an introduction to finite difference calculus
TRANSCRIPT
![Page 1: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/1.jpg)
1
Chapter 3
An Introduction to
Finite Difference Calculus
First Session Contents:
1) Approximation of Derivatives
2) Order Symbols
3) High-Order Derivatives
4) Richardson Extrapolation
![Page 2: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/2.jpg)
2
Discretization of Computational Domain
![Page 3: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/3.jpg)
3
Taylor Series Expansion
Brook Taylor
Born: August 18, 1685
Municipal Borough of Edmonton
Died: November 30, 1731
London, United Kingdom
Education: St John's College, Cambridge,
University of Cambridge
Brook Taylor was an English mathematician who is best known
for Taylor's theorem and the Taylor series
![Page 4: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/4.jpg)
4
Taylor Series Expansion
If a function 𝑓(𝑥) is infinitely differentiable at 𝑥 = 𝑥0, we can express:
To find the coefficients, initially, we put 𝑥 = 𝑥0:
Taking the first derivative gives:
So,
![Page 5: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/5.jpg)
5
Taylor Series Expansion
Similarly, we take derivative again:
Putting 𝑥 = 𝑥0 results in:
In this way, we may conclude:
Eventually, we can write Taylor Series Expansion of 𝑓(𝑥) at 𝑥 = 𝑥0 as:
where
![Page 6: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/6.jpg)
6
Taylor Series Expansion
Now, we can write Taylor Series Expansion for other functions.
since
So, we have:
As Taylor Series Expansion of sin(𝑥) and cos(𝑥) at 𝑥 = 𝑥0.
![Page 7: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/7.jpg)
7
Taylor Series Expansion
Also, we already know
So, the its Taylor Series Expansion is:
To give another example, recall that:
Hence:
![Page 8: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/8.jpg)
8
Order Symbols
Instead of saying that sin(𝑥) tends to zero at the same rate that 𝑥 tends to zero,
we say:
In general:
If
Big “Oh”
as
as
and
as
![Page 9: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/9.jpg)
9
Order Symbols
For example:
Other examples for 𝑥 → 0
as
since
![Page 10: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/10.jpg)
10
Finite Difference Calculus
Let’s write Taylor Series Expansion for 𝑓(𝑥 + ℎ) at 𝑥
So,
Collecting all terms of 𝑂 ℎ :
Let’s re-write based on index notation:
Finite Difference Truncation Error
Forward Difference
![Page 11: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/11.jpg)
11
Finite Difference Calculus
Defining “Forward Difference” Operator:
We have:
NOTE:
Truncation error is the difference between the derivative and its finite difference
Approximation.
For the “Forward Difference”:
Truncation Error
as
Limited
![Page 12: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/12.jpg)
12
Finite Difference Calculus
Similarly for 𝒇(𝒙 − 𝒉) at P we have
So,
Collecting all terms of 𝑂 ℎ :
Let’s re-write based on index notation:
Finite Difference Truncation Error
Backward Difference
![Page 13: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/13.jpg)
13
Finite Difference Calculus
Defining “Backward Difference” Operator:
We have:
Truncation Error
![Page 14: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/14.jpg)
14
Finite Difference Calculus
Let’s re-write both Taylor Series Expansion at P
So,
We may write 𝑓′(𝑥) explicitely:
Finite Difference Truncation Error
![Page 15: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/15.jpg)
15
Finite Difference Calculus
Finite Difference Truncation Error
Central Difference
Collecting all terms of 𝑂 ℎ2 :
Based on index notation we have:
![Page 16: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/16.jpg)
16
Higher-Order Derivatives
Let’s take a look at these two Taylor Series:
So, we have:
× (−𝟐)
Solving for 𝑓′′(𝑥) yields:
or
![Page 17: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/17.jpg)
17
Higher-Order Derivatives
Recall:
So,
Therefore, we can write forward difference for 𝑓′′(𝑥) in operator notation
Similarly, backward difference for 𝑓′′(𝑥) in operator notation would be:
![Page 18: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/18.jpg)
18
Higher-Order Forward and Backward Difference
Recall:
Solving for 𝑓′ 𝑥 gives:
Simplifying of this equation results in:
or:
![Page 19: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/19.jpg)
19
Forward Difference O(h)
Backward Difference O(h)
Central Difference O(h2)
Finite Difference Discretization
![Page 20: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/20.jpg)
20
Forward Difference O(h)
Backward Difference O(h)
Central Difference O(h4)
Finite Difference Discretization
![Page 21: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/21.jpg)
21
Example1
If 𝑓 𝑥 = 𝑒𝑥 find 𝑓′(1) using forward difference, choosing ℎ = 0.1
Using Central Difference:
Forward Difference
Central Difference
Exact Forward Difference Central
Difference
Value 2.718282 2.85844 2.72282
Relative Error - 5.15% 0.17%
![Page 22: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/22.jpg)
22
Example1
Now, let’s choose ℎ = 0.05
Exact value at 𝑥 = 1 Slope of line in 𝑓′-ℎ2 coordinate system is:
So, we can find exact value of 𝑓′ x = 1 :
![Page 23: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/23.jpg)
23
Example2
If 𝑓 𝑥 = sin(𝑥) find 𝑓′(1) using central difference, choosing ℎ = 0.2
Let’s choose ℎ = 0.1
Let’s choose ℎ = 0.05
![Page 24: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/24.jpg)
24
Example2
The exact answer is:
Using the method introduced in previous example:
Which gives:
![Page 25: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/25.jpg)
25
Example3
Consider 𝑓 𝑥 = sin(10𝜋𝑥), find 𝑓′ 0 by choosing ℎ = 0.2
The exact answer is:
So,
Forward Difference
Central Difference
![Page 26: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/26.jpg)
26
Example3
Consider 𝑓 𝑥 = sin(10𝜋𝑥), find 𝑓′ 0 by choosing ℎ = 0.2
The exact answer is:
So,
Forward Difference
Central Difference
Problem?
𝑇 = ℎ
![Page 27: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/27.jpg)
27
Richardson Extrapolation
Born: October 11, 1881
Died: September 30, 1953
Newcastle upon Tyne, United Kingdom
Kilmun, United Kingdom
Is it possible to obtain the exact
solution from numerical solution?
He was an English mathematician, physicist, meteorologist, psychologist and
pacifist who pioneered modern mathematical techniques of weather forecasting.
Lewis Fry Richardson
Question
Education: Bootham School, Newcastle University, King's
College, Cambridge, University of London,
Durham University
![Page 28: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/28.jpg)
28
xo x0+h x0+2h x0+nh
yn
xo
x0+h
x0+2h x0+2nh
y2n
![Page 29: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/29.jpg)
29
Richardson Extrapolation
Analytical
Solution
Numerical
Solution T.E.
∝ is constant
xo
x0+h
x0+2h x0+nh
yn
xo
x0+h/2
x0+2(h/2) x0+2n(h/2)
y2n
![Page 30: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/30.jpg)
30
Richardson Extrapolation
Regardless of terms, C can be obtained as
By Replacing, C in
we have
Consequently, the error for step size of h/2 can be determined as
![Page 31: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/31.jpg)
31
Richardson Extrapolation Second order approximation
Higher order approximation
![Page 32: An Introduction to Finite Difference Calculus](https://reader034.vdocument.in/reader034/viewer/2022042112/62573d7ab1222f5278416c48/html5/thumbnails/32.jpg)
32