0% developed

Using High Order Finite Differences/Definitions and Basics

From Wikibooks, open books for an open world
Jump to navigation Jump to search


Definitions and Basics[edit | edit source]

Vector Norms[edit | edit source]

Definition of a Vector Norm[edit | edit source]

The most ordinary kind of vectors are those consisting of ordered n-tuples of real or complex numbers. They may be written in row                       

or column     

forms. Commas or other seperators of components or coordinates may or may not be used. When a vector has many elements, notations like

  or     are often used.

A most popular notation to indicate a vector is  .

Vectors are usually added component-wise, for

   and   ,

 .

Scalar multiplication is defined by   .

A vector norm is a generalization of ordinary absolute value   of a real or complex number.

For   and  vectors, and  ,  a scalar, a vector norm is a real value   associated with a vector for which the following properties hold.

.

Common Vector Norms[edit | edit source]

The most commonly used norms are:

.

Any two norms on dimensional vectors of complex numbers are topologically equivalent in the sense that, if   and   are two different norms, then there exist positive constants   and   such that  .

Inner Product of Vectors[edit | edit source]

The inner product (or dot product), of two vectors

   and   ,

is defined by

 ,

or when   and   are complex valued, by

 

.

(1.3.3.0)


It is often indicated by any one of several notations:

   or    .

Besides the dot product, other inner products are defined to be a rule that sssigns to each pair of vectors  ,  a complex number with the following properties.





for    is real valued and positive

and

An inner product defines a norm by


 .

(1.3.3.1)


Inequalities Involving Norms[edit | edit source]

The Cauchy Schwarz and Holder's inequalities are commonly employed.



   for 

Matrix Norms[edit | edit source]

Definition of a Matrix Norm[edit | edit source]

The most ordinary kind of matices are those consisting of rectangular arrays of real or complex numbers. They may be written in element form           and be considered as collections of column  or  row  vectors.

Matrices are usually added element-wise, for



Scalar multiplication is defined by   .

The notation   means that all elements of   are identically zero.

A matrix norm is a generalization of ordinary absolute value   of a real or complex number, and can be considered a type of a vector norm.

For   and  matrices, and  ,  a scalar, a matrix norm is a real value   associated with a matrix for which the following properties hold.

.

Common Matrix Norms[edit | edit source]

The most commonly used norms are:

.

Like vector norms any two matrix norms on matrices of complex numbers are topologically equivalent in the sense that, if   and   are two different norms, then there exist positive constants   and   such that  .

The norm   on matrices   is an example of an induced norm. An induced norm is for a vector norm   defined by

.

This could cause the same subscript notation to be used fot two different norms, sometimes.

Positive definite matrices[edit | edit source]

 matrix    is said to be positive definite if for any vector  



for some positive constant   , not depending on   .

The property of being positive definite insures the numerical stability of a variety of common numerical techniques used to solve the equation   .

Taking    then

 .

so that

     and      .

This gives

and

 .

Consistency of Norms[edit | edit source]

A matrix norm    and a vector norm    are said to be consistent when

 .

When    is the matrix norm induced by the vector norm    then the two norms will be consistent.

When the two norms are not consistent there will still be a positive constant    such that

 .

Finite Difference Operators[edit | edit source]

approximation of a first derivative[edit | edit source]

The defition of the derivative of a function gives the first and simplest finite difference.

.

So the finite difference 

can be defined. It is an approximation to   when  is near .

The finite difference approximation   for   is said to be of order  ,  if there exists   such that

,

when  is near .

For practical reasons the order of a finite difference will be described under the assumption that   is sufficiently smooth so that it's Taylor's expansion up to some order exists. For example, if

then

so that

,

meaning that the order of the approximation of   by   is  .

The finite difference so far defined is a 2-point operator, since it requires 2 evaluations of . If

then another 2-point operator

can be defined. Since

,

this   is of order 2, and is referred to as a centered difference operator. Centered difference operators are usually one order of accuracy higher than un-centered operators for the same number of points.


More generally, for   points   a finite difference operator

is usually defined by choosing the coefficients   so that   has as high of an order of accuracy as possible. Considering

.

