Namespaces
Variants
Actions

Difference between revisions of "Projection methods"

From Encyclopedia of Mathematics
Jump to: navigation, search
(Importing text file)
 
m (tex encoded by computer)
Line 1: Line 1:
 +
<!--
 +
p0751401.png
 +
$#A+1 = 101 n = 0
 +
$#C+1 = 101 : ~/encyclopedia/old_files/data/P075/P.0705140 Projection methods
 +
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}}
 +
 
Methods for finding an approximate solution of an operator equation in a prescribed subspace, based on projecting the equation onto some (generally speaking, different) subspace. Projection methods constitute the basis for various computational schemes for solving boundary value problems, including the finite element and collocation methods (cf. [[Galerkin method|Galerkin method]]; [[Collocation method|Collocation method]]).
 
Methods for finding an approximate solution of an operator equation in a prescribed subspace, based on projecting the equation onto some (generally speaking, different) subspace. Projection methods constitute the basis for various computational schemes for solving boundary value problems, including the finite element and collocation methods (cf. [[Galerkin method|Galerkin method]]; [[Collocation method|Collocation method]]).
  
Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p0751401.png" /> be an operator with domain of definition <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p0751402.png" /> in a Banach space <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p0751403.png" /> and with range <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p0751404.png" /> in a Banach space <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p0751405.png" />. To solve the equation
+
Let $  L $
 +
be an operator with domain of definition $  D ( L) $
 +
in a Banach space $  X $
 +
and with range $  R ( L) $
 +
in a Banach space $  Y $.  
 +
To solve the equation
  
<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/p/p075/p075140/p0751406.png" /></td> <td valign="top" style="width:5%;text-align:right;">(1)</td></tr></table>
+
$$ \tag{1 }
 +
Lx  = y
 +
$$
  
by a projection method, one chooses two sequences <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p0751407.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p0751408.png" /> of subspaces
+
by a projection method, one chooses two sequences $  \{ X _ {n} \} $
 +
and $  \{ Y _ {n} \} $
 +
of subspaces
  
<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/p/p075/p075140/p0751409.png" /></td> </tr></table>
+
$$
 +
X _ {n}  \subset  D ( L)  \subset  X ,\  Y _ {n}  \subset  Y ,\ \
 +
n = 1 , 2 \dots
 +
$$
  
as well as projectors <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514010.png" /> projecting <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514011.png" /> onto <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514012.png" />. Equation (1) is replaced by the approximate equation
+
as well as projectors $  P _ {n} $
 +
projecting $  Y $
 +
onto $  Y _ {n} $.  
 +
Equation (1) is replaced by the approximate equation
  
<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/p/p075/p075140/p07514013.png" /></td> <td valign="top" style="width:5%;text-align:right;">(2)</td></tr></table>
+
$$ \tag{2 }
 +
P _ {n} L x _ {n}  = P _ {n} y ,\ \
 +
x _ {n} \in X _ {n} .
 +
$$
  
In the case <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514014.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514015.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514016.png" /> the projection method (2) is usually called the Galerkin method (sometimes the latter method is interpreted in a wider sense, see [[Galerkin method|Galerkin method]]).
+
In the case $  X = Y $,
 +
$  X _ {n} = Y _ {n} $,  
 +
$  n = 1 , 2 \dots $
 +
the projection method (2) is usually called the Galerkin method (sometimes the latter method is interpreted in a wider sense, see [[Galerkin method|Galerkin method]]).
  
A convergence theorem holds for projection methods for linear equations (in the case of finite-dimensional subspaces <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514017.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514018.png" />). Suppose that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514019.png" /> is linear and takes <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514020.png" /> onto <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514021.png" /> bijectively, with <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514022.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514023.png" /> dense in <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514024.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514025.png" />, respectively. Suppose that the subspaces <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514026.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514027.png" /> are finite-dimensional, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514028.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514029.png" /> and that the projectors <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514030.png" /> are uniformly bounded in <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514031.png" />, that is, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514032.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514033.png" />. Then the following condition a) is equivalent to the conditions b) and c) combined.
+
A convergence theorem holds for projection methods for linear equations (in the case of finite-dimensional subspaces $  X _ {n} $
 +
and $  Y _ {n} $).  
 +
Suppose that $  L $
 +
is linear and takes $  D ( L) $
 +
onto $  R ( L) $
 +
bijectively, with $  D ( L) $
 +
and $  R ( L) $
 +
dense in $  X $
 +
and $  Y $,  
 +
respectively. Suppose that the subspaces $  X _ {n} $
 +
and $  Y _ {n} $
 +
are finite-dimensional, $  \mathop{\rm dim}  X _ {n} = \mathop{\rm dim}  Y _ {n} $,
 +
$  n = 1 , 2 \dots $
 +
and that the projectors $  P _ {n} $
 +
are uniformly bounded in $  n $,  
 +
that is, $  \| P _ {n} \| \leq  \textrm{ const } $,
 +
$  n = 1 , 2 , . . . $.  
 +
Then the following condition a) is equivalent to the conditions b) and c) combined.
  
