PYTHON FOR M.

S STUDENTS

V.N.Purushothaman

Department of Physi s

Sree Kerala Varma College, Tri hur

31-03-2013
2
Contents

1 Unit-1: Basi s of Python language 1

1.1 Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.3 Variables and data types . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.4 operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.5 Strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.6 Mutable and Immutable Types . . . . . . . . . . . . . . . . . . . . . . 5

1.7 Conditional Exe ution . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.8 Iteration and looping . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.9 Fun tions and Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.10 File input and Output . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

1.11 Pi kling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

1.12 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2 Advan ed Python Programming 29

2.1 NumPy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.1.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3
4 CONTENTS

2.1.2 Array reation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.1.3 Array modi ation . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.1.4 Printing arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.1.5 Saving and restoring arrays . . . . . . . . . . . . . . . . . . . . 35

2.1.6 Indexing, Sli ing and Iterating . . . . . . . . . . . . . . . . . . . 39

2.1.7 Arrays as matri es . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.1.8 Arrays as polynomial oe ients . . . . . . . . . . . . . . . . . 44

2.1.9 Linear Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3 Plotting and visualization 49

3.1 Matplotlib . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.1.1 The Matplotlib Module . . . . . . . . . . . . . . . . . . . . . . . 49

3.1.2 Plotting mathemati al fun tions . . . . . . . . . . . . . . . . . . 55

4 Numeri al Analysis 61
4.1 Numeri al methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4.1.1 Inverse of a fun tion . . . . . . . . . . . . . . . . . . . . . . . . 61

4.1.2 Interpolation with Cubi Spline . . . . . . . . . . . . . . . . . . 66

4.1.3 Zeros of polynomials . . . . . . . . . . . . . . . . . . . . . . . . 70

4.1.4 Monte Carlo Methods . . . . . . . . . . . . . . . . . . . . . . . 77

4.1.5 Sampled Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

4.1.6 Dis rete Fourier Transform . . . . . . . . . . . . . . . . . . . . 82

4.1.7 Fast Fourier Transform(FFT) . . . . . . . . . . . . . . . . . . . 84

CONTENTS 5

4.1.8 Shooting method . . . . . . . . . . . . . . . . . . . . . . . . . . 85

4.1.9 Relaxation method: . . . . . . . . . . . . . . . . . . . . . . . . 87

5 Simulations 89
5.1 A omputational approa h to physi s . . . . . . . . . . . . . . . . . . . 89

5.1.1 Simple harmoni os illator . . . . . . . . . . . . . . . . . . . . . 90

5.1.2 Central eld motion . . . . . . . . . . . . . . . . . . . . . . . . 92

5.1.3 Monte-Carlo simulations- value of π . . . . . . . . . . . . . . . 95

5.1.4 Logisti map . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

5.1.5 Driven LCR ir uit . . . . . . . . . . . . . . . . . . . . . . . . 100

5.1.6 Standing waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

5.1.7 Simulation of radioa tivity . . . . . . . . . . . . . . . . . . . . . 105

6 CONTENTS
CONTENTS i

Prefa e

This is a rough olle tion of the le ture notes I had prepared to tea h the ourse
PHY2C08: COMPUTATIONAL PHYSICS pres ribed for M.S physi s students of
olleges aliated to Cali ut University. Due to la k of time, it is done at a terri pa e
whi h might have aused a few mistakes here and there. Shortage of explanations and
examples is another asualty of su h hurry. I hope that it may be useful in some small
way to students and tea hers. Please be kind enough to inform me when you ome
a ross mistakes in the ideas or language used in this monograph. Purushothaman.V.N,
Department of Physi s,
Sree Kerala Varma College, Tri hur
email:vadakkedamredimail. om
Mob: 9446723810
ii CONTENTS
Chapter 1

Unit-1: Basi s of Python language

1.1 Inputs

A program is a set of statements used to produ e an output from the input data.
Numbers (real and omplex), are read from terminal using the fun tion input('Prompt')
. Strings are read using the fun tion raw_input('prompt'). For example

>>> b=input('Give a number: ')

Give a number: 10
>>>> b
10
>>> a=raw_input('Give a text: ')
Give a text: hopeless
>>> a
'hopeless'

1.2 Outputs

The output of a program an be a number, text or graphi s. For text and numbers print
statement is employed. For graphi output fun tions like show(),saveg(),imshow() et
are dened in relevant modules.

>>> x=5
>>> print x

1
2 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

5
>>> y=[2,5,7,9℄
>>> print y
[2, 5, 7, 9℄
>>> z='beamer'
>>> print z
beamer
>>> print x,y,z
5 [2, 5, 7, 9℄ beamer

Formatted output is possible just as in -language. The general form of format string
is %m.nx where m is an integer showing the total width to be used for printing, n is
an integer representing the number of de imal pla es to be used while printing oating
point numbers so that |m| ≥ 1 + n(de imal point+de imal pla es) and x is c for single
hara ter, f for oat, e for oat in s ienti format, s for string, x for hexade imal, o
for o tal, d or i for integer and 0d for integer with zeros on the left to ll the width.

>>> print '1)%5d 2)%5i 3)%05d '%(23,23,23)

1) 23 2) 23 3)00023
>>> print '1)%4 '%('z')
1) z
>>> print '1)%12s, 2)%-12s'%('zoology','zoology')
1) zoology, 2)zoology
>>> print '1)%12.5f, 2)%-12.5f, 3)%12.7f'%(24.5,24.5,24.5)
1) 24.50000, 2)24.50000 , 3) 24.5000000
>>> print '1)%12.5x, 2)%12.5o'%(24.5,24.5)
1) 00018, 2) 00030

1.3 Variables and data types

A omputer program to solve a problem is designed using variables belonging to the

supported data types. Python supports numeri data types like integers, oating point
numbers and omplex numbers. To handle hara ter strings, it uses the String data
type. Python also supports other ompound data types likelists, tuples, di tionaries.
In the previous example, x is numeri , y is a list and z is a string.
1.4. OPERATORS 3

1.4 operators

Operators are fun tionality that do something and an be represented by symbols su h

as + or by spe ial keywords. Operators require some data to operate on and su h
data are alled operands. In x = 2 + 3, 2 and 3 are the operands and '=' and '+' are
operators. The other operators are

or,and,not(Boolean OR,AND,NOT) : returns True or Fals

in(Membership): returns True or Fals,
not in(Non-membership): returns True or False,
<, <=, >, >=, !=, == (Comparisons): returns True or False
|,^,&(Bitwise OR, XOR,AND),
<<, >>(Bitwise Shifting left and right)
+,-,* (Add, Subtra t, Multiply)
/,%,**(divide, reminder, Exponentiation)
+x (Positive), -x(Negative),
~(Bitwise NOT),
x[index℄(Subs ription)

Bitwise operator works on bits and perform bit by bit operation. Assume a = 60 and
b = 13. Now in binary format they will be as follows:

a = 0011 1100,
b = 0000 1101
Binary AND Operator opies a bit to the result if it exists in both operands.
a&b = 0000 1100
Binary OR Operator opies a bit if it exists in either operand.
a|b = 0011 1101
Binary XOR Operator opies the bit if it is set in one operand but not both.
a^b = 0011 0001
Binary Ones Complement Operator is unary and has the effe t of 'flipping' bits.
~a = 1100 0011
Binary Left Shift Operator moves left operand by the number of bits spe ified by the
a << 2 will give 1111 0000 whi h is 240
Binary Right Shift Operator moves left operand by the number of bits spe ified by th
a >> 2 will give 0000 1111whi h is 15

Pre eden e Rules of operators: These rules give the sequen e of exe utions of an
expression ontaining more than one operator.
4 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

1. ( ) (anything in bra kets is done rst. Highest pre eden e)

2. ** (exponentiation is done next)

3. -x, +x (unary ±)

4. *, /, %, // (multipli ation, division, remainder after division, su essive division)

5. +, - (addition, subtra tion)

6. relational operators: <, >, <=, >=, !=, ==

7. logi al not

8. logi al and

9. logi al or (Lowest pre eden e)

1.5 Strings

String is a olle tion of same kind of elements ( hara ters). It is a ompound or

olle tion data type. The individual elements of a string an be a essed by indexing.

>>> s = 'hello world'

>>> s[0℄
'h'
>>> s[7℄
'o'
>>> s[5℄
' '
>>> s[-1℄ # will print the last hara ter
'd'

Strings an be added and multiplied by integers.

>>> p,q,r='Eating ','troubles',' meeting '

>>> p+q+r
'Eating troubles meeting'
>>> 2*p+q
'Eating Eating troubles'
1.6. MUTABLE AND IMMUTABLE TYPES 5

1.6 Mutable and Immutable Types

There is one major dieren e between String, tuple, list,di tionary types. List and
di tionary are mutable but string and tuple are immutable. We an hange the value
of an element in a list , add new elements to them and remove any existing element.
This is not possible with String and tuple types.In the ase of sets one variety alled
frozenset is immutable while set is mutable.

List

List is mu h more exible than String. The individual elements an be of any type,
even another list. Lists are dened by en losing the elements inside a pair of square
bra kets and separated by ommas.

>>> l=[2.3,'A',3,'khan'℄
>>> type(l)
<type 'list'>
>>> 2*l
[2.29, 'A', 3, 'khan', 2.29, 'A',3, 'khan'℄
>>> l[3℄=28
>>> l
[2.299, 'A', 3, 28℄

Lists respond to the + ( on atenation) and * (repetition) operators like strings.The

result is a new list.
Python Expression Results Des ription
len([1, 2, 3℄) 3 Length
[1, 2, 3℄ + [4, 5, 6℄ [1, 2, 3, 4, 5, 6℄ Con atenation
['Hi!'℄ * 4 ['Hi!', 'Hi!', 'Hi!', 'Hi!'℄ Repetition
3 in [1, 2, 3℄ True Membership
for x in [1, 2, 3℄: print x, 1 2 3 Iteration

List Methods

A method is a fun tion that is oupled to some obje t, be it a list, a number, a string,
or whatever. In general, a method is alled like this: obje t.method(arguments). If a
is list obje t and max() is a method dened on list lass to nd the maximum value
in the list, then a.max() returns maximum of list a. A method all is like a fun tion
6 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

all, ex ept that the obje t is put before the name of the method method with a dot
separating them. Lists have several methods that an be used to examine or modify
their ontents.

1. append: The append method is used to append an obje t to the end of a list:
>>>lst = [1, 2, 3℄
>>>lst.append(4)
>>>lst
>>>[1,2, 3, 4℄

The same an be a hieved using + operator.

>>> x=[3,5,4℄
>>> x+=[2℄
>>> x
[3, 5, 4, 2℄
2. ount: The ount method ounts the o urren es of an element in a list:
>>>['to', 'be', 'or', 'not', 'to', 'be'℄. ount('to')
>>>2
>>>x = [[1, 2℄, 1, 1, [2, 1, [1, 2℄℄℄
>>>x. ount(1)
>>>x. ount([1, 2℄)
>>>2.
3. extend: The extend method allows you to append several values at on e by
supplying a sequen e of the values you want to append. In other words, your
original list has been extended by the other one:
>>>a = [1, 2, 3℄
>>>b = [4, 5, 6℄
>>>a.extend(b)
>>>a
[1,2, 3, 4, 5, 6℄
This may seem similar to on atenation, but the important dieren e is that the
extend method modies a list without reating a new one. In ordinary on ate-
nation, a ompletely new list is returned:
a = [1, 2, 3℄
b = [4, 5, 6℄
a + b
[1, 2, 3, 4, 5, 6℄
a
[1,2, 3℄
1.6. MUTABLE AND IMMUTABLE TYPES 7

4. index: The index method is used for sear hing lists to nd the index of the
rst o urren e of a value:

>>> knights = ['We', 'are', 'the', 'knights', 'who', 'say', 'ni'℄

>>> knights.index('who')
4
>>> knights.index('herring')
>>> ValueError: list.index(x): x not in list

When you sear h for the word 'who', you nd that it's lo ated at index 4:

>>> knights[4℄
'who'

5. insert: The insert method is used to insert an obje t into a list: > numbers
= [1, 2, 3, 5, 6, 7℄ > numbers.insert(3, 'four') > numbers > [1, 2, 3, 'four', 5,
6, 7℄ As with extend, you an implement insert with sli e assignments:

numbers = [1, 2, 3, 5, 6, 7℄
numbers[3:3℄ = ['four'℄
numbers
[1, 2, 3, 'four', 5, 6, 7℄

6. pop: The pop method removes an element (by default, the last one) from the
list and returns it:

>>> x = [1, 2, 3℄
>>> x.pop()
>>> x
>>> [1, 2℄
>>> x.pop(0)
>>> x
>>> [2℄

The pop method is the only list method that both modies the list and returns
a value.

>>> x = [1, 2, 3℄
>>> x.append(x.pop())
>>> x
>>> [1, 2, 3℄

7. remove: The remove method is used to remove the rst o urren e of a value:
8 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

>>> x = ['to', 'be', 'or', 'not', 'to', 'be'℄

>>> x.remove('be')
>>> x
['to', 'or', 'not', 'to', 'be'℄
>>> x.remove('bee')
>>> ValueError: list.remove(x): x not in list

8. del: To remove a list element, you an use either the del statement if you know
exa tly whi h element(s) you are deleting or the remove() method if not known.
The reverse method reverses the elements in the list.
> x = [1, 2, 3℄ > x.reverse() > x > [3, 2, 1℄ Note that reverse hanges the
list and and saves under the same name. > x = [1, 2, 3℄ > list(reversed(x)) [3,
2, 1℄
9. sort: The sort method is used to sort lists in pla e. Sorting 'in pla e' means
hanging the original list so its elements are in sorted order, rather than simply
returning a sorted opy of the list:

>>> x = [4, 6, 2, 1, 7, 9℄
>>> x.sort()
>>> x
>>> [1, 2, 4, 6, 7, 9℄
>>> x = [4, 6, 2, 1, 7, 9℄
>>> x.sort(reverse=True)
>>> x
>>> [9, 7, 6, 4, 2, 1℄

10. sorted: Another way of getting a sorted opy of a list is using the sorted
fun tion :

>>> x = [4, 6, 2, 1, 7, 9℄
>>> y=sorted(x)
>>> x,y
([4, 6, 2, 1, 7, 9℄, [1, 2, 4, 6, 7, 9℄)

This fun tion an a tually be used on any sequen e, but will always return a list:

>>> sorted('Python')
>>> ['P', 'h', 'n', 'o', 't', 'y'℄

11. len: Gives the number of elements in a list.

12. max: Gives the element having maximum ASCII value in a list.
1.6. MUTABLE AND IMMUTABLE TYPES 9

13. min: Gives the element having minimum ASCII value in a list.
14. mp: This fun tion is the basis for sorting. mp(a, b) returns -1 if a < b, 0 if a
== b and 1 if a > b.
15. list: Converts a string or tuple into a list.
16. sum: Returns the sum of elements in a numeri list.

For example

>>> a=[2,5,'A','a','ab'℄
>>> max(a)
'ab'
>>> a.append('z')
>>> a
[2,5,'A','a','ab','z'℄
>>> max(a)
'z'
>>> min(a)
2
>>> x=[4, 7, 8, 2, 3, 12℄
>>> mp(a,x)
-1
>>> mp(x,a)
1
>>> y=x
>>> y
[4, 7, 8, 2, 3, 12℄
>>> mp(x,y)
0
>>> ='design'
>>> list( )
['d', 'e', 's', 'i', 'g', 'n'℄
>>> d=('j','k',3,5)
>>> list(d)
['j', 'k', 3, 5℄
>>> sum(x)
36

Sli ing: Elements from a list an be sele ted using sli ing operator ':'. If x is a list,
then x[m : n : p] represents the set of elements of x with indi es [mth , (m + p)th , (m +
2p)th , ..] ex luding nth element.
10 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

>>> x=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9℄
>>> x[0:-1:2℄
[0, 2, 4, 6, 8℄
>>> x[1:-1:2℄
[1, 3, 5, 7℄
>>> sum(x[1:-1:2℄)
16
>>> x[0:5:2℄
[0, 2, 4℄
>>> sum(x[0:5:2℄)
6

set:

A set obje t is an unordered olle tion of immutable values. They annot be indexed
by any subs ript. The built-in fun tion len() returns the number of items in a set.
There are urrently two intrinsi set types:

(1)Sets: These represent a mutable set. They are reated by the built-in set()
onstru tor and an be modied afterwards by several methods, su h as add(), lear(),
dis ard().

Frozen sets: These represent an immutable set. They are reated by the built-in
frozenset() onstru tor. As a frozenset is immutable, it an be used as an element
of another set, or as a di tionary key. Common uses of sets in lude membership
testing, removing dupli ates from a sequen e, and omputing mathemati al operations
su h as interse tion, union, dieren e, and symmetri dieren e. The main fun tions
are len(), union(), interse tion(), dieren e(), symmetri _dieren e(), issubset() and
issuperset(). There is also an operator equivalent for many of these fun tions. Let s
and t be two sets. Then

1. x in s: test element x for membership in s (True/False)

2. x not in s: test element x for non-membership in s (True/False)
3. s.issubset(t): [s <= t℄: test whether every element in s is in t
4. s.issuperset(t): [s >= t℄: test whether every element in t is in s
5. s.union(t): [s | t℄: new set with elements from both s and t
6. s.interse tion(t): [s & t℄: new set with elements ommon to s and t
7. s.dieren e(t): [s - t℄: new set with elements in s but not in t
1.6. MUTABLE AND IMMUTABLE TYPES 11

8. s.symmetri _dieren e(t) : [s t℄: new set with elements in either s or t but not
both

9. s. opy(): new set with a opy of s

10. s. lear(): Remove all elements from the set s.

