n $,  
 +
then $  [ X, Y] = H  ^ {n} ( X;  \pi _ {n} ( Y)) $.
  
In studying extension problems one has considered the possibility of extending  "by skipping back one step" . A complete solution of the problem requires the analysis of the possibility of skipping back an arbitrary number of steps. Cohomology operations (cf. [[Cohomology operation|Cohomology operation]]) and Postnikov systems (cf. [[Postnikov system|Postnikov system]]) are used to this end. E.g., in order to describe the set <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070102.png" />, where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070103.png" /> for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070104.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070105.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070106.png" />, it is required, in general, to study the possibility of skipping back <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070107.png" /> steps, for which it is necessary to study the first <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070108.png" /> levels of the Postnikov system for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070109.png" />, i.e. to use cohomology operations of orders <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070110.png" /> (in the article [[Cohomology operation|Cohomology operation]] this problem is outlined for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070111.png" />).
+
In studying extension problems one has considered the possibility of extending  "by skipping back one step" . A complete solution of the problem requires the analysis of the possibility of skipping back an arbitrary number of steps. Cohomology operations (cf. [[Cohomology operation|Cohomology operation]]) and Postnikov systems (cf. [[Postnikov system|Postnikov system]]) are used to this end. E.g., in order to describe the set $  [ X, Y] $,  
 +
where $  \pi _ {i} ( Y) = 0 $
 +
for $  i < n $,  
 +
$  \pi _ {n} ( Y) \neq 0 $,  
 +
$  \mathop{\rm dim}  X = n + r $,  
 +
it is required, in general, to study the possibility of skipping back $  r + 1 $
 +
steps, for which it is necessary to study the first $  n + r $
 +
levels of the Postnikov system for $  Y $,  
 +
i.e. to use cohomology operations of orders $  \leq  r $(
 +
in the article [[Cohomology operation|Cohomology operation]] this problem is outlined for $  r = 1 $).
  
The theory of obstructions is also used in the more general situation of extension of sections (cf. [[Section of a mapping|Section of a mapping]]). Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070112.png" /> be a fibration with fibre <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070113.png" /> (where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070114.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070115.png" /> acts trivially on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070116.png" />), let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070117.png" /> and let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070118.png" /> be a section (i.e. a continuous mapping such that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070119.png" />). Can one extend <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070120.png" /> to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070121.png" />? The corresponding obstructions lie in the groups <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070122.png" />. An extension problem is obtained from this problem if one puts <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070123.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070124.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070125.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070126.png" />. Analogously one can also study the classification problem for sections using obstruction theory.
+
The theory of obstructions is also used in the more general situation of extension of sections (cf. [[Section of a mapping|Section of a mapping]]). Let $  p: E \rightarrow B $
 +
be a fibration with fibre $  F $(
 +
where $  \pi _ {1} ( F  ) = 0 $
 +
and $  \pi _ {1} ( B) $
 +
acts trivially on $  \pi _ {i} ( F  ) $),  
 +
let $  A \subset  B $
 +
and let $  s: A \rightarrow E $
 +
be a section (i.e. a continuous mapping such that $  ps ( a) = a $).  
 +
Can one extend $  s $
 +
to $  B $?  
 +
The corresponding obstructions lie in the groups $  H ^ {n + 1 } ( B;  \pi _ {n} ( F  )) $.  
 +
An extension problem is obtained from this problem if one puts $  B = X $,
 +
$  E = X \times Y $,
 +
$  p ( x, y) = x $,  
 +
$  s ( a) = ( a, g ( a)) $.  
 +
Analogously one can also study the classification problem for sections using obstruction theory.
  
Finally, one can remove the restriction of homotopic simplicity of the space <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070127.png" /> in the extension problem (as well as in the problem on sections); then one must use cohomology with local coefficients.
+
Finally, one can remove the restriction of homotopic simplicity of the space $  Y $
 +
in the extension problem (as well as in the problem on sections); then one must use cohomology with local coefficients.
  