Then

.

where the   are the right hand side of the Vandermonde system

and

.

When the   are chosen so that 

then

.

so that the operator is of order  .

At the end of the section a table for the first several values of  ,  the number of points, will be provided. The discussion will move on to the approximation of the second derivative.

approximation of a second derivative[edit | edit source]

The definition of the second derivative of a function

.

used together with the finite difference approximation for the first derivative

gives the finite difference 

In view of

for the operator just defined

.

If instead, the difference operator

is used

If the other obvious possibility is tried

In view of

,

.

So  

is a second order centered finite difference approximation for  .


The reasoning applied to the approximation of a first derivative can be used for the second derivative with only a few modifications.

For   points   a finite difference operator

is usually defined by choosing the coefficients   so that   has as high of an order of accuracy as possible. Considering

.

Then

.

where the   are the right hand side of the Vandermonde system

and

.

When the   are chosen so that 

then

.

so that the operator is of order  .

The effect of centering the points will be covered somewhere below.

approximation of higher derivatives[edit | edit source]

Although approximations to higher derivatives can be defined recursively from those for derivatives of lower order, the end result is the same finite difference operators. The Vandermonde type system will be used again for this purpose.

.

The number of points needed to approximate   by finite differences is at least  .


For   points   a finite difference operator

is usually defined by choosing the coefficients   so that  approximates   to as high of an order of accuracy as possible. Considering

.

Then

.

where the   are the right hand side of the Vandermonde system

and

.

When the   are chosen so that 

then

.

so that the operator is of order  .

An alternative analysis is to require that the finite difference operator differentiates powers of   exactly, up to the highest power possible.

effect of the placement of points[edit | edit source]

Usually the   are taken to be integer valued, since the points are intended to coincide with those of some division of an interval or 2 or 3 dimensional domain. If these points and hence   are chosen with only accuracy in mind, then a higher accuracy of only one order can be achieved.

So start by seeing how high is the accuracy that   can be approximated with three points.

Then accuracy of order 4 can not be achieved, because it would require the solution of

which can not be solved since the matrix

is non-singular. The possibility of an   being  can be ruled out otherwise.

For accuracy of order 3

So the matrix

is singular and   are the roots of some polynomial 

 .

Two examples are next.




To see what the accuracy that   can be approximated to with three points.

Then accuracy of order 3 can not be achieved, because it would require the solution of

which can not be solved since the matrices

     and     

would both need to be singular.

If the matrix

is singular, then   are the roots of some polynomial 

 ,

implying

meaning that elementary row operations can transform

to

which is non-singular.

Conversely, if   are the roots of some polynomial 

 , then

can be solved and    approximated to an order of 2 accuracy.

See how high is the accuracy that   can be approximated with   points.

Then accuracy of order   can not be achieved, because it would require the solution of

which can not be solved since the matrix

is non-singular. The possibility of an   being  can be ruled out otherwise, because, for example, if  ,  then the non-singularity of the block

would force  .

For accuracy of order 

So the matrix

is singular and   are the roots of some polynomial 

 .

The progression for the second, third, ... derivatives goes as follows.

If   are the roots of some polynomial 

 

then the system

can be solved, and

approximates   to an order of accuracy of  .

If   are the roots of some polynomial 

 

then the system

can be solved, and

approximates   to an order of accuracy of  .

Now, the analysis is not quite done yet. Returning to the approximation of  .  If for the polynomial

it were that  ,  then the system can be solved for one more order of accuracy. So the question arises as to whether polynomials of the form

exist that have   distinct real roots. When   there is not. So consider  .

If   has 4 distinct real roots, then

has 3 distinct real roots, which it does not. So the order of approximation can not be improved. This is generally the case.

Returning to the approximation of  .  If for the polynomial

it were that  ,  then the system can be solved for one more order of accuracy. So the question arises as to whether polynomials of the form

exist that have   distinct real roots.

If   has   distinct real roots, then

has   distinct real roots, which it does not. So the order of approximation can not be improved.

For functions of a complex variable using roots of unity, for example, may obtain higher orders of approximation, since complex roots are allowed.

centered difference operators[edit | edit source]

For   points   a finite difference operator

is said to be centered when the points are symmetrically placed about  .