11. s.dis ard(x): Remove element x from set s if it is a member. If x is not a member,
nothing happens.

12. s.update():Update s with the union of itself and others.

>>> a=[1,2.0,6.1,'l'℄
>>> b=set(a)
>>> b
set([1, 2.0, 'l', 6.0999999999999996℄)
>>> =set([1,3,6.1,'k'℄)
>>>
set([1, 'k', 3, 6.0999999999999996℄)
>>> d=set('domain')
>>> d
set(['a', 'd', 'i', 'm', 'o', 'n'℄)
>>> f=set((1,3,5,9.1))
>>> f
set([1, 3, 9.0999999999999996, 5℄)
>>> b|
set([1, 2.0, 3, 'k', 'l', 6.0999999999999996℄)
>>> b&
set([1, 6.0999999999999996℄)
>>> b.differen e( )
set([2.0, 'l'℄)
>>> .differen e(b)
set(['k', 3℄)
>>> .symmetri _differen e(b)
set(['k', 2.0, 3, 'l'℄)
>>> s=set([1,3,5,7,9℄)
>>> s.add(11)
>>> s
set([1, 3, 5, 7, 9, 11℄)
>>> s.remove(5)
>>> s
set([1, 3, 7, 9, 11℄)
12 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

Tuples

Tuples are data stru tures that are very similar to lists, but they annot be modied
(immutable). They an only be reated. Tuples have important roles as keys for
di tionaries. A tuple is a sequen e that is en losed by parentheses ( ). The following
line of ode reates a three-element tuple

>>> x = ('a', 'b', ' ')

Inter onversion between lists and tuples is possible using list() and tuple() fun tions.

>>> list((1, 2, 3, 4))

[1,2, 3, 4℄
>>> tuple([1, 2, 3, 4℄)
(1,2, 3, 4)

Di tionaries

Di tionaries are asso iative arrays. It is a group of {key : value} pairs. The elements
in a di tionary are indexed by keys. Keys in a di tionary are required to be unique.
Keys an be almost any Python type, but are usually numbers or strings. Values, on
the other hand, an be any arbitrary Python obje t. Di tionaries are en losed by urly
bra es - { } and values an be assigned and a essed using square bra es [℄. They are
dierent from sequen e type ontainers like lists and tuples in the method of storing
data. There is no on ept of order among elements. They are unordered.Their main use
in lude storing time of modi ation of les as values and le name as keys, telephone
dire tory with name as value and phone number as key, address book with name as key
and address as value, the oordinate of a point(tuple) as key and its olour as value in
a graphi s reen et . Example for a di tionary is given below.

>>> d t={} #Creates an empty di tionary

>>> d t['host'℄='Earth'
#'host' is the key and 'earth' is the value.
>>> d t
{'host': 'Earth'}
>>> d t['port'℄=80
>>> d t
{'host': 'Earth', 'port': 80}
>>> d t.keys()
1.6. MUTABLE AND IMMUTABLE TYPES 13

['host', 'port'℄
>>> d t.values()
['Earth', 80℄
>>> print d t['host'℄

Di tionary fun tions and methods:

1. mp(di t1, di t2): Compares elements of both di tionaries.

2. len(di t): Gives the total length of the di tionary. This would be equal to the
number of Key-value pairs in the di tionary.

3. str(di t): Produ es a printable string representation of a di tionary

4. type(variable): Returns the type of the passed variable. If passed variable is

di tionary then it would return a di tionary type.

Python in ludes following di tionary methods

1. di t. lear(): Removes all elements of di tionary di t

2. di t. opy(): Returns a shallow opy of di tionary di t

3. di t.fromkeys(): Create a new di tionary with keys from seq and values set to
value.

4. di t.get(key, default=None): For key key, returns value or default if key not in
di tionary

5. di t.has_key(key): Returns true if key in di tionary di t, false otherwise

6. di t.items(): Returns a list of di t's (key, value) tuple pairs

7. di t.keys(): Returns list of di tionary di t's keys

8. di t.setdefault(key, default=None): Similar to get(), but will set di t[key℄=default

if key is not already in di t

9. di t.update(di t2): Adds di tionary di t2's key-values pairs to di t

10. di t.values(): Returns list of di tionary di t's values

14 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

1.7 Conditional Exe ution

The most fundamental aspe t of a programming language is the ability to ontrol the
sequen e of operations. One of this ontrol is the ability to sele t one a tion from a set
of spe ied alternatives. The other one is the fa ility to repeat a series of a tions any
number of times or till some ondition be omes false. To exe ute some se tion of the
ode only if ertain onditions are true python uses if, elif,...,else onstru t.

>>> x = input('Enter a number ')

10
>>> if x>10:
print 'x>10' # Note the Colon and indentation.
elif x<10:
print 'x<10'
else:
print 'x=10'
>>>x=10

1.8 Iteration and looping

when a ondition remains true, if a set of statements are to be repeated, the while
and for onstru ts are employed. The general syntax of a while loop may be given as
follows.
while ondition:
set of statements to be repeated
for elements in list or tuple :
set of statements to be repeated

>>> x=10
>>> while x>0:
print x,
x=x-1
>>>10 9 8 7 6 5 4 3 2 1
>>> for i in range(10,0,-1):
print i,
>>>10 9 8 7 6 5 4 3 2 1
1.9. FUNCTIONS AND MODULES 15

1.9 Fun tions and Modules

A fun tion is a blo k of ode that performs a spe i task. Using a fun tion in a
program is alled ' alling' the fun tion. Python has two tools for building fun tions:
def and lambda. For example, we an build a fun tion that returns the square root of
a number as follows:
(1) def squareroot(x): return math.sqrt(x)
(2) squareroot = lambda x: math.sqrt(x)
(3) g = lambda x: x*2
g(3)=6
(4) (lambda x: x*2)(3)= 6
If a fun tion is used only on e ( alled from only one pla e in your program) Lambda
fun tions are useful and onvenient for two reasons: (1)There is no need to give the
fun tion a name.(2) It an be dened where it is used.

The next method of dening fun tions is illustrated below. For nding the largest
of x,y,z

>>> def large(x,y,z):

if y>x:
x,y=y,x
if z>x:
z,x=x,z
return(x)
>>> large(3,4,2)
4

In Python, the denitions of fun tions, variables, onstants may be saved in a le
and use them in a s ript or in an interpreter just like header les in C-language.
Su h a le is alled a module. The le name is the module name with the sux .py
appended. Within a module, the name of the module is assigned to the global variable
_name_. Denitions from a module an be imported into other modules or into the
main module. Examples of some standard modules are math, os,random, pylab, numpy
et . The main advantage of reating and using modules is that longer programs an
be split into several les so that maintenan e of ode is easy and an be reused in
several programs by in luding the le with the keyword import at the beginning of the
program.
16 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

1.10 File input and Output

Files are used to store data and program for later use. This program reates a new le
named 't.txt' (any existing le with the same name will be deleted) and writes a String
to it. The le is losed and then reopened for reading data. The relevant fun tions are
open, write, read and lose.

>>> f=open('t.txt','w')
>>> f.write('breaking into the file')
>>> f. lose()
>>> f=open('t.txt','r')
>>> f.read()
'breaking into the file'

1.11 Pi kling

Strings an easily be written to and read from a le. Numbers take more eort,
sin e the read() method only returns strings, whi h will have to be onverted into a
number expli itly. However, it is very ompli ated when trying to save and restore
data types like lists, di tionaries et . Rather than onstantly writing and debugging
ode to save ompli ated data types, Python provides a standard module alled pi kle.
Fun tions dump and load are used in pi kle. pi kle.dump(a,f ) will save obje t a to le
f. a=pi kle.load(f ) retrieves data from le f.

Pi kling- Examples

>>>import pi kle
>>>a=10.1
>>>b='sh'
>>> =[5,3,2℄
>>>f = open("state", 'w')
>>>pi kle.dump(a, f)
pi kle.dump(b, f)
pi kle.dump( , f)
file. lose()

>>>file = open("state", 'r')

Reading and writing files
1.12. PROBLEMS 17

a = pi kle.load(file)
b = pi kle.load(file)
= pi kle.load(file)
file. lose()

Any data that was previously in the variables a, b, or is restored to them by

pi kle.load.

1.12 Problems

1. To sort a set of numbers

a=[℄
n=input('Give the ount of numbers n')
for i in range(n):a.append(input('Type the numbers'))
a.sort() #As ending order
print a
a.reverse() # Des ending order
print a

2. Simultaneous arrays: Constru t two 100-element arrays su h that ith element of

one array is sin (2πi/100) and the other cos (2πi/100).

Program:
from math import *
x=[sin(2*pi*i/100) for i in range(1,101)℄
y=[ os(2*pi*i/100) for i in range(1,101)℄
print x,y

3. To reate a triangle of equispa ed stars(*)

Program:
n=input('Howmany rows ? ')
for i in range(n):
print
for j in range(i+1): print '*',

When the program is run

18 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

Howmany rows ? 7

*
* *
* * *
* * * *
* * * * *
* * * * * *
* * * * * * *

4. Fibona i series:

n=input('Number below whi h series is required: ')

a, b = 0, 1
while b < n:
print b,
a, b = b, a+b

5. To read a m × n matrix using list-methods only

m,n=input('order of matrix m,n = ')

a=[℄
for i in range(m):
b=[℄
for j in range(n):
b.append(input('give elements a(i,j)'))
a.append(b)
print 'output: ',a

When the program is run

order of matrix m,n = 2,3

give elements a(i,j)1
give elements a(i,j)2
give elements a(i,j)3
give elements a(i,j)4
give elements a(i,j)5
give elements a(i,j)6
output: [[1, 2, 3℄, [4, 5, 6℄℄

6. To generate values of a fun tion, say, x sin x

import math
x=[0.1*i for i in range(10)℄
1.12. PROBLEMS 19

y=[i*math.sin(i) for i in x℄
for i in range(10):
print '(%0.2f,%0.5f)'%(x[i℄,y[i℄),

The (x, y) values obtained are given below

(0.00,0.00000), (0.10,0.00998), (0.20,0.03973), (0.30,0.08866),

(0.40,0.15577), (0.50,0.23971), (0.60,0.33879), (0.70,0.45095),
(0.80,0.57388), (0.90,0.70499),

7. To nd fa torial of a number

n=input('Give the number whose fa torial is required')

f=1
for i in range(1,n+1):f*=i
print f

8. Permutations

n,r=input('Give n and r in nPr: ')

p=1
for i in range(n,n-r,-1):p*=i
print p

9. ombinations The denition is

n n! n(n − 1)(n − 2)....(n − r + 1) n (n−1)

Cr = = = C(r−1)
(n − r)!r! r(r − 1)(r − 2)....3.2.1 r

def ombination(n,r):
if r==0:return 1
else:return n* ombination(n-1,r-1)/r
n,r=input('Give n and r in nCr: ')
print ombination(n,r)

When program is run

Give n and r in nCr: 6,3

20
>>>
Give n and r in nCr: 7,4
35
20 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

Aliter:
def f(a):
if a==1:return(1)
else: return(a*f(a-1))
n,r=input('Give n and r in nCr: ')
print f(n)/(f(n-r)*f(r))

10. Pas al's triangle

Pin iple: It is the set of binomial oe ients arranged in rows. The, nth row
orresponds to the oe ients of the expansion (a + b)n

def f(a):
if a==0:return(1)
else: return(a*f(a-1))
n=input('Howmany rows ? ')
for i in range(n):
print
for j in range(i+1): print f(i)/(f(i-j)*f(j)),

Aliter:
def ombination(n,r):
if r==0:return 1
else:return n* ombination(n-1,r-1)/r
m=input('Howmany rows ? ')
for i in range(m+1):
print
for j in range(i+1):print ombination(i,j),

when the program is run

Howmany rows ? 7

1
1 1
1 2 1
1 3 3 1
1 4 6 4 1
1 5 10 10 5 1
1 6 15 20 15 6 1

11. Generate the triangular sequen e 0,1,3,6,10,15,21,...n

1.12. PROBLEMS 21

Prin iple: Obviously the numbers are ombinations

 
n n(n − 1)
=
2 1×2

where n = 0, 1, 2, 3...

Program:
m=input('Give maximum number upto whi h the series is required: ')
print [i*(i+1)/2 for i in range(m) if i*(i+1)/2 < m℄

When the program is run

Give maximum number upto whi h the series is required: 150

[0, 1, 3, 6, 10, 15, 21, 28, 36, 45, 55, 66, 78, 91, 105, 120, 136℄

Aliter :
The relation between adja ent elements is

a[0] = 0, n > 0, a[n] = a[n − 1] + (n − 1)

n,a=input('n: '),0
for i in range(n):
a+=i
print a

12. Hailstorm numbers

Prin iple: Pi k any whole number. If it's odd, multiply the number by 3, then
add 1. If it's even, divide it by 2. Now, apply the same rules to the answer that
you just obtained. Do this over and over again, applying the rules to ea h new
answer. Hen e these are the set of numbers obtained by the following rule of
iteration.
If si is even, si+1 = si /2, else si+1 = 3si + 1

Program:
s=input('Seed number for Hailstorm series: ')
while s!=1:
if s%2==0:s/=2
else:s=(3*s+1)
print s,
22 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

When the program is run

Seed number for Hailstorm series: 17

52 26 13 40 20 10 5 16 8 4 2 1

13. To nd largest and smallest in a set of numbers

x=[℄
n=input('Give the ount of numbers')
for i in range(n): x.append(input('Type numbers one at a time '))
print 'Largest of the series is',max(x)
print 'Smallest of the series is',min(x)

14. To solve quadrati equation

from math import *

a,b, =input("Give oeffi ients in the order a,b, seperated by omma. "
d=sqrt(b*b-4*a* )
print "Root 1 = ",(-b+d)/(2*a),"Root 2 = ",(-b-d)/(2*a)

When program is run

Give oeffi ients in the order a,b, seperated by omma. 1,2,3

Root 1 = (-1+1.41421356237j) Root 2 = (-1-1.41421356237j)

15. To verify orthogonality of sine and osine fun tions

Prin iple: If the fun tions are orthogonal, then
Z π π
X
sin θ cos θdθ ≈ sin θi cos θi ≈ 0
0 i=0

from math import*

print sum([sin(pi*i/180)* os(pi*i/180) for i in range(180)℄),'is negligi

Output: -4.85722573274e-15 is negligible

16. To he k whether a given number is prime

1.12. PROBLEMS 23

Prin iple: n is a prime number if it is exa tly divisible only by 1 and n. To

he k this, see whether it is divisible by any number between 2 and n/2 (integer
division).

n=input("Give the number n: ")

m=1+n/2
for i in range(2,m):
if n%i!=0: ontinue
else:
print '%d is not a prime. It is divisible by %d'%(n,i)
break
if i==n/2: print '%d is a prime'%n
When the program is run

Give the number n: 83

83 is a prime
>>>
Give the number n: 245791
245791 is not a prime. It is divisible by 7
17. To nd the ount of prime numbers in a given range.

Prin iple: Even numbers annot be prime. Che k in the set of odd numbers
for prime numbers.

j,k=input("Give the range between whi h prime numbers are required: ")
x=range(j,k,1)
p=[℄
for n in x:
m=n/2
if n>0 and n<4:p.append(n)
for i in range(2,m+1):
if n%i==0: break
if i==m: p.append(n)
print 'There are %d prime numbers in the range(%d, %d). They are'%(len(p),j,k)
print p
When the program is run

Give the range between whi h prime numbers are required: 49,150
There are 20 prime numbers in the range(49, 150). They are
[53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101,
103, 107, 109, 113, 127, 131, 137, 139, 149℄
18. To he k whether a given word is a palindrome.
24 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

Prin iple: A sequen e is a palindrome if it is the same when read from left or
right. For example, xyzzyx,123321 are palindromes.

Program:
x=raw_input('Give the word')
y=[x[-i-1℄ for i in range(len(x))℄
if list(x)==y:print 'palindrome'
else: print 'Not a palindrome'

When the program is run

Give the word: ab 121 ba

palindrome
>>>
Give the word: ele tion
Not a palindrome

19. To nd square root of a number

Prin iple: Let x be the square root of n. Then x2 − n = 0. Now x is the root
of this equation. It an be al ulated using Newton-Raphson method.

f (x) = x2 − n, f ′ (x) = 2x
 
f (xk ) 1 n
xk+1 = xk − ′ = xk +
f (xk ) 2 xk

Program:
n,e=input('Give the number n and a ura y required e: ')
x0,x=0,1
while abs(x-x0)>e:
x0=x
x=(x0+n/x0)/2.0
print 'Square root of %0.3f =%0.5f '%(n,x0)

When the program is run

Give the number n and a ura y required e: 24,.001

Square root of 24.000 =4.89900

20. To nd mth root of a number

1.12. PROBLEMS 25

Prin iple: Let x be the mth root of n. Then xm − n = 0. Now x is the mth
root of this equation. It an be al ulated using Newton-Raphson method.

f (x) = xm − n, f ′ (x) = mxm−1

 
f (xk ) 1 n
xk+1 = xk − ′ = (m − 1)xk + m−1
f (xk ) m xk

Program:

n,m,e=input('Give the number n, order of root m and a ura y e required : ')

x0,x=0.0,1.0
while abs(x-x0)>e:
x0=x
x=((m-1)*x0+n/x0**(m-1))/m
print '%0.2f (th) root of %0.3f = %0.5f '%(m,n,x0)

When the program is run

Give the number n, order of root m and a ura y e required : 65,3,.01

3.00 th root of 65.000 =4.03005
>>>
Give the number n, order of root m and a ura y e required : 65,4,.01
4.00 th root of 65.000 =2.84527
>>>
Give the number n, order of root m and a ura y e required : 2.88,1.5,.01
1.50 th root of 2.880 =2.03016

Aliter : Using bise tion method.

def f(x,m,n):return x**m-n

a,b,m,n,k=input('Intervel (a,b),order of root m, number n and no. of iterations
i=0
while i<k:
=(a+b)/2.0
if f(a,m,n)*f( ,m,n)<0:b=
else:a=
i+=1
print 'The %d root of %f after %d iterations is %20.15f'%(m,n,i, )

21. To onvert temperature in Fahrenheit into entigrade

26 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE

Prin iple: 32o F = 00 C, 212o F = 100o C . Therefore the onversion formulae

are
5
C = (F − 32)
9
9
F = C + 32
5

Program:

t=input('Give the temperature: ')

=input('Is the given temperature is in 1. entigrade or 2.farenheit (1/2
if ==1: print '%0.2f C= %0.3f F'%(t,t*9.0/5+32.0)
else: print '%0.2f F= %0.3f C'%(t,(t-32.0)*5.0/9)

When the program is run

Give the temperature: 25

Is the given temperature is in 1. entigrade or 2.farenheit (1/2): 1
25.00 C= 77.000 F
>>>
Give the temperature: 180
Is the given temperature is in 1. entigrade or 2.farenheit (1/2): 2
180.00 F= 82.222 C

22. To nd value of π

Prin iple: The value of π an be al ulated using tan 45o = tan (π/4) = 1 as
follows.
∞
X (−1)i
π 1 1 1
= tan−1 1 = 1 − + − + ....... =
4 3 5 7 i=0
2i + 1

Program:

x=[(-1)**i/(2.0*i+1) for i in range(100000)℄

pi=4*sum(x)
print 'The value of pi=%0.10f'%pi

When the program is run, the following output is obtained.