 
Obstruction theory was initiated by S. Eilenberg [[#References|[2]]]. It was also known to L.S. Pontryagin, who did not formulate it explicitly but used it for the solution of concrete problems, see [[#References|[1]]].
 
Obstruction theory was initiated by S. Eilenberg [[#References|[2]]]. It was also known to L.S. Pontryagin, who did not formulate it explicitly but used it for the solution of concrete problems, see [[#References|[1]]].
Line 39: Line 189:
 
====References====
 
====References====
 
<table><TR><TD valign="top">[1]</TD> <TD valign="top">  L.S. Pontryagin,  "Classification of continuous transformations of a complex into a sphere"  ''Dokl. Akad. Nauk SSSR'' , '''19'''  (1938)  pp. 361–363  (In Russian)</TD></TR><TR><TD valign="top">[2]</TD> <TD valign="top">  S. Eilenberg,  "Cohomology and continuous mappings"  ''Ann. of Math.'' , '''41'''  (1940)  pp. 231–251</TD></TR><TR><TD valign="top">[3]</TD> <TD valign="top">  S.-T. Hu,  "Homotopy theory" , Acad. Press  (1959)</TD></TR><TR><TD valign="top">[4]</TD> <TD valign="top">  E. Thomas,  "Seminar on fibre spaces" , Springer  (1966)</TD></TR></table>
 
<table><TR><TD valign="top">[1]</TD> <TD valign="top">  L.S. Pontryagin,  "Classification of continuous transformations of a complex into a sphere"  ''Dokl. Akad. Nauk SSSR'' , '''19'''  (1938)  pp. 361–363  (In Russian)</TD></TR><TR><TD valign="top">[2]</TD> <TD valign="top">  S. Eilenberg,  "Cohomology and continuous mappings"  ''Ann. of Math.'' , '''41'''  (1940)  pp. 231–251</TD></TR><TR><TD valign="top">[3]</TD> <TD valign="top">  S.-T. Hu,  "Homotopy theory" , Acad. Press  (1959)</TD></TR><TR><TD valign="top">[4]</TD> <TD valign="top">  E. Thomas,  "Seminar on fibre spaces" , Springer  (1966)</TD></TR></table>
 
 
  
 
====Comments====
 
====Comments====
The fundamental group <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070128.png" /> acts on the homotopy groups <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070129.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070130.png" />, cf. [[Homotopy group|Homotopy group]]. The space <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070131.png" /> is called <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070133.png" />-simple if this action (for this <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070134.png" />) is trivial; <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070135.png" /> is called simple or homotopy simple if it is path connected and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070136.png" />-simple for all <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070137.png" />. Then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070138.png" /> is Abelian and acts trivially on all <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070139.png" />. A path-connected [[H-space|<img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/o/o068/o068070/o068070140.png" />-space]] is simple.
+
The fundamental group $  \pi _ {1} ( X , x _ {0} ) $
 +
acts on the homotopy groups $  \pi _ {n} ( X , x _ {0} ) $,  
 +
$  n \geq  1 $,  
 +
cf. [[Homotopy group|Homotopy group]]. The space $  X $
 +
is called $  n $-
 +
simple if this action (for this $  n $)  
 +
is trivial; $  X $
 +
is called simple or homotopy simple if it is path connected and $  n $-
 +
simple for all $  n \geq  1 $.  
 +
Then $  \pi _ {1} ( X , x _ {0} ) $
 +
is Abelian and acts trivially on all $  \pi _ {n} ( X , x _ {0} ) $.  
 +
A path-connected [[H-space| $  H $-
 +
space]] is simple.
  