a) From some <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514034.png" /> onwards there exists a unique solution <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514035.png" /> of (2), and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514036.png" /> for any <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514037.png" />;
+
a) From some $  n = n _ {0} $
 +
onwards there exists a unique solution $  x _ {n} $
 +
of (2), and $  \| L x _ {n} - y \| \rightarrow 0 $
 +
for any $  y \in Y $;
  
b) The sequence of subspaces <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514038.png" /> is dense in the limit in <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514039.png" />, that is, the distance <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514040.png" /> as <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514041.png" /> for every <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514042.png" />;
+
b) The sequence of subspaces $  L X _ {n} $
 +
is dense in the limit in $  Y $,  
 +
that is, the distance $  d ( y , L X _ {n} ) \rightarrow 0 $
 +
as $  n \rightarrow \infty $
 +
for every $  y \in Y $;
  
c) <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514043.png" />, where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514044.png" />.
+
c) $  \tau \equiv \lim\limits _ {n \rightarrow \infty }  \tau _ {n} > 0 $,
 +
where  $  \tau = \inf  \{ {\| P _ {n} y _ {n} \| } : {y _ {n} \in L X _ {n} , \| y _ {n} \| = 1 } \} $.
  
 
The rate of convergence under the conditions b) and c) is characterized by the inequality
 
The rate of convergence under the conditions b) and c) is characterized by the inequality
  
<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/p/p075/p075140/p07514045.png" /></td> <td valign="top" style="width:5%;text-align:right;">(3)</td></tr></table>
+
$$ \tag{3 }
 +
d ( y , L X _ {n} )  \leq  \| L x _ {n} - y \|
 +
< \left ( 1 +
 +
\frac{c}{\tau _ {n} }
 +
\right ) d ( y , L X _ {n} ) .
 +
$$
 +
 
 +
If  $  X $
 +
and  $  Y $
 +
are Hilbert spaces and  $  P _ {n} $
 +
and  $  Q _ {n} $
 +
are orthoprojectors projecting  $  Y $
 +
onto  $  Y _ {n} $
 +
and  $  L X _ {n} $,
 +
respectively, condition c) is equivalent to the condition
  
If <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514046.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514047.png" /> are Hilbert spaces and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514048.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514049.png" /> are orthoprojectors projecting <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514050.png" /> onto <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514051.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514052.png" />, respectively, condition c) is equivalent to the condition
+
c')  $  \theta \equiv {\overline{\lim\limits}\; } _ {n \rightarrow \infty }  \theta _ {n} < 1 $,
 +
where  $  \theta _ {n} = \| P _ {n} - Q _ {n} \| $
 +
is the gap (angle) between the subspaces  $  Y _ {n} $
 +
and $  L X _ {n} $;
 +
instead of (3) one obtains the estimate
  
c') <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514053.png" />, where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514054.png" /> is the gap (angle) between the subspaces <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514055.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514056.png" />; instead of (3) one obtains the estimate
+
$$
 +
\| y - Q _ {n} y \|  \leq  \| L x _ {n} - y \|  \leq  \
  
<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/p/p075/p075140/p07514057.png" /></td> </tr></table>
+
\frac{1}{\sqrt {1 - \theta _ {n}  ^ {2} }}
 +
\| y - Q _ {n} y \| .
 +
$$
  
In case <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514058.png" /> (the least-squares method) one has <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514059.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514060.png" /> and the convergence criterion is the condition b).
+
In case $  Y _ {n} = L X _ {n} $(
 +
the least-squares method) one has $  \theta _ {n} = 0 $,
 +
$  n = 1 , 2 \dots $
 +
and the convergence criterion is the condition b).
  
The theorem yields a condition for convergence of the discrepancy <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514061.png" />. If <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514062.png" /> is bounded and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514063.png" />, then convergence of the discrepancy implies convergence of the approximations <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514064.png" /> themselves to the solution <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514065.png" /> of equation (1). From the theorem, a convenient criterion for the convergence of the Galerkin method can be extracted; for the Galerkin–Petrov method an additional condition of the type c') should be imposed.
+
The theorem yields a condition for convergence of the discrepancy $  \| L x _ {n} - y \| $.  
 +
If $  L  ^ {-} 1 $
 +
is bounded and $  y \in R ( L) $,  
 +
then convergence of the discrepancy implies convergence of the approximations $  x _ {n} $
 +
themselves to the solution $  x = L  ^ {-} 1 y $
 +
of equation (1). From the theorem, a convenient criterion for the convergence of the Galerkin method can be extracted; for the Galerkin–Petrov method an additional condition of the type c') should be imposed.
  