When   is odd  .

To find the centered difference operators, consider

.

Then

.

where the   are the right hand side of the over-determined system

and

.

When the   are chosen so that 

then

.

so that the operator is of order  .

Since, for the centered case, the system is over-determined, some restriction is needed for the system to have a solution. A solution occurs when the   are the roots of a polynomial

with

.

Observing that when   is even


and when   is odd

.

So when   is even   has   for all odd  and when   is odd   has   for all even  .

So a centered difference operator will achieve the one order extra of accuracy when the number of points   is even and the order of the derivative   is odd or when the number of points   is odd and the order of the derivative   is even.


Shift Operators[edit | edit source]

Shift of a Function[edit | edit source]

Trigometric polynomials[edit | edit source]

Let 

 

(3.2.0)

be trigometric polynomial defined on  .

Define the inner product on such trigometric polynomials by


 

.

(3.2.1)


In light of the orthogonalities

,

and       when  ,

inner products can be calculated easily.


 

.

(3.2.2)


and for

the inner product is given by


 

.

(3.2.3)


Definition of a shift operator[edit | edit source]

Define the shift operator   on    by

 

.

(3.3.0)

Since

and

,

so that


 

.

(3.3.1)


Approximation by trigometric polynomials[edit | edit source]

Let    be a function defined on and periodic with respect to the interval  .  That is  .

The    degree trigometric polynomial approximation to    is given by

where


 

     and      .

(3.4.0)


  approximates    in the sense that

is minimized over all trigometric polynomials, of degree    or less, by  .

In fact



.

The term in the center



,

so that


 



.

(3.4.1)


 

linearity property

(3.4.2)


If    and    are the    degree trigometric polynomial approximations to    and  ,  then the    degree trigometric polynomial approximation to    is given by  .

This follows immediately from (3.4.0) since

     and      .

Fundamental Property[edit | edit source]

 

(3.5.0)


Generally if    is the    degree trigometric polynomial approximation to a function  periodic on  ,  then    is the    degree trigometric polynomial approximation to  .

To see this calculate the trigometric polynomial approximations for  .









,

where

     and      .









.

Comparing the results with (3.3.1) finishes the observation.

Another detail of use is



which is


 

.

(3.5.1)


Error Estimation[edit | edit source]

A result used in error estimation is

 .

(3.6.0)


When   is a sine polynomial

then


 

(3.6.1)


and


 .

(3.6.2)


Simple Functions[edit | edit source]

Let 

so that

is a partition of  .

The function   is said to be simple on  ,  if


 

(3.7.0)


Of particular interest is when the points have equal spacing  .

The intent is to make estimates of  .

Begin by making an odd extension of    to   by setting    and continue the definition by extending    periodically.

Then approximate    with a sine polynomial

where

.

When   is large enough so that some    are divided by   then, for  

,

and letting  ,

 ,

so that

 .

Now, return to the sum

with    for  .

If    and  ,  then  ,  and in this case

  and  .

So for     with   ,



 .

(3.7.1)


Next observe that if    is the    degree sine polynomial approximation to    then    is the    degree trigometric polynomial approximation to  .  The assumption that    is odd and periodic is still in effect.

Finally the intended results follow.









so that

.

Making use of (3.7.1)

.

Being that simple functions can be approximated by trigometric polynomials to arbitrary accuracy,


 

.

(3.7.2)


Now, for  ,  and using the definition of the simple function  

To find the sum for    list tthe values of    over the values of    on the whole interval  .

This gives

.

So the inequality follows


 

.

(3.7.3)



Specialized Norms[edit | edit source]

Energy and Heat Norms[edit | edit source]

There are a number of less well known, but important norms. These norms are important in the analysis of many physical problems and are used in error estimation for finite difference and finite element methods. Examples are the energy and heat norms.

These norms are usually expressed in an integral form.

When     the inequality below holds.

This follows from a completely elementary, but lengthy calculation, as shown in appendix a). When additional assumptions on   are made this inequality can be improved somewhat.

See appendix b) for an explanation.

