Namespaces
Variants
Actions

Difference between revisions of "Yosida representation theorem"

From Encyclopedia of Mathematics
Jump to: navigation, search
m (link)
m (tex encoded by computer)
 
Line 1: Line 1:
Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990501.png" /> be a [[Topological space|topological space]], <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990502.png" /> the set of continuous real-valued functions on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990503.png" /> (cf. [[Continuous functions, space of|Continuous functions, space of]]). Using the pointwise defined partial order: <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990504.png" /> if and only if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990505.png" /> for all <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990506.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990507.png" /> becomes a [[Riesz space|Riesz space]]. The question arises whether it is possible to represent an arbitrary Riesz space by continuous functions with this order relation where, possibly, more general (extended) functions that can also take the values <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990508.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y0990509.png" /> may be used. Answers are given by various representation theorems. Below the Yosida representation theorem for the case of Archimedean Riesz spaces with a strong unit is described. For the Yosida representation theorem for Riesz spaces <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905010.png" />, where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905011.png" /> is a weak unit, see [[#References|[a1]]], and for the more general Johnson–Kist representation theorem, see [[#References|[a2]]].
+
<!--
 +
y0990501.png
 +
$#A+1 = 97 n = 0
 +
$#C+1 = 97 : ~/encyclopedia/old_files/data/Z099/Z.0909050 Yosida representation theorem
 +
Automatically converted into TeX, above some diagnostics.
 +
Please remove this comment and the {{TEX|auto}} line below,
 +
if TeX found to be correct.
 +
-->
  
A strong unit in a Riesz space <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905012.png" /> is an element <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905013.png" /> such that for all <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905014.png" /> there is an <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905015.png" /> such that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905016.png" />, i.e. the principal ideal generated by <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905017.png" /> should equal the whole space. A weak unit in <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905018.png" /> is an element of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905019.png" /> such that the principal band generated by <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905020.png" /> is all of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905021.png" />.
+
{{TEX|auto}}
 +
{{TEX|done}}
  
==The Riesz space <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905022.png" />.==
+
Let  $  X $
Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905023.png" /> be a compact Hausdorff space. Then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905024.png" /> is an Archimedean Riesz space and the function <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905025.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905026.png" /> for all <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905027.png" />, is a strong unit. Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905028.png" /> be a second compact Hausdorff space. The Banach–Stone theorem says that if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905029.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905030.png" /> are isomorphic as Riesz spaces, then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905031.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905032.png" /> are homeomorphic. As immediate corollaries one obtains that if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905033.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905034.png" /> are isomorphic as algebras ([[pointwise multiplication]]), then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905035.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905036.png" /> are homeomorphic; also, if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905037.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905038.png" /> are isomorphic as Banach spaces (sup-norm), then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905039.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905040.png" /> are homeomorphic.
+
be a [[Topological space|topological space]],  $  C( X) $
 +
the set of continuous real-valued functions on  $  X $(
 +
cf. [[Continuous functions, space of|Continuous functions, space of]]). Using the pointwise defined partial order: $  f \geq  g $
 +
if and only if  $  f( x) \geq  g( x) $
 +
for all $  x \in X $,  
 +
$  C( X) $
 +
becomes a [[Riesz space|Riesz space]]. The question arises whether it is possible to represent an arbitrary Riesz space by continuous functions with this order relation where, possibly, more general (extended) functions that can also take the values  $  + \infty $
 +
and $  - \infty $
 +
may be used. Answers are given by various representation theorems. Below the Yosida representation theorem for the case of Archimedean Riesz spaces with a strong unit is described. For the Yosida representation theorem for Riesz spaces $  ( L , e) $,
 +
where  $  e $
 +
is a weak unit, see [[#References|[a1]]], and for the more general Johnson–Kist representation theorem, see [[#References|[a2]]].
  