 
====References====
 
====References====
 
<table><TR><TD valign="top">[a1]</TD> <TD valign="top">  N.E. Steenrod,  "The topology of fibre bundles" , Princeton Univ. Press  (1951)</TD></TR><TR><TD valign="top">[a2]</TD> <TD valign="top">  E.H. Spanier,  "Algebraic topology" , McGraw-Hill  (1966)  pp. 269–276; 429–432</TD></TR><TR><TD valign="top">[a3]</TD> <TD valign="top">  G.W. Whitehead,  "Elements of homotopy theory" , Springer  (1978)</TD></TR><TR><TD valign="top">[a4]</TD> <TD valign="top">  H.J. Baues,  "Obstruction theory" , Springer  (1977)</TD></TR></table>
 
<table><TR><TD valign="top">[a1]</TD> <TD valign="top">  N.E. Steenrod,  "The topology of fibre bundles" , Princeton Univ. Press  (1951)</TD></TR><TR><TD valign="top">[a2]</TD> <TD valign="top">  E.H. Spanier,  "Algebraic topology" , McGraw-Hill  (1966)  pp. 269–276; 429–432</TD></TR><TR><TD valign="top">[a3]</TD> <TD valign="top">  G.W. Whitehead,  "Elements of homotopy theory" , Springer  (1978)</TD></TR><TR><TD valign="top">[a4]</TD> <TD valign="top">  H.J. Baues,  "Obstruction theory" , Springer  (1977)</TD></TR></table>

Latest revision as of 08:03, 6 June 2020


A concept in homotopy theory: An invariant that equals zero if a (step in a) corresponding problem is solvable and is non-zero otherwise.

Let $ ( X, A) $ be a pair of cellular spaces (cf. Cellular space) and let $ Y $ be a simply-connected (more generally, a homotopy-simple) topological space. Can one extend a given continuous mapping $ g: A \rightarrow Y $ to a continuous mapping $ f: X \rightarrow Y $? The extension can be attempted recursively, over successive skeletons $ X ^ {n} $ of $ X $. Suppose one has constructed a mapping $ f: X ^ {n} \cup A \rightarrow Y $ such that $ f \mid _ {A} = g $. For any oriented $ ( n + 1) $- dimensional cell $ e ^ {n+} 1 \rightarrow Y $ the mapping $ f \mid _ {\partial e ^ {n + 1 } } $ gives a mapping $ S ^ {n} \rightarrow Y $( where $ S ^ {n} $ is the $ n $- dimensional unit sphere) and an element $ \alpha _ {e} \in \pi _ {n} ( Y) $( it is here that one uses that $ Y $ is homotopy simple, which allows one to ignore the base point). This defines a cochain

$$ c _ {f} ^ {n + 1 } \in \ C ^ {n + 1 } ( X; \pi _ {n} ( Y)),\ \ c _ {f} ^ {n + 1 } ( e ^ {n + 1 } ) = \ \alpha _ {e} . $$

Since for $ e ^ {n + 1 } \subset A $ one clearly has $ c _ {f} ^ {n + 1 } ( e ^ {n + 1 } ) = 0 $, it follows that

$$ c _ {f} ^ {n + 1 } \in \ C ^ {n + 1 } ( X, A; \pi _ {n} ( Y)). $$

Clearly $ c _ {f} ^ {n + 1 } = 0 $ if and only if $ f $ can be extended to $ X ^ {n + 1 } $, i.e. $ c _ {f} ^ {n + 1 } $ is an obstruction to extending $ f $ to $ X ^ {n + 1 } $.

The cochain $ c _ {f} ^ {n + 1 } \in C ^ {n + 1 } ( X, A; \pi _ {n} ( Y)) $ is a cocycle. The fact that $ c _ {f} ^ {n + 1 } \neq 0 $ does not, in general, imply that $ g $ cannot be extended to $ X $: It is possible that $ f $ cannot be extended to $ X ^ {n + 1 } $ because of an unsuccessful choice of an extension of $ g $ to $ X ^ {n} $. It may turn out that, e.g., the mapping $ f \mid _ {X ^ {n - 1 } \cup A } $ can be extended to $ X ^ {n + 1 } $, i.e. that extension is possible by skipping back one step. It can be shown that the cohomology class

$$ [ c _ {f} ^ {n + 1 } ] \in \ H ^ {n + 1 } ( X, A; \pi _ {n} ( Y)) $$

is an obstruction to this, i.e. $ [ c _ {f} ^ {n + 1 } ] = 0 $ if and only if there is a mapping $ \widetilde{f} : X ^ {n + 1 } \cup A \rightarrow Y $ such that $ \widetilde{f} \mid _ {X ^ {n - 1 } \cup A } = f \mid _ {X ^ {n - 1 } \cup A } $( in particular, $ \widetilde{f} \mid _ {A} = g $). The construction of difference chains and cochains is used in the proof of this statement (cf. Difference cochain and chain).