The value of pi=3.1415826536

1.12. PROBLEMS 27

Python keywords

Core Python has 30 keywords:

(1)and (2)as (3)break (4) lass (5) ontinue (6)def (7)del (8)elif (9)else (10)ex-
ept (11)nally (12)for (13)from (14)global (15)if (16)import (17)in (18)is
(19)lambda (20)nonlo al (21)not (22)or (23)pass (24)raise (25)return (26)as-
sert (27)try (28)while (29)with (30)yield
28 CHAPTER 1. UNIT-1: BASICS OF PYTHON LANGUAGE
Chapter 2

Advan ed Python Programming

2.1 NumPy

2.1.1 Introdu tion

NumPy's main lass is the homogeneous multidimensional array alled ndarray. This
is a table of elements (usually numbers), all of the same data type. Ea h element
is indexed by a tuple of positive integers. Examples of multidimensional array ob-
je ts in lude ve tors, matri es, spreadsheets et . The term multidimensional refers to
arrays having several dimensions or axes. The number of axes is often alled rank
( not a tensor rank).

For example, the oordinates of a point in 3-D spa e (x, y, z) is an array of rank
1. This also gives the position ve tor of  that point.
 The array ([1., 0., 0.], [0., 1., 2.])
1 0 0
is one of rank 2. It is equivalent to (it is 2-dimensional). The rst
0 1 2
dimension (rows) has a length of 2, the se ond dimension( olumn) has  a length  of 3.
(1 0) 0
The array ([[1., 0.], 0.], [[0., 1.], 2.]) is one of rank 3. It is equivalent to (it
(0 1) 2
is 3-dimensional).

2.1.2 Array reation

There are many ways to reate arrays. For example, you an reate an array from a
regular Python list or tuple using the array fun tion.

29
30 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

>>> a = array( [2,3,4℄ )

>>> a
array([2, 3, 4℄)
>>> type(a) # a is an obje t of the ndarray lass
<type 'numpy.ndarray'>

The fun tion array() transforms sequen es of sequen es into two-dimensional arrays,
and it transforms sequen es of sequen es of sequen es into three-dimensional arrays,
and so on. The type of the resulting array is dedu ed from the type of the elements in
the sequen es.

>>> b = array( [ (1.5,2,3), (4,5,6) ℄ )

>>> b
array([[ 1.5, 2. , 3. ℄,
[ 4. , 5. , 6. ℄℄)

To reate an array whose elements are sequen es of numbers, NumPy provides a fun -
tion arange(x1 , x2 , dx) and returns x1 , x1 + dx, ...., x2 − dx. It is analogous to range
fun tion but a epts oating point numbers also.

>>> arange( 10, 30, 5 )

array([10, 15, 20, 25℄)
>>> arange( 0, 2, 0.3 )
array([ 0. , 0.3, 0.6, 0.9, 1.2, 1.5, 1.8℄)

array and arange are not the only fun tions that reate arrays. Usually the elements
of the array are not known from the beginning, and a pla eholder array(empty array)
is needed. There are some fun tions to reate arrays with some initial ontent. By
default, the type of the reated array is oat64. The fun tion zeros((m,n)) reates a
2-D array of m rows and n olumns with zeros as elements. Similarly the fun tion
ones((m,n)) reates an array full of ones, the fun tion empty((m,n)) reates an array
without lling it in and the fun tion random((m,n)) reates an array lling it with
random numbers between 0 and 1. identity (n) reates an n-dimensional unit matrix.
Then the initial ontent is random and it depends on the state of the memory. In these
fun tions the arguments m, n spe ies the size along ea h axis of the array.

Using arange with oating point arguments, it is generally not possible to predi t
the number of elements obtained be ause of the oating point pre ision. Hen e it is
better to use the fun tion linspace(x1 , x2 , nx) whi h returns equispa ed nx numbers
from x1 to x2 .
2.1. NUMPY 31

The general syntax of these fun tions are empty (shape=, dtype=int) Return an
uninitialized array of data type, dtype, and given shape.

An array of zeros an be reated with a spe ied shape using zeros() fun tion.
zeros(shape=, dtype=) :Return an array of data type dtype and given shape lled with
zeros. An array of ones an be reated with a spe ied shape using ones() fun tion.
zeros(shape=, dtype=) : Return an array of data type dtype and given shape lled with
zeros. an identity matrix an be reated using identity() fun tion
identity (n, dtype=int) : Return a 2-d square array of shape (n,n) and data type, dtype
with ones along the main diagonal.

>>> empty( (2,3) )

array([[ 3.73603959e-262, 6.02658058e-154, 6.55490914e-260℄,
[ 5.30498948e-313, 3.14673309e-307, 1.00000000e+000℄℄)

>>> empty( (2,3) ) # the ontent may hange in different invo ations
array([[ 3.14678735e-307, 6.02658058e-154, 6.55490914e-260℄,
[ 5.30498948e-313, 3.73603967e-262, 8.70018275e-313℄℄)

>>> zeros( (3,4) )

array([[0., 0., 0., 0.℄,
[0., 0., 0., 0.℄,
[0., 0., 0., 0.℄℄)

>>> ones( (2,3,4), dtype=int16 )

array([[[ 1, 1, 1, 1℄,
[ 1, 1, 1, 1℄,
[ 1, 1, 1, 1℄℄,
[[ 1, 1, 1, 1℄,
[ 1, 1, 1, 1℄,
[ 1, 1, 1, 1℄℄℄, dtype=int16)

>>> a=identity(4,dtype=float)
>>> a
array([[ 1., 0., 0., 0.℄,
[ 0., 1., 0., 0.℄,
[ 0., 0., 1., 0.℄,
[ 0., 0., 0., 1.℄℄)

>>> linspa e( 0, 2, 9 )
array([ 0. , 0.25, 0.5 , 0.75, 1. , 1.25, 1.5 , 1.75, 2. ℄)
>>> x = linspa e( 0, 2*pi, 10 )
>>> x
32 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

array([ 0. , 0.6981317 , 1.3962634 , 2.0943951 , 2.7925268 ,

3.4906585 , 4.1887902 , 4.88692191, 5.58505361, 6.28318531℄)
>>> f = sin(x)
array([ 0.00000000e+00, 6.42787610e-01, 9.84807753e-01,
8.66025404e-01, 3.42020143e-01, -3.42020143e-01,
-8.66025404e-01, -9.84807753e-01, -6.42787610e-01,
-2.44921271e-16℄)

Array attributes

The important attributes of any ndarray obje t are:

1. b.ndim :It gives the rank of the array.

2. b.shape :It
returns a tuple of integers indi ating the size of the array in ea h
dimension. For a matrix with m rows and n olumns, shape returns (m, n).
3. b.size . Returns the total number of elements in all dimensions of the array. This
is equal to the produ t of the elements of shape ommand.
4. b.dtype Returns the data type of the elements in the array. NumPy provides the
following datatypes: bool, character, int, int8, int16, int32, int64, f loat, f loat8, f loat16, f
5. b.itemsize :Returns the size in bytes of ea h element of the array.For example,
an array of elements of type oat64 has itemsize 8 (=64/8), while one of type
omplex32 has itemsize 4 (=32/8).

6. b.data : Returns the buer ontaining the a tual elements of the array.

Example for these methods : We dene the following array:

>>> from numpy import *

>>> a = array([(0, 1, 2),(3, 2, 1)℄,)
>>>a.shape, a.ndim, a.size, a.itemsize, a.dtype
( (2, 3), 2, 6, 4, 'dtype('int32') )

The type of the array an also be expli itly spe ied at reation time:

>>> = array( [ [1,2℄, [3,4℄ ℄, dtype= omplex )

>>>
2.1. NUMPY 33

array([[ 1.+0.j, 2.+0.j℄,

[ 3.+0.j, 4.+0.j℄℄)
>>> .dtype
dtype(' omplex128')

A frequent error onsists in alling array with multiple numeri arguments, rather than
providing a single list of numbers as an argument.

>>> a = array(1,2,3,4) # WRONG be ause numbers within () are taken as arguements.

>>> a = array([1,2,3,4℄) \# RIGHT be ause [1,2,3,4℄ is a single list.

2.1.3 Array modi ation

The shape of an array an be hanged with various ommands:

ravel(),transpose(),reshape(m,n,...),resize(m,n,...)
Here (m, n, ....) is the size of the multidimensional array. For example

>>> from numpy import*

>>> a=array([[ 0, 1, 2, 3℄,
[ 4, 5, 6, 7℄,
[ 8, 9, 10, 11℄℄)
>>> a.ravel()
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11℄)
>>> a
array([[ 0, 1, 2, 3℄,
[ 4, 5, 6, 7℄,
[ 8, 9, 10, 11℄℄) # No permanent hange to shape
>>> a.reshape(4,3)
array([[ 0, 1, 2℄,
[ 3, 4, 5℄,
[ 6, 7, 8℄,
[ 9, 10, 11℄℄)
>>> a
array([[ 0, 1, 2, 3℄,
[ 4, 5, 6, 7℄,
[ 8, 9, 10, 11℄℄) # No permanent hange to shape
>>> a.resize(4,3)
>>> a
array([[ 0, 1, 2℄,
[ 3, 4, 5℄,
34 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

[ 6, 7, 8℄,
[ 9, 10, 11℄℄)
>>> a
array([[ 0, 1, 2℄,
[ 3, 4, 5℄,
[ 6, 7, 8℄,
[ 9, 10, 11℄℄) # Permanent hange to shape
>>> a.transpose()
array([[ 0, 3, 6, 9℄,
[ 1, 4, 7, 10℄,
[ 2, 5, 8, 11℄℄) # No permanent hange to shape

The reshape fun tion returns its argument with a modied shape, whereas the resize
method modies the array itself:

2.1.4 Printing arrays

When you print an array, NumPy displays it in a similar way to nested lists, but with
the following layout:

1. the last dimension is printed from left to right,

2. the last but one, from top to bottom,

3. and the rest, also from top to bottom, separating ea h sli e by an empty line.

One dimensional arrays are then printed as rows, two dimensional as matri es and
three dimensional as lists of matri es.

>>> a = arange(6)
>>> print a
[0 1 2 3 4 5℄
>>>
>>> b = arange(12).reshape(4,3)
>>> print b
[[ 0 1 2℄
[ 3 4 5℄
[ 6 7 8℄
[ 9 10 11℄℄
>>>
2.1. NUMPY 35

>>> = arange(24).reshape(2,3,4)
>>> print
[[[ 0 1 2 3℄
[ 4 5 6 7℄
[ 8 9 10 11℄℄
[[12 13 14 15℄
[16 17 18 19℄
[20 21 22 23℄℄℄

If an array is too large to be printed, NumPy automati ally skips the entral part of
the array and only prints the orners:

>>> print arange(10000)

[ 0 1 2 ..., 9997 9998 9999℄
>>>
>>> print arange(10000).reshape(100,100)
[[ 0 1 2 ..., 97 98 99℄
[ 100 101 102 ..., 197 198 199℄
[ 200 201 202 ..., 297 298 299℄
...,
[9700 9701 9702 ..., 9797 9798 9799℄
[9800 9801 9802 ..., 9897 9898 9899℄
[9900 9901 9902 ..., 9997 9998 9999℄℄

2.1.5 Saving and restoring arrays

The simplest way to store arrays is to write it to a text le as text using the numpy
fun tion savetxt(). The array an be retrieved using the fun tion genfromtxt(). The
syntax of these fun tions are
savetxt(fname,array,fmt= ,delimiter=)
Here fname is the name of the le to be reated and opened for writing, array is the
name of the array, fmt is the format spe i ation of the data to be stored, delimiter is
the hara ter used to distinguish elements of the array.

genfromtxt(fname,dtype=, omments=# ,delimiter= ,skiprows= ). Here dtype is the

datatype of array elements and skiprows a epts a number whi h refers to the number
of rows to skip from 0th row.

>>> b
36 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

array([[ 0., 1., 2., 3.℄,

[ 4., 5., 6., 7.℄,
[ 8., 9., 10., 11.℄,
[ 12., 13., 14., 15.℄℄)
>>> savetxt('f1.txt',b,fmt='%8.6f',delimiter='&')
>>> savetxt('f2.txt',b,fmt='%8.4f',delimiter=' ')

When f1.txt and f2.txt are opened in a text editor, the ontents of the file
f1.txt
0.000000 &1.000000 &2.000000 &3.000000
4.000000 &5.000000 &6.000000 &7.000000
8.000000 &9.000000 &10.000000&11.000000
12.000000&13.000000&14.000000&15.000000

f2.txt
0.0000 1.0000 2.0000 3.0000
4.0000 5.0000 6.0000 7.0000
8.0000 9.0000 10.0000 11.0000
12.0000 13.0000 14.0000 15.0000

>>> genfromtxt('f1.txt',dtype='float')
array([[ 0., 1., 2., 3.℄,
[ 4., 5., 6., 7.℄,
[ 8., 9., 10., 11.℄,
[ 12., 13., 14., 15.℄℄)

>>> genfromtxt('f1.txt',skiprows=2)
array([[ 8., 9., 10., 11.℄,
[ 12., 13., 14., 15.℄℄)

If the arrays are too large, saving them in text format onsumes large volume of
memory. In that ase they an be saved in binary format.

>>> from numpy import *

>>> a=genfromtxt('f1.txt')
>>> save('f3.npy',a)
When the file f3.npy is opened in a word pro essor, the following output is o
ï¾ 12 NUMPY##F#{'des r': '<f8', 'fortran_order': False, 'shape': (4, 4), }
##############ï¾ 12 ?################################################ ###

>>> b=load('f3.npy')
>>> b
2.1. NUMPY 37

array([[ 0., 1., 2., 3.℄,

[ 4., 5., 6., 7.℄,
[ 8., 9., 10., 11.℄,
[ 12., 13., 14., 15.℄℄)

Basi Arithmeti Operations on arrays

Arithmeti operators apply elementwise on arrays. A new array is reated and lled
with the result.

>>> a = array( [20,30,40,50℄ )

>>> b = arange( 4 )
>>>b
array([0,1,2,3℄)
>>> = a-b
>>>
array([20, 29, 38, 47℄)
>>> b**2
array([0, 1, 4, 9℄)
>>> 10*sin(a)
array([ 9.12945251, -9.88031624, 7.4511316 , -2.62374854℄)
>>> a<35
array([True, True, False, False℄, dtype=bool)
>>> i = identity( 3 )
>>> i
array([[1, 0, 0℄,
[0, 1, 0℄,
[0, 0, 1℄℄)
>>> i + i # add element to element
array([[2, 0, 0℄,
[0, 2, 0℄,
[0, 0, 2℄℄)
>>> i + 4 # add a s alar to every entry
array([[5, 4, 4℄,
[4, 5, 4℄,
[4, 4, 5℄℄)
>>> a = array( range(1,10)).reshape(3,3)
>>> a
array([[1, 2, 3℄,
[4, 5, 6℄,
[7, 8, 9℄℄)
38 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

>>> i * a # element to element

array([[1, 0, 0℄,
[0, 5, 0℄,
[0, 0, 9℄℄)
>>> x = array( [1,2,3℄)
>>> x
array([1, 2, 3℄)
>>> y = array( [ [4℄, [5℄, [6℄ ℄)
>>> y
array([[4℄,
[5℄,
[6℄℄)
>>> x + y
array([[5, 6, 7℄,
[6, 7, 8℄,
[7, 8, 9℄℄)
#This is equivalent to ([[1,2,3℄,[1,2,3℄,[1,2,3℄℄+
#([[4,4,4℄,[5,5,5℄,[6,6,6℄℄)
>>> x*y
array([[ 4, 8, 12℄,
[ 5, 10, 15℄,
[ 6, 12, 18℄℄)
>>> x/y
array([[0, 0, 0℄,
[0, 0, 0℄,
[0, 0, 0℄℄)
>>> y/x
array([[4, 2, 1℄,
[5, 2, 1℄,
[6, 3, 2℄℄)
>>> x%y
array([[1, 2, 3℄,
[1, 2, 3℄,
[1, 2, 3℄℄)
>>> y%x
array([[0, 0, 1℄,
[0, 1, 2℄,
[0, 0, 0℄℄)
>>> x**y
array([[ 1, 16, 81℄,
[ 1, 32, 243℄,
[ 1, 64, 729℄℄)
>>> s=arange(1,6,1)
2.1. NUMPY 39

>>> s
array([1, 2, 3, 4, 5℄)
>>> s.sum() #sum of all elements
15
>>> s.prod() # produ t of all elements
120
>>> s.mean() # Mean of all elements
3.0
>>> s.var() #Varian e
2.0
>>> s.std() #Standard deviation
1.4142135623730951

2.1.6 Indexing, Sli ing and Iterating

One dimensional arrays an be indexed, sli ed and iterated over like lists and other
Python sequen es.