A topological space <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905041.png" /> is extremely disconnected if every open subset <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905042.png" /> has an open closure (i.e. <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905043.png" /> is both open and closed). Nakano's theorem says that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905044.png" /> is Dedekind complete if and only if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905045.png" /> is extremely disconnected. It was also obtained independently by T. Ogasawara and M.H. Stone, cf. [[#References|[a2]]].
+
A strong unit in a Riesz space  $  L $
 +
is an element  $  e \in L  ^ {+} = \{ {f \in L } : {f \geq  0 } \} $
 +
such that for all  $  f \in L $
 +
there is an  $  n \in \mathbf N $
 +
such that  $  | f | \leq  ne $,
 +
i.e. the principal ideal generated by  $  e $
 +
should equal the whole space. A weak unit in  $  L $
 +
is an element of  $  L  ^ {+} $
 +
such that the principal band generated by  $  e $
 +
is all of  $  L $.
 +
 
 +
==The Riesz space  $  C( X) $.==
 +
Let  $  X $
 +
be a compact Hausdorff space. Then  $  C( X) $
 +
is an Archimedean Riesz space and the function  $  \mathbf 1 : X \rightarrow \mathbf R $,
 +
$  x \mapsto 1 $
 +
for all  $  x $,
 +
is a strong unit. Let  $  Y $
 +
be a second compact Hausdorff space. The Banach–Stone theorem says that if  $  C( X) $
 +
and  $  C( Y) $
 +
are isomorphic as Riesz spaces, then  $  X $
 +
and  $  Y $
 +
are homeomorphic. As immediate corollaries one obtains that if  $  C( X) $
 +
and  $  C( Y) $
 +
are isomorphic as algebras ([[pointwise multiplication]]), then  $  X $
 +
and  $  Y $
 +
are homeomorphic; also, if  $  C( X) $
 +
and  $  C( Y) $
 +
are isomorphic as Banach spaces (sup-norm), then  $  X $
 +
and  $  Y $
 +
are homeomorphic.
 +
 
 +
A topological space  $  X $
 +
is extremely disconnected if every open subset $  U $
 +
has an open closure (i.e. $  \overline{U}\; $
 +
is both open and closed). Nakano's theorem says that $  C( X) $
 +
is Dedekind complete if and only if $  X $
 +
is extremely disconnected. It was also obtained independently by T. Ogasawara and M.H. Stone, cf. [[#References|[a2]]].
  
 
==Representation of Archimedean Riesz spaces with strong unit.==
 
==Representation of Archimedean Riesz spaces with strong unit.==
Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905046.png" /> be an Archimedean Riesz space. Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905047.png" /> be the set of maximal ideals of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905048.png" />. For each <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905049.png" />, let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905050.png" />. Define a topology on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905051.png" /> by taking the <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905052.png" /> as a [[subbase]] (cf. [[Pre-base|Pre-base]]). The closed sets of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905053.png" /> are the sets <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905054.png" />, where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905055.png" /> runs through all subsets of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905056.png" />. This topology is called the hull-kernel topology on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905057.png" />. The construction is a fairly familiar one and occurs in several parts of mathematics. It is originally due to M.H. Stone. Depending on which sets of ideals are used, the mathematical specialism involved, and the ideosyncracies of authors it is also called the Zariski topology, Gel'fand topology, Gel'fand–Kolmogorov topology, Jacobson topology, Grothendieck topology, etc.
+
Let $  L $
 +
be an Archimedean Riesz space. Let $  X = \mathop{\rm MSpec} ( L) $
 +
be the set of maximal ideals of $  L $.  
 +
For each $  f \in L  ^ {+} $,  
 +
let $  X _ {f} = \{ {M \in X } : {f \notin M } \} $.  
 +
Define a topology on $  X $
 +
by taking the $  X _ {f} $
 +
as a [[subbase]] (cf. [[Pre-base|Pre-base]]). The closed sets of $  X $
 +
are the sets $  C _ {D} = \{ {M \in X } : {D \subset  M } \} $,  
 +
where $  D $
 +
runs through all subsets of $  L  ^ {+} $.  
 +
This topology is called the hull-kernel topology on $  X $.  
 +
The construction is a fairly familiar one and occurs in several parts of mathematics. It is originally due to M.H. Stone. Depending on which sets of ideals are used, the mathematical specialism involved, and the ideosyncracies of authors it is also called the Zariski topology, Gel'fand topology, Gel'fand–Kolmogorov topology, Jacobson topology, Grothendieck topology, etc.
  