Suppose that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514066.png" /> is a bounded linear form and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514067.png" /> is a bounded bilinear form on a real Hilbert space <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514068.png" /> (or sesquilinear in the case of a complex <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514069.png" />). It is assumed that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514070.png" /> is representable as <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514071.png" />, such that
+
Suppose that $  l $
 +
is a bounded linear form and $  a $
 +
is a bounded bilinear form on a real Hilbert space $  H $(
 +
or sesquilinear in the case of a complex $  H $).  
 +
It is assumed that $  a $
 +
is representable as $  a = \widehat{a}  + b $,  
 +
such 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/p/p075/p075140/p07514072.png" /></td> </tr></table>
+
$$
 +
\widehat{a}  ( u , u )  \geq  \gamma  \| u \|  ^ {2} \ \
 +
\forall u \in H ,\  \dot \gamma  = \textrm{ const }  > 0 ,
 +
$$
  
while the bilinear form <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514073.png" /> is completely continuous, i.e. the weak convergences <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514074.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514075.png" /> in <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514076.png" /> imply the convergence <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514077.png" /> (the forms <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514078.png" /> are not necessarily symmetric). Suppose that the following problem is posed: Find a <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514079.png" /> such that
+
while the bilinear form $  b $
 +
is completely continuous, i.e. the weak convergences $  u _ {n} \rightarrow u $,  
 +
$  v _ {n} \rightarrow v $
 +
in $  H $
 +
imply the convergence $  b ( u _ {n} , v _ {n} ) \rightarrow b ( u , v ) $(
 +
the forms $  a , \widehat{a}  , b $
 +
are not necessarily symmetric). Suppose that the following problem is posed: Find a $  u \in H $
 +
such 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/p/p075/p075140/p07514080.png" /></td> <td valign="top" style="width:5%;text-align:right;">(4)</td></tr></table>
+
$$ \tag{4 }
 +
a ( u , v )  = l ( v) \ \
 +
\forall v \in H .
 +
$$
  
The Galerkin method for solving (4) consists in the following. One chooses (closed) subspaces <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514081.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514082.png" /> and finds a <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514083.png" /> such that
+
The Galerkin method for solving (4) consists in the following. One chooses (closed) subspaces $  H _ {n} \subset  H $,
 +
$  n = 1 , 2 \dots $
 +
and finds a $  u _ {n} \in H _ {n} $
 +
such 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/p/p075/p075140/p07514084.png" /></td> <td valign="top" style="width:5%;text-align:right;">(5)</td></tr></table>
+
$$ \tag{5 }
 +
a ( u _ {n} , v _ {n} )  = l ( v _ {n} ) \ \
 +
\forall v _ {n} \in H _ {n} .
 +
$$
  
The following theorem holds: Suppose that <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514085.png" /> is dense in the limit in <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514086.png" />, that the conditions imposed above on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514087.png" /> are satisfied and that problem (4) has a unique solution <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514088.png" /> (an equivalent condition is: the homogeneous problem of finding <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514089.png" /> from the condition <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514090.png" /> for every <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514091.png" /> has only the trivial solution <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514092.png" />); then problem (5) for sufficiently large <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514093.png" /> has a unique solution <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514094.png" />, and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514095.png" /> with the estimate
+
The following theorem holds: Suppose that $  \{ H _ {n} \} $
 +
is dense in the limit in $  H $,  
 +
that the conditions imposed above on $  a $
 +
are satisfied and that problem (4) has a unique solution $  u \in H $(
 +
an equivalent condition is: the homogeneous problem of finding $  u $
 +
from the condition $  a ( u , v ) = 0 $
 +
for every $  v \in H $
 +
has only the trivial solution $  u = 0 $);  
 +
then problem (5) for sufficiently large $  n $
 +
has a unique solution $  u _ {n} \in H _ {n} $,  
 +
and $  \| u _ {n} - u \| \rightarrow 0 $
 +
with the estimate
  
<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/p/p075/p075140/p07514096.png" /></td> </tr></table>
+
$$
 +
\| u - O _ {n} u \|  \leq  \| e _ {n} - u \|  \leq  \
 +
c  \| u - O _ {n} u \| ,
 +
$$
  
where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514097.png" /> is the orthoprojector projecting <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514098.png" /> onto <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p07514099.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p075140100.png" />.
+
where $  O _ {n} $
 +
is the orthoprojector projecting $  H $
 +
onto $  H _ {n} $
 +
and $  c = \textrm{ const } $.
  