>>> a = arange(10)**3
>>> a
array([ 0, 1, 8, 27, 64, 125, 216, 343, 512, 729℄)
>>> a[2℄
8
>>> a[2:5℄
array([ 8, 27, 64℄)
>>> a[:6:2℄ = -1000 \# modify elements in a
>>> a[::-1℄ \# reversed a
array([ 729, 512, 343, 216, 125, -1000, 27, -1000, 1, -1000℄)
>>> for i in a: print i**(1/3.),
nan 1.0 nan 3.0 nan 5.0 6.0 7.0 8.0 9.0

2.1.7 Arrays as matri es

As the produ t operator '*' operates elementwiseP (produ t of orresponding elements)

in NumPy arrays, the matrix produ t (cij = k aik bkj ) an be found using the dot
fun tion. It also gives theP
dot produ t of two ve tors.The fun tion inner (x,y) omputes
the inner produ t (zij = k xik .yjk ) between two arrays. For 1-D arrays dot and inner
fun tions give the same result. Similarly a ross fun tion is dened whi h returns the
40 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

ross produ t of two ve tors. outer (x, y) omputes an outer produ t of two ve tors
(zij = xi .yj ). In matrix lass to reate matri es mat method and matrix method
are dened.mat(data, dtype=),matrix(data, dtype=). This data an be any list, tuple,
string or array . This fun tion interprets the input as a matrix.

>>> from numpy import*

>>> mat(range(2,7))
matrix([[2, 3, 4, 5, 6℄℄)
>>> a,b=mat([[1,2℄,[3,4℄℄),mat('1,2;3,4')
>>> a
matrix([[1, 2℄,
[3, 4℄℄),
>>> b
matrix([[1, 2℄,
[3, 4℄℄)
>>> r=mat('1,2,3,4')
>>> r
matrix([[1, 2, 3, 4℄℄) # Row matrix
>>> =mat('1;2;3;4')
>>>
matrix([[1℄,
[2℄,
[3℄,
[4℄℄) # olumn matrix

>>> k=arange(15).reshape(3,5)
>>> k
array([[ 0, 1, 2, 3, 4℄,
[ 5, 6, 7, 8, 9℄,
[10, 11, 12, 13, 14℄℄)

>>> k=mat(k) #array k be omes 3X5 matrix k

>>> t=transpose(k) # transposed 5X3 matrix
>>>l
matrix([[ 0, 5, 10℄,
[ 1, 6, 11℄,
[ 2, 7, 12℄,
[ 3, 8, 13℄,
[ 4, 9, 14℄℄))
>>> k*t #It must be a 3X3 matrix
matrix([[ 30, 80, 130℄,
[ 80, 255, 430℄,
[130, 430, 730℄℄)
2.1. NUMPY 41

>>> dot(k,t)
matrix([[ 30, 80, 130℄,
[ 80, 255, 430℄,
[130, 430, 730℄℄)
>>> t.fill(3)
>>> t
array([[3, 3, 3℄,
[3, 3, 3℄,
[3, 3, 3℄,
[3, 3, 3℄,
[3, 3, 3℄℄)

>>> s=range(1,6)
>>> m= mat(s)
>>> m
matrix([[1, 2, 3, 4, 5℄℄) #The two square bra kets are there as most of the matri es
>>> n=matrix(range(1,6))
>>> n
matrix([[1, 2, 3, 4, 5℄℄)
>>> x=[1,2,3℄
>>> y=[3,2,1℄
>>> dot(x,y) # like dot produ t of ve tors
10
>>> ross(x,y) # like ross produ t of ve tors
array([-4, 8, -4℄)
>>> inner([1,2,3℄,[10,100,1000℄)
3210 # 1.10+2.100+3.1000
>>> a=arange(9).reshape(3,3)
>>> b=a.T # another method to find transpose.
>>> a
array([[0, 1, 2℄,
[3, 4, 5℄,
[6, 7, 8℄℄)
>>> b
array([[0, 3, 6℄,
[1, 4, 7℄,
[2, 5, 8℄℄)
>>> inner(a,b) #Ordinary inner produ t of ve tors for 1-D arrays (without omplex o
produ t over the last axes.

array([[ 15, 18, 21℄,

[ 42, 54, 66℄,
[ 69, 90, 111℄℄)
42 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

It is possible to perform in rement and de rement operations without reating

>>> a = ones((2,3), dtype=int) #integer array
>>> b = random.random((2,3)) #float array
>>> a *= 3
>>> a
array([[3, 3, 3℄,
[3, 3, 3℄℄)
>>> b += a
>>> b
array([[ 3.69092703, 3.8324276 , 3.0114541 ℄,
[ 3.18679111, 3.3039349 , 3.37600289℄℄)
>>> a += b \# b is onverted to integer type
>>> a
array([[6, 6, 6℄,
[6, 6, 6℄℄)

When operating with arrays of dierent numeri data types, the type of the resulting
array orresponds to the more general or pre ise one.

>>> a = ones(3, dtype=int32)

>>> b = linspa e(0,pi,3)
>>> b.dtype.name
float64
>>> = a+b
>>>
array([ 1. , 2.57079633, 4.14159265℄)
>>> .dtype.name
'float64'
>>> d = exp( *1j)
>>> d
array([ 0.54030231+0.84147098j, -0.84147098+0.54030231j,
-0.54030231-0.84147098j℄)
>>> d.dtype.name
' omplex128'

Many unary operations, like omputing the sum of all the elements in the array, are
implemented as methods of the ndarray lass.

>>> a = random.random((2,3))
>>> a
array([[ 0.6903007 , 0.39168346, 0.16524769℄,
2.1. NUMPY 43

[ 0.48819875, 0.77188505, 0.94792155℄℄)

>>> a.sum()
3.4552372100521485 #sum of all elements
>>> a.min()
0.16524768654743593
>>> a.max()
0.9479215542670073

By default, these operations apply to the array as if it were a list of numbers, regardless
of its shape. However, by spe ifying the axis parameter you an apply an operation
along the spe ied axis(dimension) of an array:

>>> b = arange(12).reshape(3,4)
>>> b
array([[ 0, 1, 2, 3℄,
[ 4, 5, 6, 7℄,
[ 8, 9, 10, 11℄℄)
>>>
>>> b.sum(axis=0) # Give sum as a 1-D array(sum of ea h olu
array([12, 15, 18, 21℄)
>>> b.sum(axis=1) # Give sum as a 1-D array(sum of ea h row)
array([6, 22, 38℄)
>>> b.min(axis=1) # minimum of ea h row
array([0, 4, 8℄)
>>> b.min(axis=0) # minimum of ea h olumn
array([0,1,2,3℄)
>>> p.max(axis=0)
matrix([[12, 13, 14, 15℄℄)

>>> b. umsum(axis=1) # umulative sum along the rows

array([[ 0, 1, 3, 6℄,
[ 4, 9, 15, 22℄,
[ 8, 17, 27, 38℄℄)
>>> p=arange(16).reshape(4,4)
>>> p
array([[ 0, 1, 2, 3℄,
[ 4, 5, 6, 7℄,
[ 8, 9, 10, 11℄,
[12, 13, 14, 15℄℄)
>>> tra e(p)
30
>>> p=mat(p)
44 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

>>> p
matrix([[ 0, 1, 2, 3℄,
[ 4, 5, 6, 7℄,
[ 8, 9, 10, 11℄,
[12, 13, 14, 15℄℄)
>>> tra e(p)
30

2.1.8 Arrays as polynomial oe ients

There are four methods dened in polynomial lass to reate and manipulate polyno-
mials.

1. poly1d( , r, v) : reates a one-dimensional polynomial. Here c represents an array

or list. If r is True, c represents roots of the polynomial. If r is False (whi h is
the default), polynomial oe ients zeroth element orresponding to the highest
power of variable. v is the hara ter to be used as polynomial variable.

>>> =[3,1,-1,-3℄
>>> p=poly1d( )
>>> print p
3 2
3 x + 1 x - 1 x - 3

>>> p1=poly1d( ,True)

>>> print p1
4 2
1 x - 10 x + 9
>>> p2=poly1d( ,False,'y')
>>> print p2
3 2
3 y + 1 y - 1 y - 3

2. Here p is the polynomial and x is the value at whi h p is to be

polyval(p, x) :
evaluated. The polynomial value at x is returned. For polynomials dened above

>>> polyval(p2,2)
23
>>> polyval(p1,2)
-15
2.1. NUMPY 45

>>> polyval(p,2)
23

3. poly(s) : This fun tion returns the oe ients of the polynomial with the s as the
set of roots in the form of an array.

>>> d=[0,0℄
>>> poly(d)
array([1, 0, 0℄) #x*x=0
>>> d=[1,-1,2℄
>>> f=poly(d) #(x-1)(x+1)(x-2)
>>> f
array([ 1, -2, -1, 2℄)
>>> print poly1d(f)
3 2
1 x - 2 x - 1 x + 2

4. roots(p): Returns the roots of polynomial p

>>> g=[1,-2,-1,2℄
>>> p=poly1d(g)
>>> roots(p)
array([-1., 2., 1.℄)
Roots an also be found using the following ommand.
>>> p.r
array([-1., 2., 1.℄)

2.1.9 Linear Algebra

The linear algebra module is a sub lass of numpy. It is alled linalg. A few fun tions
are dened in the NumPy.linalg sub-pa kage. The important fun tions are

1. norm(x) : Returns norm of a ve tor x, norm =

pP
i x2i
2. det(a) : Returns determinant of a square matrix
3. inv(a) : Returns inverse of a non-singular square matrix
4. pinv(a) : Returns pseudoinverse of a singular square matrix. For invertible ma-
tri es, this is the same as the inverse.
5. solve(a,y) : Returns the solution ve tor x to the linear equation ax = y
46 CHAPTER 2. ADVANCED PYTHON PROGRAMMING

6. eig(a) :Returnall solutions (λ, x) to the equation ax = λx. The rst element of
the return tuple ontains all the eigenvalues. The se ond element of the return
tuple ontains the eigenve tors (ith eigenve tor as ith olumn).

7. eigvals(a) : Returns all eigenvalues of square matrix a as an array

8. eigh(h) : Return all solutions (λ, x) to the equation hx = λx where h is a hermitian

matrix.

9. eigvalsh(h) :Returns all eigenvalues of hermitian matrix h as an array

These are also in luded in the sub-pa kage numpy.dual.

 
1 1 2
Let a be a square matrix −1 0 1. It is reated as
2 3 0

>>> from numpy import*

>>> from numpy.linalg import*
>>> a=array([[1,0,1℄,
[2,1,0℄,
[0,2,4℄℄)
>>> det(a)
8.0
>>> inv(a)
array([[ 0.5 , 0.25 , -0.125℄,
[-1. , 0.5 , 0.25 ℄,
[ 0.5 , -0.25 , 0.125℄℄)
>>> eigvals(a)
array([ 0.8223493+1.07730381j, 0.8223493-1.07730381j, 4.3553014+0.j℄)
>>> eig(a)
(array([ 0.8223493+1.07730381j, 0.8223493-1.07730381j, 4.3553014+0.j ℄
#eigenvalues

array([[-0.06908062+0.41891651j, -0.06908062-0.41891651j, 0.28156897+0.j ℄,

[ 0.77771286+0.j , 0.77771286+0.j , 0.16783528+0.j ℄,
[-0.43902814-0.14884162j, -0.43902814+0.14884162j, 0.94474877+0.j ℄℄)
# Three eigenve tors

To solve the simultaneous equations

2x + 3y + 4z = 8, 3x + 4y + 5z = 10, 4x − 5y + 6z = 32
2.1. NUMPY 47

>>> a=mat('2,3,4;3,4,5;4,-5,6') # The oeffi ient matrix

>>> a
matrix([[ 2, 3, 4℄,
[ 3, 4, 5℄,
[ 4, -5, 6℄℄)
>>> s=mat('8;10;32') # Constant ve tor S in matrix equation AX=S
>>> s
matrix([[ 8℄,
[10℄,
[32℄℄)
>>> b=solve(a,s)
>>> b
matrix([[ 1.℄,
[-2.℄,
[ 3.℄℄)
48 CHAPTER 2. ADVANCED PYTHON PROGRAMMING
Chapter 3

Plotting and visualization

3.1 Matplotlib

Graphs, harts, surfa e plots et are visual presentations of numeri al data. It is useful
to ompare, ontrast, dete t trends, predi t and omprehend huge amounts of data.
Dierent Python modules are used for generating two and three dimensional graphs.

3.1.1 The Matplotlib Module

The python pa kage Matplotlib produ es graphs and gures in dierent hard opy for-
mats like jpeg, bmp, eps, png et . Most of the fun tions of NumPy and matplotlib.pyplot
are dened in the module pylab also. It also provides many fun tions for matrix ma-
nipulation. The data for plotting are supplied as Python lists or Numpy arrays. This
module ontains a lot of methods for plotting and annotating graphs. Some of these
methods are used frequently in s ienti omputing. They are listed below with exam-
ples.

1. plot()Fun tion : The general format is

plot(x, y, color =, linestyle =, marker =, markerf acecolor =, markersize =)

x,y are lists or arrays, olor olor of graph, linestyle spe ies dashedline(''),solid
line('-'), dotted line (':') et ., marker marking data points on the graph ('.','*'),
markerfa e olor the olor of marker and markersize is the size of marker.

2. show() : Sends graphi output to the s reen.

49
50 CHAPTER 3. PLOTTING AND VISUALIZATION

3. xlabel() :  is name of variable plotted along x-axis.

4. ylabel() :  is name of variable plotted along y-axis.

5. title() :  is name of Graph.

6. legend(names,lo =) : The names of dierent urves as a list or tuple of strings,

lo is the lo ation in the graph where the legend must appear. Upper right, lower
left et . lo =0 ts the legend at the most onvenient lo ation.

7. grid(True) : Shows grid lines if True

8. axis(z) : Used to set or get the axis properties.

axis() Returns the urrent axes limits [xmin, xmax, ymin, ymax℄.
axis(z) :Sets the min and max of the x and y axes, with z = [xmin, xmax, ymin,
ymax℄.
axis('o ' ): Removes the axis lines and labels.
axis('equal') : Changes limits of x or y axis so that equal in rements of x and y
have the same length so that a ir le will appear ir ular.

9. gure()

f igure(num, f igsize = (w, h), dpi = N, f acecolor =′ b′ , edgecolor =′ k ′ )

num is an integer variable. This fun tion reates a new gure if gure(num)
does not exist. If it already exists, it be omes a tive. gsize=(w,h) is a tuple
of width and height in in hes, dpi=N reates gure with resolution N dots per
square in h, fa e olor sets the ba kground olor edge olor sets the border olor.
All arguments ex ept the rst are optional.

10. subplot(m, n, N) : Where m is the number of rows, n number of olumns and

N = 1 is the rst plot number and in reasing N ll rows rst. Nmax = m × n

11. Writes 'string' at (x, y) with respe t to bottom-left orner as

text(x,y,'string')
origin and s aled as in the gure.

12. bar(left, height, width=0.8, bottom=0, olor=None, edge olor=None, linewidth=None,ye

xerr=None, e olor=None, apsize=3,align='edge', orientation='verti al', log=False)
Make a bar plot with re tangles representing the two arrays left, height bounded
by left, left + width, bottom, bottom + height with optional arguments whose
default values are given.

13. barh() Like bar() ex ept that the bars are horizontal.

14. ontour() For plotting impli it fun tions.

15. ontourf() For olor-lled plotting of impli it fun tions.

3.1. MATPLOTLIB 51

16. loglog() Make

a plot with log s aling on the x and y axis. It supports all the
keyword arguments of plot() fun tion.

17. semilogx()Graph with x-axis plotted in log s ale. It supports all the keyword
arguments of plot() fun tion.

18. semilogy() Graph with y-axis plotted in log s ale.

19. ogrid[minx:maxx:nxj,miny:maxy:nyj℄ Creates a grid of nx x-vlues in the range

(xmin,xmax) and ny y-values in the range (ymin,ymax).eg. x,y=ogrid[1:10:100j,5:15:150j℄

20. pie()pie(x, explode=None, labels=None, olors=None, autop t=None, p tdistan e=0.6,shadow=

labeldistan e=1.1, hold=None) Makes a pie hart of array x. The fra tional area
of ea h wedge is given by x/sum(x). If sum(x) <= 1, then the values of x give
the fra tional area dire tly and the array will not be normalized.

21. Make a polar plot. theta and r are lists or arrays. Multiple
polar(theta, r, args)
theta, r arguments are supported, with format strings, as in plot().

22. s atter(x, y, s=20, ='b', marker='o', map=None, norm=None,vmin=None,

vmax=None, alpha=1.0, linewidths=None,verts=None, **kwargs) Make a s atter
plot of x versus y , where x, y are onverted to 1-D sequen es whi h must be of
the same length,N .