Since the problem of homotopy classification of mappings $ X \rightarrow Y $ can be interpreted as an extension problem, obstruction theory is applicable also to the description of the set $ [ X, Y] $ of homotopy classes of mappings from $ X $ into $ Y $. Let $ I = [ 0, 1] $ and let $ A = X \times \{ 0, 1 \} $ be a subspace of $ X \times I $. Then a pair of mappings $ f _ {0} , f _ {1} : X \rightarrow Y $ is interpreted as a mapping $ G: A \rightarrow Y $, $ G ( x, i) = f _ {i} ( x) $, $ i = 0, 1 $, and the presence of a homotopy between $ f _ {0} $ and $ f _ {1} $ means the presence of a mapping $ F: X \times I \rightarrow Y $ extending $ G $. If the homotopy $ F $ has been constructed on the $ n $- dimensional skeleton of $ X $, then the obstruction to its extension to $ X $ is the difference cochain

$$ d ^ {n} ( f _ {0} , f _ {1} ) \in \ C ^ {n} ( X; \pi _ {n} ( Y)). $$

As an application one may consider the description of the set $ [ X, Y] = [ X, K ( \pi , n)] $, $ n > 1 $, where $ K ( \pi , n) $ is the Eilenberg–MacLane space: $ \pi _ {i} ( K ( \pi , n)) = 0 $ for $ i \neq n $; $ \pi _ {n} ( K ( \pi , n)) = \pi $. Let $ f _ {0} : X \rightarrow K ( \pi , n) $ be a constant mapping and $ f: X \rightarrow K ( \pi , n) $ an arbitrary continuous mapping. Since $ H ^ {i} ( X; \pi _ {i} ( Y)) = 0 $ for $ i < n $, the mappings $ f _ {0} $ and $ f $ are homotopic on $ X ^ {n - 1 } $ and, after having chosen such a homotopy, one can define the difference cochain

$$ d ^ {n} ( f, f _ {0} ) \in \ C ^ {n} ( X; \pi _ {n} ( Y)) = \ C ^ {n} ( X; \pi ). $$

The cohomology class $ [ d ^ {n} ( f, f _ {0} )] \in H ^ {n} ( X; \pi ) $ is well-defined, i.e. does not depend on the choice of a homotopy between $ f _ {0} $ and $ f $( since $ \pi _ {i} ( Y) = 0 $ for $ i < n $). Further, if two mappings $ f, g: X \rightarrow Y $ are such that $ [ d ^ {n} ( f, f _ {0} )] = [ d ^ {n} ( g, f _ {0} )] $, then $ [ d ^ {n} ( f, g)] = 0 $, and hence $ f $ and $ g $ are homotopic on $ X ^ {n} $. The obstructions to extending this homotopy to $ X $ lie in the groups $ H ^ {i} ( X; \pi _ {i} ( Y)) = 0 $( since $ i > n $), and hence $ f $ and $ g $ are homotopic. Thus, the homotopy class of $ f $ is completely determined by the element $ [ d ^ {n} ( f, f _ {0} )] \in H ^ {n} ( X; \pi ) $. Finally, for any $ x \in H ^ {n} ( X; \pi ) $ there is a mapping $ f $ with $ [ d ^ {n} ( f, f _ {0} )] = x $, hence $ [ X, K ( \pi , n)] = H ^ {n} ( X; \pi ) $. Similarly, if $ \pi _ {i} ( Y) = i $ for $ i < n $ and if $ \mathop{\rm dim} X \leq n $, then $ [ X, Y] = H ^ {n} ( X; \pi _ {n} ( Y)) $.