In applications to boundary value problems for equations of elliptic type, as a rule, the energy space of the principal part of the corresponding differential operator is chosen as <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/p/p075/p075140/p075140101.png" />.
+
In applications to boundary value problems for equations of elliptic type, as a rule, the energy space of the principal part of the corresponding differential operator is chosen as $  H $.
  
 
====References====
 
====References====
 
<table><TR><TD valign="top">[1]</TD> <TD valign="top">  M.A. Krasnosel'skii,  G.M. Vainikko,  P.P. Zabreiko,  et al.,  "Approximate solution of operator equations" , Wolters-Noordhoff  (1972)  (Translated from Russian)</TD></TR></table>
 
<table><TR><TD valign="top">[1]</TD> <TD valign="top">  M.A. Krasnosel'skii,  G.M. Vainikko,  P.P. Zabreiko,  et al.,  "Approximate solution of operator equations" , Wolters-Noordhoff  (1972)  (Translated from Russian)</TD></TR></table>
 
 
  
 
====Comments====
 
====Comments====
 
  
 
====References====
 
====References====
 
<table><TR><TD valign="top">[a1]</TD> <TD valign="top">  C.A.J. Fletcher,  "Computational Galerkin methods" , Springer  (1984)</TD></TR></table>
 
<table><TR><TD valign="top">[a1]</TD> <TD valign="top">  C.A.J. Fletcher,  "Computational Galerkin methods" , Springer  (1984)</TD></TR></table>

Revision as of 08:08, 6 June 2020


Methods for finding an approximate solution of an operator equation in a prescribed subspace, based on projecting the equation onto some (generally speaking, different) subspace. Projection methods constitute the basis for various computational schemes for solving boundary value problems, including the finite element and collocation methods (cf. Galerkin method; Collocation method).

Let $ L $ be an operator with domain of definition $ D ( L) $ in a Banach space $ X $ and with range $ R ( L) $ in a Banach space $ Y $. To solve the equation

$$ \tag{1 } Lx = y $$

by a projection method, one chooses two sequences $ \{ X _ {n} \} $ and $ \{ Y _ {n} \} $ of subspaces

$$ X _ {n} \subset D ( L) \subset X ,\ Y _ {n} \subset Y ,\ \ n = 1 , 2 \dots $$

as well as projectors $ P _ {n} $ projecting $ Y $ onto $ Y _ {n} $. Equation (1) is replaced by the approximate equation

$$ \tag{2 } P _ {n} L x _ {n} = P _ {n} y ,\ \ x _ {n} \in X _ {n} . $$

In the case $ X = Y $, $ X _ {n} = Y _ {n} $, $ n = 1 , 2 \dots $ the projection method (2) is usually called the Galerkin method (sometimes the latter method is interpreted in a wider sense, see Galerkin method).

A convergence theorem holds for projection methods for linear equations (in the case of finite-dimensional subspaces $ X _ {n} $ and $ Y _ {n} $). Suppose that $ L $ is linear and takes $ D ( L) $ onto $ R ( L) $ bijectively, with $ D ( L) $ and $ R ( L) $ dense in $ X $ and $ Y $, respectively. Suppose that the subspaces $ X _ {n} $ and $ Y _ {n} $ are finite-dimensional, $ \mathop{\rm dim} X _ {n} = \mathop{\rm dim} Y _ {n} $, $ n = 1 , 2 \dots $ and that the projectors $ P _ {n} $ are uniformly bounded in $ n $, that is, $ \| P _ {n} \| \leq \textrm{ const } $, $ n = 1 , 2 , . . . $. Then the following condition a) is equivalent to the conditions b) and c) combined.

a) From some $ n = n _ {0} $ onwards there exists a unique solution $ x _ {n} $ of (2), and $ \| L x _ {n} - y \| \rightarrow 0 $ for any $ y \in Y $;

b) The sequence of subspaces $ L X _ {n} $ is dense in the limit in $ Y $, that is, the distance $ d ( y , L X _ {n} ) \rightarrow 0 $ as $ n \rightarrow \infty $ for every $ y \in Y $;

c) $ \tau \equiv \lim\limits _ {n \rightarrow \infty } \tau _ {n} > 0 $, where $ \tau = \inf \{ {\| P _ {n} y _ {n} \| } : {y _ {n} \in L X _ {n} , \| y _ {n} \| = 1 } \} $.