Examples

from matplotlib.pyplot import*

from numpy import*
subplot(2,2,1)
plot([1,2,3,4℄,'*-')
subplot(2,2,2)
plot([4,2,3,1℄,'^-')
subplot(2,2,3)
plot([4,3,2,1℄,'^-')
subplot(2,2,4)
plot([2,4,3,1℄,'^-')
show()
52 CHAPTER 3. PLOTTING AND VISUALIZATION

4.0 4.0

3.5 3.5

3.0 3.0

2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

4.0 4.0

3.5 3.5

3.0 3.0

2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

Polar plots: Polar oordinates lo ate a point on a plane with one distan e and
one angle. The distan e `r' is measured from the origin. The angle θ is measured
from positive dire tion of x-axis in the anti- lo kwise sense. Plotting is done using
polar(theta, radius, f ormat string) fun tion. An example is given below.

Polar rhodonea A or rose urve is a sinusoid r = sin (nθ) where n is a

rhodonea
onstant. If n is an integer the urve will have 2n 'petals' and n 'petals' if n is odd.
If n is a rational number(=p/q, p,q integers), then the urve is losed and has nite
length. If n is an irrational number, then it is losed and has innite length.

from matplotlib.pyplot import*

from numpy import*
n=2
th = linspa e(0, 10*pi,1000)
r = sin(n*th)
polar(th,r)
show()
3.1. MATPLOTLIB 53

90

135 45

1.0
0.8
0.6
0.4
0.2

180 0

225 315

270

Pie Charts: A pie hart is a ir ular hart in whi h a ir le is divided into se tors.
Ea h se tor visually represents an item in a data set to mat h the per entage or fra tion
of the item in the total data set. Pie harts are useful to ompare dierent parts of
a whole amount. They are often used to present nan ial information. The fun tion
pie(list of per entages or fra tions , labels=list of labels) produ es a pie hart. Both
the lists must have the same length.

from matplotlib.pyplot import*

from numpy import*
labs = ['A+', 'A', 'B+', 'B', 'C+', 'C','D'℄
fra s = [5,8,18, 19, 20,17,14℄
pie(fra s, labels=labs)
show()

B+

B A

A+

D
C+

Parametri plots: A parametri plot is a visual des ription of a set of parametri

equations. If x and y are both fun tions of a variable t, then they reate a set of
parametri equations. For example, the two equations y = t sin t2 and x = t cos t2 form
a set of parametri equations in whi h y and x are fun tions of t, the graph of whi h
will be in this form.

from matplotlib.pyplot import*

from numpy import*
54 CHAPTER 3. PLOTTING AND VISUALIZATION

t=arange(0,6.3,0.001)
x=t* os(t*t)
y=t*sin(t*t)
plot(x,y)
show()

2

4

6

8

6
4
2 0 2 4 6 8

2-D plots in olours: Two dimensional matrix an be represented graphi ally by

assigning a olor to ea h point proportional to the value of that element. The fun tion
imshow(matrix) is employed to reate su h plots.

from matplotlib.pyplot import*

from numpy import*
m=linspa e(0,1,900).reshape(30,30)
imshow(m)
show()

10

15

20

25

0 5 10 15 20 25
3.1. MATPLOTLIB 55

3.1.2 Plotting mathemati al fun tions

sine fun tion:

from pylab import*

x=linspa e(0,2*pi,200)
y=sin(x)
plot(x,y)
xlabel('x')
ylabel('sin(x)')
title('Plot of sine fun tion')
grid(True)
show()

Logarithm fun tion: Logarithm fun tion log(x) gives logarithm of a variable to the
base exponential e.

from pylab import*

x=linspa e(0,200,200)
y=log(x)
plot(x,y)
xlabel('x')
ylabel('log(x)')
title('Plot of log fun tion')
grid(True)
show()

Exponential fun tion: Exponential fun tion exp(x) gives ex of a variable x.

from pylab import*

x=linspa e(0,5,200)
y=exp(x)
plot(x,y)
xlabel('x')
ylabel('exp(x)')
title('Plot of exponential fun tion')
grid(True)
show()
56 CHAPTER 3. PLOTTING AND VISUALIZATION

Gaussian fun tion: Gaussian fun tion is given by y = exp(−x2 ) gives ex of a vari-
able x.

from pylab import*

x=linspa e(-5,5,200)
y=exp(-x**2)
plot(x,y)
xlabel('x')
ylabel('gaussian(x)')
title('Plot of Gaussian fun tion')
grid(True)
show()

Gamma fun tion: The gamma fun tion is dened by the integral
Z ∞
Γ(x) = tx−1 e−t dt
0

and satises the re urren e relation

Γ(x + 1) = xΓ(x)
and ree tion formula
π
Γ(x)Γ(1 − x) =
sin πx
πx
Γ(1 + x)Γ(1 − x) =
sin (πx)
πx
Γ(1 − x) =
Γ(1 + x) sin (πx)
It an be used to al ulate Γ-fun tions less than 1. Sin e sin (πx) is zero for integer
x, Γ-fun tion is unbounded for negative integers but not for negative fra tions. The
following approximation method,derived by Lan zos, is employed for al ulating the
Γ-fun tion numeri ally. For x > 0,
√
 
k−5 −k a1 a2 an
Γ(1 + x) = k e 2π a0 + + + .... + +ǫ
x+1 x+2 x+n
where k = x + 5.5, ǫ the error term and ai expansion oe ients. For |ǫ| < 2 × 10−10 ,
n = 6 is su ient. The oe ients are given by
a0 = 1.00002746310005, a1 = 76.18009172947146
a2 = −86.50532032941677, a3 = 24.01409824083091
a4 = −1.231739572450155, a5 = 1.208650973866179 × 10−3 ,
a6 = −5.395239384953 × 10−6
√
 
k−5 a1 a2 an
ln Γ(1 + x) = ln(k ) − k + ln( 2π) + ln a0 + + + .... +
x+1 x+2 x+n
3.1. MATPLOTLIB 57
"√  #
2π a1 a2 an
ln Γ(x) = ln (k k−5 ) − k + ln a0 + + + .... +
x x+1 x+2 x+n

But reasonable a ura y an be a hieved by just taking the rst ve terms and ap-
proximating elements of a to 2 de imal pla es.

from pylab import*

def gamma(x0):
a=[1,76.18,-86.505,24.014,-1.232℄
k=x0+5.5
s=(k-5)*log(k)-k
s1=a[0℄
for i in range(1,5):
s1+=a[i℄/(x0+i)
return exp(s+log(sqrt(2*pi)*s1/x0))
p=linspa e(-5,6,1000)
z=gamma(p)
plot(p,z)
axis([-5,7,-1,150℄)
xlabel('x')
ylabel('Gamma fun tion')
title('Plot of Gamma fun tion')
grid(True)
show()

Plot of Gamma function

140

120

100
Gamma function

80

60

40

20

0
−4 −2 0 2 4 6
x
58 CHAPTER 3. PLOTTING AND VISUALIZATION

Polynomial Evaluation: Legendre Fun tion Legendre fun tions Pn (x) are de-
ned through a generating fun tion as

∞
1 X
√ = Pn (x)tn
1 − 2xt + t2 n=0

where |t| ≤ 1 The zeroth term (n = 0) of the expansion is 1 = P0 (x)t0 Hen e P0 (x) = 1
for all x. Expanding left side as a binomial series and equating oe ient of t on both
sides, xt = P1 (x)t or P1 (x) = x for all x. Dierentiating and equating oe ients of tn
on both sides one gets

(2n + 1)xPn (x) − nPn−1 (x)

Pn+1 (x) =
n+1

This is a re urren e relation whi h may be used for al ulating Legendre polynomials
of any order.

from pylab import*

def legendre(m,z):
p0,p1=1,z
if m==0:p=p0
if m==1:p=p1
if m>1:
for i in range(1,m):
p=((2*i+1)*z*p1-i*p0)/(i+1)
p0,p1=p1,p
return(p)

n=input('Give Order n of Legendre fun tion')

x=linspa e(-1,1,200)
y=legendre(n,x)
plot(x,y)
xlabel('x')
ylabel('Legendre fun tion')
title('Plot of Legendre Polynomial for n=5')
grid(True)
show()
3.1. MATPLOTLIB 59

1.0
Legendre Polynomial for n=5 0.5 Bessel function for n=2

0.4
0.5
Legendre function

0.3

Bessel function
0.0

0.2

−0.5
0.1

−1.0 0.0
−1.0 −0.5 0.0 0.5 1.0 −6 −4 −2 0 2 4 6
x x

Polynomial Evaluation: Bessel's Fun tion The Bessel fun tion of order n for a
variable x is given by the series
∞
X (−1)s  x 2s+n
Jn (x) =
s=0
s!(n + s)! 2

1  x n
The zeroth term (s = 0) of the expansion is . The ratio of the pth and
n! 2
−1  x 2
(p − 1)th term of the expansion is . Using these results Bessel fun tion
p(n + p) 2
of any order an be al ulated for any desired a ura y.

from pylab import*

def bessel(m,z):
z/=2.0
60 CHAPTER 3. PLOTTING AND VISUALIZATION

fa torial=1
for i in range(1,n+1):fa torial*=i
term=[z**n/fa torial℄
for i in range(1,10):
term.append(-term[i-1℄*z*z/(i*(i+n)))
return sum(term)

n=input('Give Order n of Bessel fun tion : ')

x=linspa e(-5.0,5.,1000)
y=[bessel(n,i) for i in x℄
plot(x,y)
xlabel('x')
ylabel('Bessel fun tion')
title('Plot of Bessel fun tion for n=2')
grid(True)
show()
Chapter 4

Numeri al Analysis

4.1 Numeri al methods

4.1.1 Inverse of a fun tion

Inverse fun tions are often used in physi s. For example, onsider the length L of
the mer ury olumn in a apillary tube as fun tion of temperature T . L = f (T ).
When it is used as a thermometer, the length of the mer ury pellet is measured and
the temperature is inferred from it using the formula T = f −1 (L). Similarly, in a
piezoele tri rystal, the voltage V developed is a fun tion of stress S applied. V =
f (S). In a strain gauge, this voltage is measured to estimate the load W pla ed on
it. W = Af −1 (V ) where A is the surfa e area of the rystal on whi h the load applies
stress. If a fun tion represents a prin iple, in general, its appli ation employs the
inverse fun tion.

Denition 1 (Fun tion) For every x in a set X, if an obje t f maps exa tly one
element y in set Y then f is alled a fun tion with domain X, and o-domain or range
Y.

It is represented as
∀x ∈ X, y ∈ Y, f : x → y

y is alled the image and x, the preimage. If there is more than one preimage for a
given image, then f is alled an onto mapping. For example, f : x → y has the expli it
form y = x2 , then image y is the same for all ±x in the domain X . If for every image,
there is a unique preimage, f is a one-to-one mapping. For example, f : x → y has

61
62 CHAPTER 4. NUMERICAL ANALYSIS

the expli it form y = x3 , then image y is the unique for every x in the domain X . In
terms of domain and o-domain sets f : X → Y .

Denition 2 (Inverse Fun tion) If f is a fun tion whose domain is the set X, and
range is the set Y and there exists a fun tion g with domain Y and range X, with the
property

∀x ∈ X, f (x) = y ∈ Y if and only if ∀y ∈ Y, g(y) = x

then g is alled inverse of f.

The fun tion f sends all elements of the the domain X to the range Y . If f is invertible,
the fun tion g is unique. There is exa tly one fun tion g satises this property. Fun tion
g is alled the inverse of f , denoted formally as f −1 . (6= 1/f ). Sin e f implies unique
y for ea h x and g implies unique x for ea h y , the two mappings f and g must be
one-to-one. This is a ne essary ondition for f to have an inverse.

Test to he k whether a fun tion is one-to-one: It is alled the horizontal line

test.If no line parallel to x-axis interse ts the graph of the fun tion y = f (x) at more
than one point, that fun tion is one-to-one.

100

80

60 onto function
f(x)

40

20
one-to-one function

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
x

Method of nding inverse

There are dierent methods of nding inverse fun tion.

4.1. NUMERICAL METHODS 63

Algebrai method: The algorithm for nding an inverse fun tion g for f (x) alge-
brai ally involves the following steps.

1. Che k whether f is a one-to-one mapping.

2. Put f (x) = y

3. Swap the x and y variables

4. Solve for y. It gives f −1 (x)

5. Verify that f −1 (f (x)) = x or f (f −1(x)) = x

Example 1 Find inverse of f (x) = 3x + 4

1. f is a one-to-one fun tion.

2. Put 3x + 4 = y

3. Swap variables: 3y + 4 = x

4. solve for y : y = (x − 4)/3 = f −1 (x)

5. f (f −1 (x)) = 3 ∗ [(x − 4)/3] + 4 = x

Hen e inverse fun tion is f −1 (x) = (x − 4)/3

Problem 1 Find inverse of f (x) = (x + 1)/x

Answer:f −1 (x) = 1/(x − 1)

Problem 2 Find inverse of f (x) = log x

Answer:f −1 (x) = ex

Problem 3 Find inverse of f (x) = (x + 1)/x

Answer:f −1 (x) = 1/(x − 1)

64 CHAPTER 4. NUMERICAL ANALYSIS

Swapping method: Often the experimental data related through some fun tion will
be in the form of ordered pairs formed from tables.
x x1 x2 x3 x4 x5 x6 x7 x8 x9
y = f (x) y1 y2 y3 y4 y5 y6 y7 y8 y9
Then f (x) may be expressed as ordered pairs f (x) : (xi , yi )), i = 1, 2, 3... Here

f (x) : (x1 , y1 ), (x2 , y2 ), (x3 , y3 ), (x4 , y4 ), (x5 , y5), (x6 , y6), (x7 , y7 ), (x8 , y8 )(x9 , y9 )

If ∀i, j, xi 6= xj , and yi 6= yj , then the fun tion f (x) is one-to-one. Its inverse will exist
and an be obtained by simply swapping x and y values.

f −1 (x) : (y1 , x1 ), (y2, x2 ), (y3 , x3 ), (y4 , x4 ), (y5 , x5 ), (y6 , x6 ), (y7, x7 ), (y8, x8 )(y9 , x9 )

In tabular form, it will appear as

x y1 y2 y3 y4 y5 y6 y7 y8 y9
−1
f (x) x1 x2 x3 x4 x5 x6 x7 x8 x9

Example 2 Find the inverse of the fun tion

x 1 -2 -1 0 2 3 4 -3
f (x) 2 0 3 -1 1 -2 5 1
Swapping x and y , we get the inverse fun tion as x and y never repeats among them-
selves.

f −1 (x) = (2, 1), (0, −2), (3, −1), (−1, 0), (1, 2), (−2, 3), (5, 4), (1, −3)

Graphi al method: The basi prin iple is that the graph of an inverse relation is
the ree tion of the graph of original relation on the identity line (slope=1),y = x.

Example 3 The fun tion f (x) = 2x + 1 has f −1 (x) = (x − 1)/2. Plots are given below
4.1. NUMERICAL METHODS 65

4 ( ) =2x +1
f x

2 f x( ) =x

1
y

0 ( ) = f−1 (x) =(x−1)/2

g x

−1

−2

−3
−3 −2 −1 0 1 2 3 4 5
x
Often it is ne essary to re-
stri t the domain on ertain fun tions to guarantee that the inverse relation is also a
fun tion.

Example 4 For example if y = ax2 is the fun tion, then it is one-to-one only for
x > 0.