In studying extension problems one has considered the possibility of extending "by skipping back one step" . A complete solution of the problem requires the analysis of the possibility of skipping back an arbitrary number of steps. Cohomology operations (cf. Cohomology operation) and Postnikov systems (cf. Postnikov system) are used to this end. E.g., in order to describe the set $ [ X, Y] $, where $ \pi _ {i} ( Y) = 0 $ for $ i < n $, $ \pi _ {n} ( Y) \neq 0 $, $ \mathop{\rm dim} X = n + r $, it is required, in general, to study the possibility of skipping back $ r + 1 $ steps, for which it is necessary to study the first $ n + r $ levels of the Postnikov system for $ Y $, i.e. to use cohomology operations of orders $ \leq r $( in the article Cohomology operation this problem is outlined for $ r = 1 $).

The theory of obstructions is also used in the more general situation of extension of sections (cf. Section of a mapping). Let $ p: E \rightarrow B $ be a fibration with fibre $ F $( where $ \pi _ {1} ( F ) = 0 $ and $ \pi _ {1} ( B) $ acts trivially on $ \pi _ {i} ( F ) $), let $ A \subset B $ and let $ s: A \rightarrow E $ be a section (i.e. a continuous mapping such that $ ps ( a) = a $). Can one extend $ s $ to $ B $? The corresponding obstructions lie in the groups $ H ^ {n + 1 } ( B; \pi _ {n} ( F )) $. An extension problem is obtained from this problem if one puts $ B = X $, $ E = X \times Y $, $ p ( x, y) = x $, $ s ( a) = ( a, g ( a)) $. Analogously one can also study the classification problem for sections using obstruction theory.

Finally, one can remove the restriction of homotopic simplicity of the space $ Y $ in the extension problem (as well as in the problem on sections); then one must use cohomology with local coefficients.

Obstruction theory was initiated by S. Eilenberg [2]. It was also known to L.S. Pontryagin, who did not formulate it explicitly but used it for the solution of concrete problems, see [1].

A good discussion can be found in [3] and [4].

References

[1] L.S. Pontryagin, "Classification of continuous transformations of a complex into a sphere" Dokl. Akad. Nauk SSSR , 19 (1938) pp. 361–363 (In Russian)
[2] S. Eilenberg, "Cohomology and continuous mappings" Ann. of Math. , 41 (1940) pp. 231–251
[3] S.-T. Hu, "Homotopy theory" , Acad. Press (1959)
[4] E. Thomas, "Seminar on fibre spaces" , Springer (1966)

Comments

The fundamental group $ \pi _ {1} ( X , x _ {0} ) $ acts on the homotopy groups $ \pi _ {n} ( X , x _ {0} ) $, $ n \geq 1 $, cf. Homotopy group. The space $ X $ is called $ n $- simple if this action (for this $ n $) is trivial; $ X $ is called simple or homotopy simple if it is path connected and $ n $- simple for all $ n \geq 1 $. Then $ \pi _ {1} ( X , x _ {0} ) $ is Abelian and acts trivially on all $ \pi _ {n} ( X , x _ {0} ) $. A path-connected $ H $- space is simple.

References

[a1] N.E. Steenrod, "The topology of fibre bundles" , Princeton Univ. Press (1951)
[a2] E.H. Spanier, "Algebraic topology" , McGraw-Hill (1966) pp. 269–276; 429–432
[a3] G.W. Whitehead, "Elements of homotopy theory" , Springer (1978)
[a4] H.J. Baues, "Obstruction theory" , Springer (1977)
How to Cite This Entry:
Obstruction. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Obstruction&oldid=11605
This article was adapted from an original article by Yu.B. Rudyak (originator), which appeared in Encyclopedia of Mathematics - ISBN 1402006098. See original article