From now on, let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905058.png" /> have a strong unit <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905059.png" />. For each <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905060.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905061.png" /> and there is a unique homomorphism of Riesz spaces <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905062.png" /> such that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905063.png" />. Using this, one defines for every <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905064.png" /> a function <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905065.png" /> by <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905066.png" />. The number <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905067.png" /> can also be described as the unique real number such that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905068.png" />. One now has the following representation theorem (K. Yosida, S. Kakutani, M.G. Krein, S.G. Krein, H. Nakano). Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905069.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905070.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905071.png" /> be as just described. Then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905072.png" /> defines a continuous function on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905073.png" /> and the mapping <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905074.png" /> is a Riesz isomorphism of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905075.png" /> onto a Riesz subspace <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905076.png" />. There are a number of complementary facts. Using the [[Stone–Weierstrass theorem|Stone–Weierstrass theorem]] one obtains that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905077.png" /> is norm dense in <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905078.png" />; it is then also order dense.
+
From now on, let $  L $
 +
have a strong unit $  e $.  
 +
For each $  M \in  \mathop{\rm MSpec} ( L) $,  
 +
$  L / M \simeq \mathbf R $
 +
and there is a unique homomorphism of Riesz spaces $  \phi _ {M} : L \rightarrow \mathbf R $
 +
such that $  \phi _ {M} ( e) = 1 $.  
 +
Using this, one defines for every $  f \in L $
 +
a function $  \widehat{f}  :  X = \mathop{\rm MSpec} ( L) \rightarrow \mathbf R $
 +
by $  \widehat{f}  ( M) = \phi _ {M} ( f  ) $.  
 +
The number $  \widehat{f}  ( M) $
 +
can also be described as the unique real number such that $  f - \widehat{f}  ( M) e \in M $.  
 +
One now has the following representation theorem (K. Yosida, S. Kakutani, M.G. Krein, S.G. Krein, H. Nakano). Let $  X $,  
 +
$  L $,  
 +
$  e $
 +
be as just described. Then $  \widehat{f}  ( M) = \phi _ {M} ( f  ) $
 +
defines a continuous function on $  X $
 +
and the mapping $  f \mapsto \widehat{f}  $
 +
is a Riesz isomorphism of $  L $
 +
onto a Riesz subspace $  \widehat{L}  \subset  C( M) $.  
 +
There are a number of complementary facts. Using the [[Stone–Weierstrass theorem|Stone–Weierstrass theorem]] one obtains that $  \widehat{L}  $
 +
is norm dense in $  C( X) $;  
 +
it is then also order dense.
  
Given <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905079.png" />, where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905080.png" /> is a weak unit, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905081.png" /> defines a metric on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905082.png" />, called the uniform metric. If <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905083.png" /> is complete with respect to this metric, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905084.png" /> is called uniformly closed. A further addition to the representation theorem is then that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905085.png" /> is isomorphic to <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905086.png" /> if and only if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905087.png" /> is a strong unit and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905088.png" /> is uniformly closed. This last statement, together with that fact that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905089.png" /> is isomorphic to a sub-Riesz space of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905090.png" /> (if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905091.png" /> is a strong unit), is also referred to as the Krein–Kakutani theorem.
+
Given $  ( L, e) $,  
 +
where $  e $
 +
is a weak unit, $  \rho ( f, g) = \inf \{ {r \in \mathbf R } : {| f- g | \wedge e \leq  r e } \} $
 +
defines a metric on $  L $,  
 +
called the uniform metric. If $  L $
 +
is complete with respect to this metric, $  L $
 +
is called uniformly closed. A further addition to the representation theorem is then that $  ( L , e) $
 +
is isomorphic to $  ( C( X), \mathbf 1 ) $
 +
if and only if $  e $
 +
is a strong unit and $  L $
 +
is uniformly closed. This last statement, together with that fact that $  ( L, e) $
 +
is isomorphic to a sub-Riesz space of $  ( C( X), \mathbf 1 ) $(
 +
if $  e $
 +
is a strong unit), is also referred to as the Krein–Kakutani theorem.
  
A final complement to the Yosida representation theorem is that if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905092.png" /> has the principal projection property, i.e. <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905093.png" /> for every principal band <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905094.png" />, then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905095.png" /> is zero dimensional and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905096.png" /> contains all locally constant functions on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/y/y099/y099050/y09905097.png" />. This can also be called the Freudenthal spectral theorem, [[#References|[a3]]], in the sense that that theorem in its traditional formulation is an immediate consequence of this result, cf. (the editorial comments to) [[Riesz space|Riesz space]].
+
A final complement to the Yosida representation theorem is that if $  L $
 +
has the principal projection property, i.e. $  A + A  ^ {d} = L $
 +
for every principal band $  A $,  
 +
then $  X = \mathop{\rm MSpec} ( L) $
 +
is zero dimensional and $  \widehat{L}  $
 +
contains all locally constant functions on $  X $.  
 +
This can also be called the Freudenthal spectral theorem, [[#References|[a3]]], in the sense that that theorem in its traditional formulation is an immediate consequence of this result, cf. (the editorial comments to) [[Riesz space|Riesz space]].
  