1.0

0.8

0.6 ( ) = f−1 (x) =

g x x
y

0.4 ( ) =x
f x

0.2 f x ( ) = x2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
x
1

1 Note that all graphs will not produ e an inverse relation whi h is also a fun tion.
66 CHAPTER 4. NUMERICAL ANALYSIS

4.1.2 Interpolation with Cubi Spline

Real world numeri al data is usually di ult to analyse. Any fun tion whi h would
ee tively orrelate the data would be di ult to obtain and highly unwieldy. To this
end, the idea of the ubi spline was developed. Using this pro ess, a series of unique
ubi polynomials are tted between ea h of the data points, with the stipulation that
the urve obtained be ontinuous and appear smooth. These ubi splines an then be
used to determine rates of hange and umulative hange over an interval.

y0 yn
y1

x
x0 x1 x2 xn
Theory:
Let there be (n + 1) data points (xi , yi ), i = 0, 1, 2, ...n. The essential idea is to t a
pie ewise fun tion of the form


 p0 (x) x0 ≤ x < x1

p (x)
1 x1 ≤ x < x2
P(x) =


 ... ...
pn−1 (x) xn−1 ≤ x < xn


where ea h pi (x) is a ubi polynomial of the form

p(x) = a3 (x − xi )3 + a2 (x − xi )2 + a1 (x − xi ) + a0

. To make the interpolation ontinuous, smooth and well-behaved, we impose the

following onditions

1. The pie ewise fun tion P(x) will interpolate all data points. That is P(xi ) =
pi (xi ) = yi
4.1. NUMERICAL METHODS 67

2. P(x) will be ontinuous on the interval (x0 , xn ). That is pi (xi ) = pi−1 (xi )
3. The rst derivative P′ (x) will be ontinuous on the interval (x0 , xn ). that is,
p′i (xi ) = p′i−1 (xi )
4. The se ond derivativeP”(x) will be ontinuous on the interval (x0 , xn ). That is,
p′′i (xi ) = p′′i−1 (xi )

The se ond derivative of a ubi polynomial is of degree one ( a straight line). For the
rst pair of points (x0 , y0), (x1 , y1) the following form may be hosen.
x − x1 x − x0
p′′0 (x) = a0 + a1
x0 − x1 x1 − x0
a0 and a1 are then given by

p′′0 (x0 ) = a0 , p′′0 (x1 ) = a1

whi h are the values of the se ond derivative of p0 (x) at x = x0 and x = x1 . Integrating
p′′0 (x) with respe t to x
(x − x1 )2 (x − x0 )2
p′0 (x) = a0 + A + a1 +B
2(x0 − x1 ) 2(x1 − x0 )
where A.B are onstants of integration. The two onstants are dierent be ause the
rst term is integrated with respe t to (x − x1 ) while the se ond term is integrated
with respe t to (x − x0 ) Integrating again
(x − x1 )3 (x − x0 )3
p0 (x) = a0 + A(x − x1 ) + a1 + B(x − x0 )
6(x0 − x1 ) 6(x1 − x0 )
To nd A, B we use the two given points (x0 , y0 ) and (x1 , y1). Substituting x = x0
(x0 − x1 )3 (x0 − x0 )3
p0 (x0 ) = a0 + A(x0 − x1 ) + a1 + B(x0 − x0 )
6(x0 − x1 ) 6(x1 − x0 )
(x0 − x1 )3
y0 = a0 + A(x0 − x1 )
6(x0 − x1 )
y0 a0 (x0 − x1 )
A= −
x0 − x1 6
Similarly using p0 (x1 ) = y1
(x1 − x1 )3 (x1 − x0 )3
y1 = p0 (x1 ) = a0 + A(x1 − x1 ) + a1 + B(x1 − x0 )
6(x1 − x1 ) 6(x1 − x0 )
y1 a1 (x1 − x0 )
B= −
x1 − x0 6
68 CHAPTER 4. NUMERICAL ANALYSIS

The polynomial has the form

a0 (x − x1 )3
 
y0 a0
p0 (x) = + − (x0 − x1 ) (x − x1 )
6 (x0 − x1 ) x0 − x1 6
3
 
a1 (x − x0 ) y1 a1
+ + − (x1 − x0 ) (x − x0 )
6 (x1 − x0 ) x1 − x0 6
Similarly the polynomial between (x1 , y1 ) and (x2 , y2 ) will have the form

a1 (x − x2 )3
 
y1 a1
p1 (x) = + − (x1 − x2 ) (x − x2 )
6 (x1 − x2 ) x1 − x2 6
3
 
a2 (x − x1 ) y2 a2
+ + − (x2 − x1 ) (x − x1 )
6 (x2 − x1 ) x2 − x1 6
To evaluate the se ond derivatives ai = p′′i (xi ), we use the ondition that the rst
derivative be ontinuous at the knots. That is, p′0 (x1 ) = p′1 (x1 ).

3a0 (x − x1 )2 y0 a0
p′0 (x) = + − (x0 − x1 )
6 (x0 − x1 ) x0 − x1 6
2
3a1 (x − x0 ) y1 a1
+ + − (x1 − x0 )
6 (x1 − x0 ) x1 − x0 6
y0 a0 a1 y1 a1
p′0 (x1 ) = − (x0 − x1 ) + (x1 − x0 ) + − (x1 − x0 )
x0 − x1 6 2 x1 − x0 6
Rearranging and simplifying
y0 a0 a1 y1 a1
p′0 (x1 ) = − (x0 − x1 ) + (x1 − x0 ) + − (x1 − x0 )
x0 − x1 6 2 x1 − x0 6
y1 − y0 a0 2a1
= + (x1 − x0 ) + (x1 − x0 )
x1 − x0 6 6
Similarly
3a1 (x − x2 )2 y1 a1
p′1 (x) = + − (x1 − x2 )
6 (x1 − x2 ) x1 − x2 6
2
3a2 (x − x1 ) y2 a2
+ + − (x2 − x1 )
6 (x2 − x1 ) x2 − x1 6
At x = x1 ,
3a1 (x1 − x2 )2 y1 a1 y2 a2
p′1 (x1 ) = + − (x1 − x2 ) + − (x2 − x1 )
6 (x1 − x2 ) x1 − x2 6 x2 − x1 6
Whi h an be simplied as
y2 − y1 2a1 a2
p′1 (x1 ) = − (x2 − x1 ) − (x2 − x1 )
x2 − x1 6 6
4.1. NUMERICAL METHODS 69

Therefore ontinuity of rst derivative requires

y1 − y0 a0 2a1 y2 − y1 2a1 a2
+ (x1 − x0 ) + (x1 − x0 ) = − (x2 − x1 ) − (x2 − x1 )
x1 − x0 6 6 x2 − x1 6 6
a0 2a1 2a1 a2 y2 − y1 y1 − y0
(x1 − x0 ) + (x1 − x0 ) + (x2 − x1 ) + (x2 − x1 ) = −
6 6 6 6 x2 − x1 x1 − x0
The se ond and third terms on the left hand side may be ombined
a0 2a1 a2 y2 − y1 y1 − y0
(x1 − x0 ) + (x2 − x0 ) + (x2 − x1 ) = −
6 6 6 x2 − x1 x1 − x0
For the polynomials p1 (x) between (x1 , x2 ) and p2 (x) between (x2 , x3 ),
a1 2a2 a3 y3 − y2 y2 − y1
(x2 − x1 ) + (x3 − x1 ) + (x3 − x2 ) = −
6 6 6 x3 − x2 x2 − x1
There are four unknowns a0 , a1 , a2 , a3 and two equations. So the values of any two of
them must be assumed to get the other two. As two of the variables repeat in every
su essive equations (a1 , a2 here), if we ompute all the n-polynomials between the
(n + 1) data points, we will get (n − 1) equations and (n + 1) variables. Usually it is
assumed that a0 = an = 0 so that there are (n − 1) variables and (n − 1) equations
whi h an be exa tly solved. Cubi spline with a0 = an = 0 is alled natural splin4e

If the x-values are equispa ed, let xi+1 − xi = h, then the (n − 1) equations may be
written as
    
4 1 0 ... 0 0 0 a1 y0 − 2y1 + y2
1 4 1 ... 0 0 0  a2 
   y1 − 2y2 + y3 
  
1 4 1
 ... 0 0 0
 a3 
= 6  y2 − 2y3 + y4



.. .. .. ... .. .. ..  ..  h 
   2  .... 
 
0 0 0 ... 1 4 1 an−2  yn−3 − 2yn−2 + yn−1 
0 0 0 ... 0 1 4 an−1 yn−2 − 2yn−1 + yn
The general formula for the ubi polynomial between xi and xi+1 is given by
ai (x − xi+1 )3
 
yi ai
pi (x) = + − (xi − xi+1 ) (x − xi+1 )
6 (xi − xi+1 ) xi − xi+1 6
3
 
ai+1 (x − xi ) yi+1 ai+1
+ + − (xi+1 − xi ) (x − xi )
6 (xi+1 − xi ) xi+1 − xi 6
Problem 4 -Suppose
(
x3 + ax2 − 4x + c, 0≤x≤2
s(x) =
−x3 + 9x2 + bx + 34, 2≤x≤4

Find onstants a, b, c su h the s(x) is twi e ontinuously dierentiable on the interval

[0, 4].
70 CHAPTER 4. NUMERICAL ANALYSIS

Problem 5 Suppose
(
ax3 + x −2 ≤ x ≤ 0
s(x) =
x3 + bx, 0≤x≤2

Find onstants a, b su h the s(x) is twi e ontinuously dierentiable on the interval

[−2, 2].

Problem 6 Suppose
(
0 x≤2
s(x) =
(x − 2)3 2<x

Is s(x) a ubi spline? Justify your answer.

Problem 7 Using ubi spline interpolation te hnique, nd y(x = 0.6) from the fol-
x 0.1 0.2 0.4 0.7 1.1
lowing data.
y 0.5754 0.6796 0.8026 0.9179 1.0231

Ans: y(0.6) = 0.8846

4.1.3 Zeros of polynomials

Polynomials are used in physi s to des ribe the traje tory of proje tiles. Polynomial
integrals an be used to express energy, inertia and voltage dieren e. in quantum
me hani s, orthogonal polynomials appear as energy and momentum eigenfun tions.
The zero or root of these polynomials gives positions and instants of zero probability
for a physi al system. They are also used for interpolation of experimental data.

Denition 3
Pn
If p(x) = i=0 ai xi is a polynomial of degree n and p(b) = 0, then b is
alled a zero or root of the polynomial p(x).

There is an important theorem that relates the fa tors and zeros (roots) of a polynomial.
There are 5 theorems about roots of Polynomials. They are n-Zero theorem, Remainder
theorem, Fa tor theorem, Rational Root Theorem, Irrational Root Theorem, Complex
Root Theorem, Des artes Rule.

Theorem 1 (n-Zero's Theorem) If p(x) is of Degree n, then it has at most n zeros.

p(x)
Theorem 2 (The Remainder Theorem) If = q(x)and r, the remainder, then p(b)
x−b
r
4.1. NUMERICAL METHODS 71

Proof:
p(x) = q(x)(x − b) + r, p(b) = r
To verify remainder theorem: If p(x) = x3 − x2 − 17x − 16. Let us divide p(x) by
(x − 5). The quotient is q(x) = x2 + 4x + 3 and the remainder is r = −1. Now
p(5) = 53 − 52 − 17 × 5 − 16 = −1 = r . Hen e veried.

If remainder r = 0 wehave the fa tor theorem.

Theorem 3 (Fa tor Theorem) p(x) = ni=0 ai xi is a polynomial

P
If of degree n and
(x − b) is a fa tor of the polynomial p(x), then b is a zero of p(x).

Proof:
p(x) = q(x)(x − b), p(b) = 0

To verify fa tor theorem: If p(x) = x3 + (a − 1)x2 − (a + 6)x − 6a. Fa toring the

polynomial we get p(x) = (x + a)(x + 2)(x − 3). Hen e x = (−a, −2, +3) are the set of
roots. p(−a) = a3 + (a − 1)a2 − (a + 6)a − 6a = 0. Similarly p(−2) = p(3) = 0

Theorem 4 (Bolzano's Theorem) If p(x) is a ontinuous fun tion in the interval

x ∈ (a, b) and p(a)p(b) < 0, then there exists at least one x = c ∈ (a, b) su h that
p(c) = 0.

This is a spe ial ase of Intermediate Value Theorem. To verify Bolzano's theorem:
If p(x) = 6x3 − 20x2 − 14x + 60. (x − 2) is a zero of this fun tion.

p(1.5) = +14.25, p(2.5) = −6.25, p(1.5) × p(2.5) = −89.0625 < 0

The root 2 lies between 1.5 and 2.5

Theorem 5 (Rational Roots Theorem)

Pn
Rational Roots of p(x) = i=0 ai xi will
be of the form of (fa tors of a0 divided by fa tors of an .

A rational root is one whi h is expressible as a quotient of two integers. In the poly-
nomial p(x) = 6x3 − 13x2 + x + 2, a0 = 2, an = a3 = 6. Hen e the possible rational
roots are f actors of a0 /f actors of a3 = f actors of 2/f actors of 6. That is

±1, ±2
= ±1, ±1/2, ±1/3, ±1/6, ±2, ±2/3
±1, ±2, ±3, ±6

The theorem states that if there is a rational root , it must be one of these. In fa t,
the roots are (2, 1/2, −1/3) whi h belong to this set.
72 CHAPTER 4. NUMERICAL ANALYSIS

Theorem 6 (Irrational
√ Root Theorem) If p(x)√
is a polynomial with rational oef-
 ients and a+
√ b where a and b are rational and b is irrational is a root, then the
onjugate a − b is also a root.

√
For example, if p(x) = 2x3 − x2 − 9x − 4, its roots are x0 = 1/2, x± = (1 ± 17)/2.
Irrational roots o ur in pairs.

Theorem 7 (Complex Root Theorem) If p(x) is a polynomial fun tion with real
oe ients and a + ib is a root, then a − ib must also be a root of p(x).

This theorem states that if the Coe ients are real, all Complex Roots o ur in
√ Conju-
gate Pairs. For example, the roots of p(x) = 3x +7x +11x+3 are x± = −1±i 2, x0 =
3 2

1/3
An obvious orollary is for an odd-degree polynomial there exist at least one real root.

A method of determining the maximum number of positive and negative real roots
of a polynomial is given by Des artes Rule.
For positive roots, start with the sign of the oe ient of the lowest (or highest) power
of x. Count the number of sign hanges n as you pro eed from the lowest to the
highest power (ignoring powers whi h do not appear). Then n is the maximum number
of positive roots. Consider p(x) = x7 + x6 − x4 − x3 − x2 + x − 1. Sin e there
are three sign hanges, there are a maximum of three possible real positive roots.
A tually, there is only one real positive root x = 1.1147 and three pairs of omplex
roots −1.2 ± 0.6i, −0.3 ± i, 0.4 ± 0.5i for this polynomial.

Theorem 8 Every polynomial p of degree n with omplex oe ients ai and an 6= 0

an be represented as
n
Y
p(x) = ai (x − bi )
i=1

where bi are the roots of p .

Sum and produ t of roots of polynomials Any polynomial p(x) =

Pn
ai xi of i=0
degree n has n roots bi , i = 1, 2, 3..n. If ai are real, the sum and produ t of roots are
given by
n n
X an−1 Y a0
bi = − , bi = (−1)n
i=1
an i=1 an
4.1. NUMERICAL METHODS 73

Numeri al pro edure: Deation

Consider a polynomials of degree n > 2. Any one root b1 is found by any one of the
methods - bise tion, Newton-Raphson, se ant, Laguerre's et . The polynomial an
be written as a produ t p(x) = (x − b1 )q1 (x) where q1 (x) is a redu ed or deated
polynomial of degree n − 1. Also the roots of q1 are exa tly the remaining roots of p.
Any one root of q1 is then determined as b2 . Deating q1 , q1 (x) = (x − b2 )q2 (x). The
same pro edure may be repeated till the degree of qi is 2. Then quadrati formula gives
the last two roots. This method of su essive deation has the following advantages.

1. Deation is just polynomial division.

2. The eort of nding a root generally de reases in ea h step.

3. The method annot onverge twi e to the same non-multiple root.

4. Su essive Deation is numeri ally stable, if the root of smallest absolute value
is divided out in ea h step.

Laguerre's Method of nding polynomial roots: Polynomial roots an be ra-

tional, irrational, real or omplex, this method will onverge to all types of roots.

The basi prin iple of this iterative method is given below.

Let p(x) be a nth degree polynomial.
n
X
2 n
p(x) = a0 + a1 x + a2 x + ... + an x = ai xi
i=0

If the n roots are bi , i = 1, 2, 3, ...n.

n
Y
p(x) = (x − b1 )(x − b2 ).....(x − bn ) = (x − bi )
i=1

Taking logarithm of modulus on both sides

n
X
ln |p(x)| = ln |(x − b1 )| + ln |(x − b2 )|..... + ln |(x − bn )| = ln |(x − bi )|
i=1

The modulus is taken sin e logarithm of negative real numbers and omplex numbers
are not dened. Dierentiating with respe t to x

d ln |p(x)| 1 1 1
=+ + ..... +
dx (x − b1 ) (x − b2 ) (x − bn )
74 CHAPTER 4. NUMERICAL ANALYSIS

n
p′ (x) X 1
= = c(say)
p(x) i=1
(x − bi )

Dierentiating again

d2 ln |p(x)| 1 1 1
2
=− 2
− 2
..... −
dx (x − b1 ) (x − b2 ) (x − bn )2
 ′ 2 n
p′′ (x)p(x) − p′ (x)p′ (x) p′′ (x) p (x) X 1
= − =−
p(x).p(x) p(x) p(x) i=1
(x − bi )2
2 n
p′ (x) p′′ (x) X

1
− = = d(say)
p(x) p(x) i=1
(x − bi )2

Let xj − bj = ej where xj is the j th trial root and bj , the a tual root. To nd ej ,
we assume that all other roots bi are equidistant from xj . Let xj − bi = s, i =
1, 2, .., n and i 6= j . Then

1 1 1 1 1 n−1
c = + + ..... + + .... + = +
s s ej s ej s

Similarly
1 n−1
d= 2
+
ej s2
Eliminating s and solving for ej
n
ej = p
c± (n − 1)(n d − c2 )

For |ej | to be small, the moduluspof denominator |c ± (n − 1)(n d − c2 )| must be

p

large. If
pc < 0, denominator is c − (n − 1)(n d − c ). If c > 0, the denominator must
2

be c + (n − 1)(n d − c2 ). To redu e ej , the pro ess is repeated with xj → xj − ej till

reasonable a ura y is obtained.

The polynomial is then divided by (x − bj ) to get a lower degree polynomial q(x).

The root of q(x) is found using the same method. This pro edure is repeated, till all
the roots are obtained.

Example-1 : To nd the roots of the polynomial p(x) = 6x4 + 23x3 + 37x2 + 28x + 6
we nd that all oe ients are real and positive and rational. Hen e omplex and
irrational roots o ur in pairs. No hange of sign means no real positive root. p′ (x) =
24x3 + 69x2 + 74x + 28 and p′′ (x) = 72x2 + 138x + 74. Let us assume that xj = −1.0
4.1. NUMERICAL METHODS 75

in Legurre's pro edure

10

4
p(x)

0 b1 b2

−2

−4
−2.0 −1.5 −1.0 −0.5 0.0
x

Trial value: x=-1.0000

(1),p(-1.0000)=-2.0000,p'(-1.0000)=-1.0000,
p''(-1.0000)= 8.0000, =0.5000,d=4.2500, x=-1.5271,
(2)p(-1.5271)=0.2480,p'(-1.5271)=-9.5650,
p''(-1.5271)=31.1665, =-38.5659,d=1361.6680, x=-1.5000
(3) p(-1.5000)=-0.0000 .Hen e it is one root.