Discrete Heat Norms[edit | edit source]

In the analysis of finite difference methods for partial differential equations it is useful to have discrete analogs of norms like the energy and heat norms.

In an attempt not to have notation too cumbersome some indices are suppressed when from the discussion it is clear what they refer to. When   is a   dimensional vector of complex numbers, finite difference operators are defined as discrete approximations or analogs of derivatives.

The most appropriate definition of a discrete energy or heat norm may vary due to differences in the handling of initial or boudary conditions. So for this reason the reader should make appropriate adjustments, when needed, to apply the same reasoning to another problem.

Before the discrete versions of energy or heat norms can be defined,  finite differece operations need to be defined and explained. This was done in the section on finite difference operators.


The next discrete version of the preceding inequality has important applications to the estimation of the error when approximating a second derivative with a finite difference operator.

If    then

with

and generally the following under-estimate holds.

See appendix c) for a proof.

The inequality can be improved for general    by using (3.7.3) with    increased by 2.

If    then

and

.


appendix a)[edit | edit source]

When     the inequality below holds.

First apply the Cauchy Schwarz inequality.

,

Next observe the integrals on the right are increasing with  .

Integrate, make cancellations, and reapply the first inequality.

 .

 .

 .

After integrating again the inequality is improved.

 .

 .

Now, assume the inequality immediately below holds for some  .

After substituting the inequality above into the next

 .

the following observations are made.

 .

 .

 .

 .

 .

Repeating this iteration leads to a sequence 

which converges to  ,  the solution of  .

So:



and

 .

For   ,

 .

 .

To handle the part 

,



 .

 .

 .

Reasoning as before this inequality can be strengthend to

 .

 .

 .

Adding the results of the two main calculations

 .

appendix b)[edit | edit source]

When     the inequality below holds.


Assume the conditions that   can be approximated by a trigometric polynomial

.

such that   is also approximated by the polynomials derivative

That is to say, given  ,  there exixts  a trigometric sine polynomial  ,  such that

  and   .

Now,



  and  

.

So it is easy to see that

.

In this case the inequality is sharp, since it holds with equality for  .

Since

,

and  

,

the inequality holds for   and  .

appendix c)[edit | edit source]

If    then

with

and generally the following under-estimate holds.

The cases for 1 and 2 are simple.

and

with equality, when  .

To prove the general under-estimate do as follows.

Use    and apply the Cauchy Schwartz inequality together with inequality ().

(1)

So for  ,

and

.

After inserting the inequality above into (1)

.

So for  ,

and after using formula ()

.

Using    the right end of the sum is estimated using nearly the same procedure.

So for  ,

,

,

and

.

So for  ,

,

.

Combining the two results:

.

When   is odd, use    so that the inequality becomes

.

When   is even, use    so that the inequality becomes

.

Special Formulas[edit | edit source]

The rule named L´Hospital's Rule named by a French math teacher who published it without proof, because he said you need to be taken to the hospital if you can't prove it yourself, states.

If    and    then,

.

The rule named Leibniz's Rule states how to differentiate a product of two functions    times.

There are a number of ways to find the formulas for the sums of the powers of the first    integers. One of the most convenient is as follows. Begin with the geometric formula.

By convention  .

Apply Leibniz's Rule rule to the right hand side of the equivalent identity

to differentiate both sides    times.

Upon noticing that only the first two terms of the sum on the right are non-zero

.

Now take the limit as    to establish the formula

.

From term by term differentiation.

and

.

More generally

.

and thus

.

A progression of formulas follow by setting  .

.

.

and

.

After adjusting the index of summation and increasing    by    the progression has the form

.

Letting  

.


The formula called  summation by parts  is a discrete analog of  integration by parts . It is verified simply by inspection.

One form of the statement of the Fundamental Theorem of Integral Calculus is

For    it holds  .

Another rule which slightly generalizes the Fundamental Theorem of Integral Calculus says

.


The following simple inequality can be used to bound sums of products.

and also

,

so that

.

Letting    and   

.

The particular case with    gives

.

save scratch[edit | edit source]



Now, suppose for some 

So for 


zzzzzzzzzz

So for 

xxxxxxxxxxxxxxxx