 
====References====
 
====References====
 
<table><TR><TD valign="top">[a1]</TD> <TD valign="top">  A.W. Hager,  L.C. Robertson,  "Representing and ringifying a Riesz space" , ''Symp. Math. INDAM'' , '''21''' , Acad. Press  (1977)  pp. 411–432</TD></TR><TR><TD valign="top">[a2]</TD> <TD valign="top">  W.A.J. Luxemburg,  A.C. Zaanen,  "Riesz spaces" , '''I''' , North-Holland  (1971)  pp. Chapt. 7</TD></TR><TR><TD valign="top">[a3]</TD> <TD valign="top">  E. de Jonge,  A.C.M. van Rooy,  "Introduction to Riesz spaces" , ''Tracts'' , '''8''' , Math. Centre  (1977)</TD></TR></table>
 
<table><TR><TD valign="top">[a1]</TD> <TD valign="top">  A.W. Hager,  L.C. Robertson,  "Representing and ringifying a Riesz space" , ''Symp. Math. INDAM'' , '''21''' , Acad. Press  (1977)  pp. 411–432</TD></TR><TR><TD valign="top">[a2]</TD> <TD valign="top">  W.A.J. Luxemburg,  A.C. Zaanen,  "Riesz spaces" , '''I''' , North-Holland  (1971)  pp. Chapt. 7</TD></TR><TR><TD valign="top">[a3]</TD> <TD valign="top">  E. de Jonge,  A.C.M. van Rooy,  "Introduction to Riesz spaces" , ''Tracts'' , '''8''' , Math. Centre  (1977)</TD></TR></table>

Latest revision as of 08:29, 6 June 2020


Let $ X $ be a topological space, $ C( X) $ the set of continuous real-valued functions on $ X $( cf. Continuous functions, space of). Using the pointwise defined partial order: $ f \geq g $ if and only if $ f( x) \geq g( x) $ for all $ x \in X $, $ C( X) $ becomes a Riesz space. The question arises whether it is possible to represent an arbitrary Riesz space by continuous functions with this order relation where, possibly, more general (extended) functions that can also take the values $ + \infty $ and $ - \infty $ may be used. Answers are given by various representation theorems. Below the Yosida representation theorem for the case of Archimedean Riesz spaces with a strong unit is described. For the Yosida representation theorem for Riesz spaces $ ( L , e) $, where $ e $ is a weak unit, see [a1], and for the more general Johnson–Kist representation theorem, see [a2].

A strong unit in a Riesz space $ L $ is an element $ e \in L ^ {+} = \{ {f \in L } : {f \geq 0 } \} $ such that for all $ f \in L $ there is an $ n \in \mathbf N $ such that $ | f | \leq ne $, i.e. the principal ideal generated by $ e $ should equal the whole space. A weak unit in $ L $ is an element of $ L ^ {+} $ such that the principal band generated by $ e $ is all of $ L $.

The Riesz space $ C( X) $.

Let $ X $ be a compact Hausdorff space. Then $ C( X) $ is an Archimedean Riesz space and the function $ \mathbf 1 : X \rightarrow \mathbf R $, $ x \mapsto 1 $ for all $ x $, is a strong unit. Let $ Y $ be a second compact Hausdorff space. The Banach–Stone theorem says that if $ C( X) $ and $ C( Y) $ are isomorphic as Riesz spaces, then $ X $ and $ Y $ are homeomorphic. As immediate corollaries one obtains that if $ C( X) $ and $ C( Y) $ are isomorphic as algebras (pointwise multiplication), then $ X $ and $ Y $ are homeomorphic; also, if $ C( X) $ and $ C( Y) $ are isomorphic as Banach spaces (sup-norm), then $ X $ and $ Y $ are homeomorphic.