Hen e x = −1.5 is a root. (x + 1.5) = (2x + 3)/2 is a fa tor of given polynomial.

p(x)
q(x) = = 3x3 + 7x2 + 8x + 2
2x + 3

Trial value: x=0.0000

(1) q(0)=2.0000 q'(0)=8.0000
q''(0)=14.0000 =4.0000 d=9.0000, x=-0.3405
(2)q(0.3405)=-0.0310 q'(0.3405)=4.2761
q''(0.3405)=7.8702 =-137.7945 d=19240.9417, x=-0.3333
(3)q(0.3333)=0.0000 Hen e it is another root.

Hen e x = −0.333 is a root. (x + 0.3333) = (x + 1/3) = (3x + 1)/3 is a fa tor of given

polynomial.
q(x)
q1(x) = = x2 + 2x + 2
3x + 1
76 CHAPTER 4. NUMERICAL ANALYSIS

This quadrati equation has roots

√
−2 ± 22 − 4.1.2
x± = = −1 ± i
2
Hen e the set of roots are (x = (1/3, 3/2, 1 + i, 1 − i)

Example-2 : To nd the roots of the polynomial p(x) = x4 + 1x3 + 6x2 + 4x + 16 we

nd that all oe ients are real and positive. Hen e omplex and irrational roots o ur
in pairs. No hange of sign means no real positive root. p′ (x) = 4x3 + 3x2 + 12x + 4
and p′′ (x) = 12x2 + 6x + 12. We start with a small trial root xj = −1.0. The iterations
give following values. (1)xj = (−0.8
√ − 1.2499j), √
(2)xj = (−0.9851 − 1.7613j), (3)xj =
√ − 1.7320j) = −1 − 3j Sin e −1 − 3j is a root its omplex onjugate
(−1.0000
−1 + 3j is also a root. Hen e by fa tor theorem
√ √
p(x) = q(x)[x + 1 + 3j][x + 1 + 3j] = q(x)(x + 1)2 + 3 = q(x)(x2 + 2x + 4)

x4 + 1x3 + 6x2 + 4x + 16
q(x) = = x2 − x + 4
x2 + 2x + 4
The roots of this√quadrati equation are given by x± = 0.5 ± 1.9365j . Hen e the set of
roots are (−1 ± 3j, 0.5 ± 1.9365j)

Problem 8 Gas tank that is 10 meters in length (end to end) onsists of a right-
ylinder and is apped at either end by a hemisphere. What is the radius of the tank if
the volume is 50 ubi meters?

Answer: V olume = 2πr 3 /3 + πr 2 (L − 2r) + 2πr 3 /3 simplify to the form πr 3 − 15πr 2 +

75 = 0 The three roots are r = 14.89235732, 1.32108215, −1.21343947. The rst is not
possible as the total length is only 10m. Third value is una eptable as radius annot
be negative. So r = 1.32108215. Also 2πr 3 /3 + πr 2 (L − 2r) + 2πr 3 /3 = 50.00104m3

Example-3 : To nd the roots of a ubi polynomial p(x) = x3 + x − 10, we observe

that there is one sign hange so that there may be a positive root. Sin e oe ients
are real, omplex and/or irrational roots o ur in pairs. Sin e there are only 3 roots
one of them must be real. We start with a trial xj = 0 in Legurre's pro edure. The two
iterations give (1) xj = 1.4293 and (2) xj = 2.0068 so that p(x) = 0.0 Hen e (x − 2) is
a fa tor. Dividing with this fa tor

x3 + x − 10
q(x) = = x2 + 2x + 5
x−2
4.1. NUMERICAL METHODS 77

Using quadrati formula

√
−2 ± 4 − 20
x± = = −1 ± 2j
2
Hen e the roots are (2, −1 ± 2j)

Example-4 : To nd the roots of a ubi polynomial p(x) = 6x3 − 11x2 − 14x + 24,
we observe that there are 2 sign hanges so that there may be 2 positive roots. As
before we start with xj = 0 in Legurre's pro edure. Su essive iterations give xj =
0.9101, 1.3001, 1.3333(= 4/3) Hen e x − 4/3 or (3x − 4) is a fa tor.

6x3 − 11x2 − 14x + 24

q(x) = = 2x2 −x−6 = 2x2 −4x+3x−6 = 2x(x−2)+3(x−2) = (x−2)(2x+3)
3x − 4
The roots are x = 2 and x = −3/2. The set of 3 roots are x = 2, 4/3, −3/2.

Problem 9 Find roots of following polynomials using Legurre's method.

√
No. polynomial p(x) Answer(j = −1)
1 x4 − 10x3 + 35x2 − 50x + 24 1, 2, 3, 4
2 x4 − 2x3 + 2x − 1 1, −2, −3, 4
3 x4 − 3x3 − 3x2 + 11x − 6 2, −1/3, 3/2, 2
4 x4 − 4x3 + 6x2 − 4x + 1 1, 3, (2 ± 1j)
5 x4 + 4x3 + 6x2 + 4x + 5 2 ± j, 0 ± j
6 x3 − 3x2 + 3x − 1 (2 ± 3j), 4
7 x3 − 6x2 + 11x − 6 2, 3, 4
8 x3 − 4x2 + 5x − 2 (1 ± 5j), 3
9 x3 − 8x2 + 20x − 16 3, 3, 0.5

4.1.4 Monte Carlo Methods

Monte Carlo method is an iterative omputational method used to al ulate multi-

dimensional integrals and investigate the behaviour of physi al systems using sto hasti
methods. It is used as a statisti al tool in studying situations whi h are not amenable
to ompute using deterministi algorithms. With the development of powerful om-
puters and e ient algorithms, the Monte Carlo method is found to be very useful
in nding numeri al solutions to quantitative problems whi h are nonlinear and in-
volving un ertain parameters. In physi s, Monte Carlo methods are used in nu lear
physi s(nu lear model), mole ular dynami s, rystal physi s, statisti al physi s(Ising
model), X-ray Imaging, Ele tron Dynami s in Doped Semi ondu tors, Quantum hro-
modynami s et . This method is based on the following on epts and prin iples.
78 CHAPTER 4. NUMERICAL ANALYSIS

Denition 4 (Random variable) A random variable is an assignment of numbers

to possible out omes of random events.

For example, onsider tossing a pair of oins. The number of heads n showing when
the oins land is a random variable. It an be assigned the number 0 to the out ome
[Tail, Tail℄. That is n(T, T ) = 0. Similarly n(T, H) = n(H, T ) = 1 and n(H, H) = 2.

Denition 5 (Expe tation or Expe ted Value) The expe ted value of a random
variable is the long-term limiting average of its values in independent repeated events.

The expe ted value of the random variable X is denoted E[X℄. For example, while
tossing a oin, the fra tion of times the oin lands with head up is half when the
number of tosses are very large. It is expressed as E[n(H)] = 0.5. This idea an be
expressed mathemati ally as follows. Let a random variable X takes a values xi with
probability pi . If the sum of produ ts onverges absolutely to some value Y , that is,
n
X
lim pi .xi = Y
n→∞
i=1

, then Y is the expe tation of X . For a ontinuous random variable X , the expe tation
may be dened as Z
E[X] = xP (x)dx

where P (x) is the probability of random variable X having a value x.

Taking the expe ted value is a linear operation: if X and Y are two random
variables,E[X + Y ] = E[X] + E[Y ], and for any onstant a, E[aX] = aE[X]

Theorem 9 (Law of large numbers) As the number of trials of a random pro ess
in reases, the per entage dieren e between the expe ted values and a tual values (mean
measured values) of all random variables goes to zero.

The mean of values of x obtained from n independent trials is given by

N
1 X
hxi = xi
N i=1

The law of large numbers states that

lim E[x] − hxi = 0

N →∞
4.1. NUMERICAL METHODS 79

Simple integration

Basi Prin iple: Consider the denite integral

Rb
a
f (x)dx where f (x) is dierentiable
in the interval (a, b). By denition,
Z b N
X
f (x)dx = lim f (xi )δxi
a N →∞
i=0

where a ≤ xi ≤ b. If δxi is a onstant for all i, δxi = (b − a)/N .

b N
b−aX
Z
f (x)dx = lim f (xi )
a N →∞ N
i=0

if xi are hosen at random between a and b. But

N
1 X
f (xi ) = hf i
N i=0

N
1 X
lim f (xi ) = lim hf (xi )i = E[f ]
N →∞ N N →∞
i=0

where E[f ] is the expe tation by law of large numbers. If b − a) = L,

Z b
f (x)dx = Lhf i
a

For surfa e and volume integrals, the above relation be omes

Z Z
f (x)dx = Shf i, f (x)dx = V hf i
s v

The error e in the above al ulation is the dieren e between the expe ted value and the
mean value whi h is given by their standard deviation. For a k -dimensional integral,
it is given by r
hf 2 i − hf i2
e = ±τ
N
where τ is k -dimensional volume and hf 2 i = N i=0 f (xi ). It is only a rough estimate
2
P
of the error.

Numeri al pro edure: In order to integrate a fun tion over a ompli ated domain
D , Monte Carlo integration uses random points over some simple domain D ′ , whi h
en loses D . Ea h random number generated is then he ked to see whether it is within
80 CHAPTER 4. NUMERICAL ANALYSIS

D . Out of the n random numbers, if m are within D , then the ratio (m/n) gives the
ratio of n-dimensional volumes D/D ′ . For example, to al ulate the area D of one
quadrant of a ir le of radius r , we en lose it within a square (D ′ ) of side r . Pairs
of random numbers (xi , yi ) in the interval (0, r) are generated and he ked to see if
x2 + y 2 ≤ r 2 so that it is within D . The ratio m/n of the number of random pairs to
the total number of pairs gives the ratio of their areas D/D ′ . Then

D = D ′ (m/n)

1.0

0.8

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0

.
R 2π dθ √
For example, the integral 0 whi h is 2π/ 3 = 3.62759872847, gives the
2 + cos θ
following values.
No.of random points in (0, 2π) Value of integral
10000 3.63293517
20000 3.63805025
30000 3.63569129
40000 3.63152102
50000 3.62770939

Integration by Importan e Sampling

To redu e the varian e(standard deviation) in the al ulation of a denite integral

using Monte Carlo method, sampling of the data points is introdu ed. Sampling is
the pro ess of sele ting data from a domain of interest. By studying the sample, one
an generalize the results ba k to the data set from whi h sample hosen. Sampling
redu es volume of data to be pro essed, thereby redu ing omputing time.
4.1. NUMERICAL METHODS 81

Importan e sampling: The term importan e sampling refers to the pro ess of las-
sifying values of the input random variables in a simulation a ording to the impa t on
the quantity being estimated. If the large impa t values (important values) are empha-
sized by sampling more frequently, then the varian e of the estimated quantity an be
redu ed. Hen e, the basi methodology in importan e sampling is to hoose a distri-
bution whi h gives more weight to the important values. In a simulation, outputs are
again weighted to orre t for the use of the biased distribution. This ensures that the
new importan e sampling is unbiased. Thus the fundamental problem in importan e
sampling is to determine the distribution that is properly biased. Su h a distribution
saves large amounts of omputing time.

Integration using importan e sampling Consider the integral I = f dτ where

R
f is a ontinuous fun tion in some n-dimensional volume
R τ . Let f (x) = g(x)h(x) where
g(x) is a positive fun tion satisfying the ondition g(x)dτ = 1. Then the integral
be omes I = h(x)g(x) dτ . This an be interpreted as follows.
R

One an integrate f by sampling it with uniform probability density dτ in simple

Monte Carlo method. But the same an be done by sampling h(x) with non-uniform
probability density g(x)dτ . The generalized fundamental theorem is that the integral
of any fun tion f is estimated, using N sample points xi , ..., xN , by
b
f
Z
f (x)dτ = τ h i
a g

with an error estimate r

hf 2 /g 2 i − hf /gi2
error = ±τ
N
To hoose the probability distribution g(x), one pro eeds by sear hing a fun tion whi h
minimises the absolute value of error (variational method). This may be taken as a
onstraint so that Legrange's multiplier method an also be used for nding g(x).

4.1.5 Sampled Data

Sampled data refers to a subset of a large data system that is dis rete or ontinuous
and whi h has all the hara teristi s of the omplete data system. The size of the
sample required for this purpose is given by sampling theorem.

Theorem 10 ( Nyquist-Shannon Sampling Theorem) In order for a band-limited

signal of maximum frequen y νm to be re onstru ted fully, it must be sampled at a rate
f ≥ 2νm .
82 CHAPTER 4. NUMERICAL ANALYSIS

Here band-limited signal refers to a signal with a zero power for frequen ies ν > νm .
A signal sampled at f = 2νm is said to be Nyquist sampled, and fc is alled the
Nyquist riti al frequen y. No information is lost if a signal is sampled at fc , and no
additional information is gained by sampling faster than this rate. For example, to
sample a sine wave of frequen y ν , the minimum sampling rate is fc = 2ν . That is the
time interval ∆ between samples and period of sine wave T are related as Delta = T /2.
Hen e a onvenient hoi e is to sample at positive and negative peak of the wave. If h(t)
ontains frequen ies f outside of the range (−f c, f c) where fc is the Nyquist frequen y,
then the power ontent of these frequen ies is moved into that range −fc < f < fc so
that power spe trum is modied. This phenomenon is alled aliasing . Any frequen y
omponent outside of the range (−f c, f c) is translated into that range as a result of
dis rete sampling.

4.1.6 Dis rete Fourier Transform

Consider a fun tion h(t) of a ontinuous variable t transformed into a fun tion H of
another independent variable ω having range (−∞, ∞) by the equation
Z ∞
H(ω) = h(t)eiωt dt
−∞

where ω = 2πf . h(t) an be retrieved using the inverse transform

Z ∞
h(t) = H(ω)e−iωtdω
−∞

The integration in these transforms an be repla ed with a summation over the fun tion
values h(t) orresponding to properly sampled values of t. Consider N onse utive
sampled values at onstant separation ∆. Let

tk = k∆, hk ≡ h(tk ), k = 0, 1, 2, ..., N − 1

If the fun tion h(t) is ontinuous and not periodi , then we assume that the sampled
points are su h that h(t) is similar in stru ture at all times t. Even though all frequen-
ies in the Nyquist frequen y range (−fc , fc ) are possible, we an get Fourier transforms
H(fk ) only for N -frequen ies as there are only N -input samples. Consider N values
n
fn = , n = [N/2, (N/2) − 1, (N/2) − 2, ...(−N/2) + 1, −N/2]
N∆
The extreme values of n, that is ±N/2 exa tly orrespond to the lower and upper limits
±fc . If ωn = 2πfn , then the dis rete Fourier transform is given by
Z ∞ N −!
X N −!
X
iωn t (iωn tk )
H(ωn ) = h(t)e dt ≈ hk e ∆=∆ hk e(iωn k∆)
−∞ k=0 k=0
4.1. NUMERICAL METHODS 83

Substituting for ωn ∆ = 2πfn ∆ = 2πn/N in the exponent

N −!  
X 2πikn
H(ωn ) = ∆ hk exp
N
k=0

The quantity
N −!  
X 2πikn
hk exp = Hn
k=0
N
where Hn is alled the dis rete Fourier transform of the N points hk .

H(fn ) ≈ Hn ∆

The inverse dis rete Fourier transform is then given by

N −!  
1 X 2πikn
hk = Hn exp −
N n=0 N

Algorithm For n data points

1: Read input list x of length n
2: ω ← 2π/n
3: for p = 0 to n − 1 do
4: s←0
5: for q = 0 to n − 1 do
6: s ← s + xq eiωpq
7: end for
8: yp ← s
9: end for

Program: The following fun tion will al ulate the dis rete Fourier transform of list
of values x.

#input x is list, y is output list

from math import*
def dft(x):
n = len(x)
omega, y = 2*pi/n,[0℄*n
for p in xrange(n):
s = 0
for q in xrange(n):
s +=x[q℄*exp(omega*q*p*1j)
84 CHAPTER 4. NUMERICAL ANALYSIS

y[p℄ = s
return y
print dft([1,3,5,3,1,-1,-3℄)
Output
[(9+0j), (-3.49+11.41j), (-0.11+1.681j), (2.60-0.14j),
(2.60+0.14j), (-0.11-1.68j), (-3.49-11.41j)℄

Thus the al ulation of dis rete Fourier transform of N -sampled points requires N ×
N = N 2 omputations. There is an algorithm to redu e the number of omputations
N log2 N alled Fast Fourier Transform.

4.1.7 Fast Fourier Transform(FFT)

This algorithm was rst on eived by Gauss in the 18th entuary. With the advent of
modern omputers, Tukey and Cooley developed an algorithm to implement it on a
digital omputer. There are other algorithms similar to this one developed re ently.

The basi prin iple is the divide and onquer strategy just like any other large data
systems. The dis rete Fourier transform of n-sampled points requires N × N = N 2
omputations. If the data set is divided into two equal parts, ea h part requires N 2 /4
omputations for its Fourier transform. Hen e total number of omputations is only
N 2 /2 = N 2 /21 whi h shows a redu tion by N 2 /2. If ea h half is still divided, the
number of omputations gets redu ed to 4 × (N/4)2 = N 2 /4 = N 2 /22 In general if
the N = 2p -points are divided into M = 2q equal parts, the number of omputations
be omes N 2 /2M = 22p−q .

Pro edure: There are dierent algorithms for FFT. One of the simplest is the Sande-
Tukey algorithm. It is given below.