The rate of convergence under the conditions b) and c) is characterized by the inequality

$$ \tag{3 } d ( y , L X _ {n} ) \leq \| L x _ {n} - y \| < \left ( 1 + \frac{c}{\tau _ {n} } \right ) d ( y , L X _ {n} ) . $$

If $ X $ and $ Y $ are Hilbert spaces and $ P _ {n} $ and $ Q _ {n} $ are orthoprojectors projecting $ Y $ onto $ Y _ {n} $ and $ L X _ {n} $, respectively, condition c) is equivalent to the condition

c') $ \theta \equiv {\overline{\lim\limits}\; } _ {n \rightarrow \infty } \theta _ {n} < 1 $, where $ \theta _ {n} = \| P _ {n} - Q _ {n} \| $ is the gap (angle) between the subspaces $ Y _ {n} $ and $ L X _ {n} $; instead of (3) one obtains the estimate

$$ \| y - Q _ {n} y \| \leq \| L x _ {n} - y \| \leq \ \frac{1}{\sqrt {1 - \theta _ {n} ^ {2} }} \| y - Q _ {n} y \| . $$

In case $ Y _ {n} = L X _ {n} $( the least-squares method) one has $ \theta _ {n} = 0 $, $ n = 1 , 2 \dots $ and the convergence criterion is the condition b).

The theorem yields a condition for convergence of the discrepancy $ \| L x _ {n} - y \| $. If $ L ^ {-} 1 $ is bounded and $ y \in R ( L) $, then convergence of the discrepancy implies convergence of the approximations $ x _ {n} $ themselves to the solution $ x = L ^ {-} 1 y $ of equation (1). From the theorem, a convenient criterion for the convergence of the Galerkin method can be extracted; for the Galerkin–Petrov method an additional condition of the type c') should be imposed.

Suppose that $ l $ is a bounded linear form and $ a $ is a bounded bilinear form on a real Hilbert space $ H $( or sesquilinear in the case of a complex $ H $). It is assumed that $ a $ is representable as $ a = \widehat{a} + b $, such that

$$ \widehat{a} ( u , u ) \geq \gamma \| u \| ^ {2} \ \ \forall u \in H ,\ \dot \gamma = \textrm{ const } > 0 , $$

while the bilinear form $ b $ is completely continuous, i.e. the weak convergences $ u _ {n} \rightarrow u $, $ v _ {n} \rightarrow v $ in $ H $ imply the convergence $ b ( u _ {n} , v _ {n} ) \rightarrow b ( u , v ) $( the forms $ a , \widehat{a} , b $ are not necessarily symmetric). Suppose that the following problem is posed: Find a $ u \in H $ such that

$$ \tag{4 } a ( u , v ) = l ( v) \ \ \forall v \in H . $$

The Galerkin method for solving (4) consists in the following. One chooses (closed) subspaces $ H _ {n} \subset H $, $ n = 1 , 2 \dots $ and finds a $ u _ {n} \in H _ {n} $ such that

$$ \tag{5 } a ( u _ {n} , v _ {n} ) = l ( v _ {n} ) \ \ \forall v _ {n} \in H _ {n} . $$

The following theorem holds: Suppose that $ \{ H _ {n} \} $ is dense in the limit in $ H $, that the conditions imposed above on $ a $ are satisfied and that problem (4) has a unique solution $ u \in H $( an equivalent condition is: the homogeneous problem of finding $ u $ from the condition $ a ( u , v ) = 0 $ for every $ v \in H $ has only the trivial solution $ u = 0 $); then problem (5) for sufficiently large $ n $ has a unique solution $ u _ {n} \in H _ {n} $, and $ \| u _ {n} - u \| \rightarrow 0 $ with the estimate

$$ \| u - O _ {n} u \| \leq \| e _ {n} - u \| \leq \ c \| u - O _ {n} u \| , $$

where $ O _ {n} $ is the orthoprojector projecting $ H $ onto $ H _ {n} $ and $ c = \textrm{ const } $.

In applications to boundary value problems for equations of elliptic type, as a rule, the energy space of the principal part of the corresponding differential operator is chosen as $ H $.

References

[1] M.A. Krasnosel'skii, G.M. Vainikko, P.P. Zabreiko, et al., "Approximate solution of operator equations" , Wolters-Noordhoff (1972) (Translated from Russian)

Comments

References

[a1] C.A.J. Fletcher, "Computational Galerkin methods" , Springer (1984)
How to Cite This Entry:
Projection methods. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Projection_methods&oldid=15275
This article was adapted from an original article by G.M. Vainikko (originator), which appeared in Encyclopedia of Mathematics - ISBN 1402006098. See original article