A topological space $ X $ is extremely disconnected if every open subset $ U $ has an open closure (i.e. $ \overline{U}\; $ is both open and closed). Nakano's theorem says that $ C( X) $ is Dedekind complete if and only if $ X $ is extremely disconnected. It was also obtained independently by T. Ogasawara and M.H. Stone, cf. [a2].

Representation of Archimedean Riesz spaces with strong unit.

Let $ L $ be an Archimedean Riesz space. Let $ X = \mathop{\rm MSpec} ( L) $ be the set of maximal ideals of $ L $. For each $ f \in L ^ {+} $, let $ X _ {f} = \{ {M \in X } : {f \notin M } \} $. Define a topology on $ X $ by taking the $ X _ {f} $ as a subbase (cf. Pre-base). The closed sets of $ X $ are the sets $ C _ {D} = \{ {M \in X } : {D \subset M } \} $, where $ D $ runs through all subsets of $ L ^ {+} $. This topology is called the hull-kernel topology on $ X $. The construction is a fairly familiar one and occurs in several parts of mathematics. It is originally due to M.H. Stone. Depending on which sets of ideals are used, the mathematical specialism involved, and the ideosyncracies of authors it is also called the Zariski topology, Gel'fand topology, Gel'fand–Kolmogorov topology, Jacobson topology, Grothendieck topology, etc.

From now on, let $ L $ have a strong unit $ e $. For each $ M \in \mathop{\rm MSpec} ( L) $, $ L / M \simeq \mathbf R $ and there is a unique homomorphism of Riesz spaces $ \phi _ {M} : L \rightarrow \mathbf R $ such that $ \phi _ {M} ( e) = 1 $. Using this, one defines for every $ f \in L $ a function $ \widehat{f} : X = \mathop{\rm MSpec} ( L) \rightarrow \mathbf R $ by $ \widehat{f} ( M) = \phi _ {M} ( f ) $. The number $ \widehat{f} ( M) $ can also be described as the unique real number such that $ f - \widehat{f} ( M) e \in M $. One now has the following representation theorem (K. Yosida, S. Kakutani, M.G. Krein, S.G. Krein, H. Nakano). Let $ X $, $ L $, $ e $ be as just described. Then $ \widehat{f} ( M) = \phi _ {M} ( f ) $ defines a continuous function on $ X $ and the mapping $ f \mapsto \widehat{f} $ is a Riesz isomorphism of $ L $ onto a Riesz subspace $ \widehat{L} \subset C( M) $. There are a number of complementary facts. Using the Stone–Weierstrass theorem one obtains that $ \widehat{L} $ is norm dense in $ C( X) $; it is then also order dense.

Given $ ( L, e) $, where $ e $ is a weak unit, $ \rho ( f, g) = \inf \{ {r \in \mathbf R } : {| f- g | \wedge e \leq r e } \} $ defines a metric on $ L $, called the uniform metric. If $ L $ is complete with respect to this metric, $ L $ is called uniformly closed. A further addition to the representation theorem is then that $ ( L , e) $ is isomorphic to $ ( C( X), \mathbf 1 ) $ if and only if $ e $ is a strong unit and $ L $ is uniformly closed. This last statement, together with that fact that $ ( L, e) $ is isomorphic to a sub-Riesz space of $ ( C( X), \mathbf 1 ) $( if $ e $ is a strong unit), is also referred to as the Krein–Kakutani theorem.

A final complement to the Yosida representation theorem is that if $ L $ has the principal projection property, i.e. $ A + A ^ {d} = L $ for every principal band $ A $, then $ X = \mathop{\rm MSpec} ( L) $ is zero dimensional and $ \widehat{L} $ contains all locally constant functions on $ X $. This can also be called the Freudenthal spectral theorem, [a3], in the sense that that theorem in its traditional formulation is an immediate consequence of this result, cf. (the editorial comments to) Riesz space.

References

[a1] A.W. Hager, L.C. Robertson, "Representing and ringifying a Riesz space" , Symp. Math. INDAM , 21 , Acad. Press (1977) pp. 411–432
[a2] W.A.J. Luxemburg, A.C. Zaanen, "Riesz spaces" , I , North-Holland (1971) pp. Chapt. 7
[a3] E. de Jonge, A.C.M. van Rooy, "Introduction to Riesz spaces" , Tracts , 8 , Math. Centre (1977)
How to Cite This Entry:
Yosida representation theorem. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Yosida_representation_theorem&oldid=49242