Let N = 2p an integer power of 2. If the length of the data set is not a power of two,
zeros may be added as data elements up to the next power of two. If fn are n-data
points separated by N equal intervals, then its dis rete Fourier transform (DFT) is
given by
N
X −1
Fk = fn W nk , k = 0, 1, 2, 3..N − 1
n=0

where W = e2πi/N . Let us divide the entire data into two sets-odd and even indi es-
4.1. NUMERICAL METHODS 85

ea h of length N/2 with

(N/2)−1 (N/2)−1
X X
2nk
Fk = f2n W + f2n+1 W (2n+1)k
n=0 n=0
(N/2)−1
X
W 2nk f2n + f2n+1 W k
 
=
n=0
(e) (o)
= Fk + W k Fk

(e) (o)
where Fk = N/2−1 W 2nk f2n and Fk = N/2−1 W 2nk f2n+1 , supers ript 'o' and 'e'
P P
n=0 n=0
stands for odd and even number4ed data. The rux of the solution is to onsider these
odd and even sets of N/2 numbers as transforms of sequen es of length N/2. This is
alled Danielson-Lan zos Lemma. It is found that this Lemma an be used re ursively.
(e) (o)
Having redu ed the problem of omputing Fk to that of omputing Fk and Fk , the
(e)
same redu tion of Fk to the problem of omputing the transform of its N/4 even-
(ee)
numbered input data (even k in f2k ) as Fk and N/4 odd-numbered data (odd k in
(eo) (o) (oe) (oo)
f2k ) Fk . Similarly division of Fk an also be done into Fk and Fk . In other
words, one an dene dis rete Fourier transforms of the points whi h are respe tively
even-even, even-odd, odd-even and odd-odd on the su essive subdivisions of the data.
Sin e N is a power of 2, it is evident that one an ontinue applying the Danielson-
Lan zos Lemma until we have subdivided the data all the way down to transforms of
length 1. The Fourier transform of length one is just the identity operation that opies
its one input number into its one output slot! In other words, for every pattern of
log2 N there is a one-point transform that is just one of the input numbers fn for some
n.

4.1.8 Shooting method

Shooting method is employed to solve ordinary se ond order dierential equations with
a pair of boundary onditions . It is a two-point boundary value problem. The bound-
ary onditions at the starting point do not determine a unique solution to start with.
Starting boundary onditions is almost ertain not to satisfy the boundary onditions
at the other boundary point. In general, iteration is required to orrelate boundary
onditions into a single global solution of the dierential equation. Dierential equa-
tions are to be integrated over the interval of interest several times. Only for linear
dierential equations, the number of iterations an be predi ted. Consider the general
form of a se ond order linear dierential equation for the fun tion y(x)

y ′′ (x) = p(x)y ′ (x) + q(x)y(x) + r(x)

86 CHAPTER 4. NUMERICAL ANALYSIS

with boundary ondition y(a) = α, y(b) = β . Suppose u(x) is a fun tion whi h satises
the above equation with initial onditions u(a) = α, u′ (a) = 0

u′′ (x) = p(x)u′ (x) + q(x)u(x) + r(x)

Let v(x) be a fun tion whi h satises the equation

v ′′ (x) = p(x)v ′ (x) + q(x)v(x), v(a) = 0, v ′ (a) = 1

Then the linear ombination y = u + cv is a solution.

y ′′ (x) = u′′(x) + cv ′′ (x)

= p(x)u′ (x) + q(x)u(x) + r(x) + c[p(x)v ′ (x) + q(x)v(x)]
= p(x)[u′ (x) + cv ′ (x)] + q(x)[u(x) + cv(x)] + r(x)
= p(x)y ′(x) + q(x)y + r(x)

To evaluate , we use the boundary ondition at x = b

β − u(b)
u(b) + cv(b) = β, c =
v(b)

Hen e the solution is

β − u(b)
y(x) = u(x) + v(x)
v(b)

Pro edure: Any se ond order linear dierential equation an be split into two ou-
pled rst order equations and solved by Runge-Kutta method if two initial onditions
are known. In linear shooting method, the following pro edure is followed.

1. First solve
u′(x) = s(x), u(a) = α
s′ (x) = p(x)s(x) + q(x)u(x) + r(x), s(a) = 0

2. Then solve
v ′ (x) = t(x), v(a) = 0
t′ (x) = p(x)t(x) + q(x)v(x), t(a) = 1

3. Finally, the desired solution y(x) is the linear ombination

β − u(b)
y(x) = u(x) + v(x)
v(b)
4.1. NUMERICAL METHODS 87

Eigenvalue problems: Consider boundary value problem

y ′′(x) = p(x)y ′ (x) + q(x)y(x) + r(x), y(a) = α, y(b) = β

. If q(x) = λ a onstant and r(x) = 0, it be omes an eigenvalue problem in dierential

equations.
y ′′(x) − p(x)y ′ (x) = λy(x), y(a) = α, y(b) = β
. Here λ is an eigenvalue and y(x) is an eigenfun tion orresponding to λ. The method
of solution involves the omputation y(b) for dierent λ and nd the one for whi h
y(b) = β . This an be done e iently if we nd the roots of the logarithmi derivative
as it will an el all multipli ative onstants from y and y ′.
 ′ 
y (x, λ
=0
y(x, λ) x=β

4.1.9 Relaxation method:

It is an approa h dierent from shooting method. The dierential equations are re-
pla ed by nite-dieren e equations between a set of points in the range of integration.
The onversion of a dierential operator to a dieren e operator is done as follows. Let
y0 , y1 , y2 be three points orresponding to the x-values x0 , x0 + δx, x0 + 2δx respe tively
 
′ dy y1 − y0 y1 − y0
y (x0 ) = ≈ =
dx x=0 x0 + δx − x0 δx
 ′
y1′ − y0′
 
′′ dy y2 − y1 y1 − y0 1 y2 − 2y1 + y0
y (x0 ) = ≈ ≈ − =
dx x=0 δx δx δx δx (δx)2

Pro edure: The method of solution involves the following steps. Let y(x) be the
unknown fun tion

1. The dierential equation is onverted into a dieren e equation.

2. A trial solution is assumed whi h onsists of values for the dependent variables
at ea h mesh point. It may not satisfy the desired nite-dieren e equation, nor
the required boundary onditions.

3. It is then substituted in the dieren e equation and a solution is obtained.

4. This solution is substituted ba k to the dieren e equation and solution is again

found.
88 CHAPTER 4. NUMERICAL ANALYSIS

5. This iterative pro ess is ontinued till the solution is in lose agreement with
the true solution. This is indi ated by the agreement with dieren e equation
and boundary ondition. Also further iteration will not hange the solution
signi antly.

Relaxation method is preferred over shooting method in the following situations.

• If the boundary onditions are subtle, or involve ompli ated algebrai relations.

• If the solution is smooth and not highly os illatory.

• If the dierential equations have extraneous solutions whi h disappears during

iteration. They will not appear in the nal solution satisfying all boundary
onditions.

• If a good initial guess is possible, relaxation methods are very e ient. Often
it may be ne essary to solve a problem many times, ea h time using a slightly
dierent value of some parameter like an eigenvalue. In that ase, the previous
solution is usually a good initial guess when the parameter is hanged.
Chapter 5

Simulations

5.1 A omputational approa h to physi s

Simulation in physi s refers to the imitation of the behaviour of physi al systems with
time (temporal evolution). Time evolution of a system follow deterministi laws. These
laws are invariably dierential equations. If the dierential equations are non-linear
or they are sets of oupled equations, analyti solutions are either too di ult or
impossible without heavy approximations. Numeri al solutions are the only alternative
in those ases. To redu e the errors in the results, often one has to use a large number
of time steps. A omputer be omes an absolute ne essity in su h ases. By hanging
variables in the simulation, the behaviour of the system under dierent ir umstan es
an be studied virtually. It thus leads to the on ept of a theoreti al lab.

Steps involved in Simulation

The number of steps involved depends generally on the omplexity of the phenomena
to be simulated. But the following 7 steps are mandatory.

1. Identify the property or phenomena of interest to be studied.

Eg. Motion of masses under mutual attra tion.
2. Choose the eld of for e whi h des ribes how the atoms or other parti les within
the system intera t with ea h other and also with the external world.
In the above ase, it an be an inverse-square law for e.
k
f=
r2
89
90 CHAPTER 5. SIMULATIONS

3. Create a set of variables that you may need to onstru t a simulation.

In the above ase, initial and urrent positions (x0 , y0 ), (x, y), initial and urrent
velo ities (vx0 , vy0), (vx, vy), a eleration a = f /m and time step dt

4. Derive an equation relating dierent variables. In most of the ases of interest in

physi s, it may be a dierential equation.
In the above ase, as a eleration is radial ax = a cos θ, ay = a sin θ = ky/r 3

d2 x
ax = = k.x/r 3
dt2

d2 y
ay = 2 = k.y/r 3
dt
5. Choose a suitable method of solution of the equation. For ordinary dierential
equations any of these methods- Runge-Kutta, predi tor- orre tor, Monte-Carlo,
Euler et .-may be used. In Euler method.

vx = vx0 + ax dt
vy = vy0 + ay dt
x = x0 + vx dt
y = y0 + vy dt

6. Solve the equation for dierent values of the variable starting from initial values
and in rementing in steps of proper size.
Cal ulate [x(t + n dt), y(t + n dt)] from [x{t + (n − 1)dt}, y{t + (n − 1)dt}].

7. Either print the output as a table of values or plot the output as a graph.

5.1.1 Simple harmoni os illator

Prin iple

A simple harmoni os illator is des ribed by the equation

d2 x
+ ω 2x = 0
dt2
This an be split into 3 equations using a eleration a, velo ity v and displa ement x
as follows.
a = −ω 2 x, δv = a δt, δx = v δt
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 91

As before, the last two relations an be expressed as

vi = vi−1 + a δt

Similarly the displa ement is given by

xi = xi−1 + vi δt

The x − t graph and x − v graph (Phase urve) are drawn for the os illator.

Program

from pylab import*

n=50
x=zeros(n,dtype=float)
t=zeros(n,dtype=float)
v=zeros(n,dtype=float)
a=zeros(n,dtype=float)
x[0℄,v[0℄,omega=0,2,1 #input('initial displa ement, velo ity and angular frequen y')
dt=2.1*pi/(n*omega)
a[0℄=-x[0℄*omega**2
for i in range(1,n):
a[i℄=-x[i-1℄*omega**2
v[i℄=v[i-1℄+a[i℄*dt
x[i℄=x[i-1℄+v[i℄*dt
t[i℄=t[i-1℄+dt
subplot(2,2,1)
xlabel("time")
ylabel("displa ement")
grid(True)
plot(t,x)
subplot(2,2,2)
xlabel("time")
ylabel("velo ity")
grid(True)
plot(t,v)
subplot(2,2,3)
xlabel("displa ement")
ylabel("velo ity")
grid(True)
plot(x,v)
subplot(2,2,4)
xlabel("time")
92 CHAPTER 5. SIMULATIONS

ylabel("a eleration")
grid(True)
plot(t,a)
show()

3 3
2 2
displacement

1 1
velocity
0 0
−1 −1
−2 −2
−3 −3
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
time time
3 3
2 2
acceleration

1 1
velocity

0 0
−1 −1
−2 −2
−3 −3
−3 −2 −1 0 1 2 3 0 1 2 3 4 5 6 7
displacement time

5.1.2 Central eld motion

A for e eld having a potential fun tion V (r, θ, φ) = V (r) is alled a entral eld. As
it depends only on r it has spheri al symmetry and onsequently angular momentum
is onserved. Gravitational eld and Coulomb eld are examples of entral elds. The
general form of the potential is V (r) = kr n where k is a onstant and n a real number.
Rutherford s attering is an example of motion in a repulsive entral eld.
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 93
B(xn , yn )
.
......
......
......
......
.
..
.......
.
.
.......
.......
.......
.......
.......
..
..
.......
.
....
.......
......
.......
........
.......
..
..
........
.
....
.......
........
.........
.......
........
..
..
..
.........
.
.....
.........
.........
.........
..........
..........
..
..
..
..
...........
.
.......
...........
............
............
.............
..............
...............
.
..................
........................
................................................................................................................................................
.
..
..
..
..
..
..
..
θ
A(x1 , y1 ) C
① b

Prin iple

Rutherford's experiment is to measure the dee tion of a beam of α− parti les by gold
nu lii due to Coulomb repulsion. The ele trostati for e is given by

Ze.2e
F~ = r̂
4πǫ0 r 2

Resolving and putting r 2 = x2 + y 2 , the omponents of a eleration are

2ze2 x
ax = (1)
4mπǫ0 r 3
2ze2 y
ay =
4mπǫ0 r 3

2ze2 d2 x d2 y
Putting c = , ax = 2 and ay = 2 one gets
4mπǫ0 dt dt

d2 x cx
2
= 2 (2)
dt (x + y 2 )3/2
d2 y cy
2
= 2
dt (x + y 2 )3/2

The velo ity and position of every α−parti le at dierent instants of time are then
determined by solving the above dierential equations numeri ally. From gure, if
(x1 , y1 ) are asymptoti points , △ABC is isos eles. If θ is the angle of dee tion,

xn − x0
cot (θ/2) =
yn − y0
94 CHAPTER 5. SIMULATIONS

Path of alpha particle Relation between b and theta

4.5 5.0

4.0 4.5

3.5 4.0

3.0 3.5

cot(theta/2)
2.5 3.0
y

2.0 2.5

1.5 2.0

1.0 1.5

0.5 1.0

0.0 0.5
−6 −4 −2 0 2 4 6 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
x impact parameter b

Algorithm
1: Read the initial values of velo ities and positions of parti les in the beam, the
impa t parameter and the time step dt.
2: for b=0 to 20 step 1 do
3: while kyk is below a xed value yn do
4: Cal ulate ax and ay using formulae 3 and 5.1.2
5: Cal ulate x and y using 4th -order Runge-Kutta Method.
6: plot (x, y )
7: end while
8: Cal ulate cot (θ/2) and b/ cot (θ/2)
9: plot cot (θ/2) against b
10: end for
11: End

Program

from pylab import*

=21.82743562;
dt=0.0001
x=zeros(10001,'float')
y=zeros(10001,'float')
b=linspa e(0,1,10)
otthetaby2=zeros(10,'float')
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 95

for j in range(10):
x[0℄,y[0℄,t,vx,vy=-5,b[j℄,0,10,0
for i in range(10000):
vx+=x[i℄* *dt/(x[i℄*x[i℄+y[i℄*y[i℄)**1.5
vy+=y[i℄* *dt/(x[i℄*x[i℄+y[i℄*y[i℄)**1.5
x[i+1℄=x[i℄+vx*dt
y[i+1℄=y[i℄+vy*dt
subplot(1,2,1)
xlabel('x')
ylabel('y')
title("Path of alpha parti le")
plot(x,y)
otthetaby2[j℄=(x[i℄-x[0℄)/(y[i℄-y[0℄)
subplot(1,2,2)
xlabel('impa t parameter b')
ylabel(' ot(theta/2)')
title("Relation between b and theta")
grid(True)
plot(b, otthetaby2)
show()

5.1.3 Monte-Carlo simulations- value of π

Prin iple

Points in the rst quadrant of a unit ir le entered at origin satises the following
inequalities.

1. 0 ≤ x ≤ +1
2. 0 ≤ y ≤ +1
3. x2 + y 2 ≤ 1 .

.The rst two onditions are satised by all points in a unit square in the rst quadrant
as shown in gure. The ratio of area of the se tor(= π/4) to the area of the square(= 12 )
is π/4. Four times this ratio gives the value of π .

To determine this ratio in Monte Carlo Method, pairs of pseudo-random numbers

in the range (0, 1) are generated for oordinates (x, y). Allpthese pairs fall within the
square, but only those points belong to the ir le for whi h x2 + y 2 ≤ 1. The number
96 CHAPTER 5. SIMULATIONS

of su h points N are ounted. The ratio of number N to the total number of points
gives the ratio of their areas.

In PYTHON there is one in-built random number generator using the multipli ative
ongruential re ursive method developed by D.Lehmer. The ith and (i + 1)th random
numbers are related as

xi+1 = (axi + c)mod m

where the multiplier a = 75 − 1 = 16, 806, the in rement c = 0 and the modulus
m = 231 − 1 = 2, 14, 74, 83, 647 are alled magi numbers. The re urren e relation
suggests that the random numbers will repeat with a period less than m. The magi
numbers for the set (a, m, c) are hosen so that period is ≈ m and every number
between 0 and m − 1 o ur at some point. The role of initial seed x0 has only little
ee t.

(0,1)
..................... (1,1)
................
............
...........
..........
.........
........
.......
......
.......
......
......
.....
.....
.....
.....
.....
....
....
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
..
.. (1,0)
(0,0)

Algorithm
1: Read N ,the ount of random numbers to be generated
2: j←0
3: for i = 1 to N do
4: x ← rand()
5: y ← rand()
6: if (x2 + y 2) ≤ 1 then
7: j ←j+1
8: end if
9: end for
10: π ← 4.j/i
11: Print π
12: End
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 97

Program

from random import random

j=0
for i in range(1000000):
if (random()**2+random()**2)<=1:j+=1
print "Value of pi = ",4.0*j/i

5.1.4 Logisti map

Prin iple

A map is a fun tion whi h relates the oordinates of a point Pn+1 in terms of those
of the previous point Pn . A map is always dis rete as it uses the previous value of
the dependent variable as the present value of independent variable. There is thus no
question of dierentiability for a map.

The logisti map is developed by Robert May in 1876 as a mathemati al model of

population growth whose generations do not overlap with a xed environment. It is
given by
xn+1 = cxn (1 − xn )
This is alled logisti map. where 0 < x < 1 and c > 1. A ontinuous form of this map
is the logisti equation f (x) = cx(1 − x)

Chara teristi s of logisti equation and map:

1. The roots of logisti equation are obtained by setting f (x) = 0. They are x=0
and x=1.

2.

df (x)
= c(1 − 2x)
dx

Extremum o urs at df(x)/dx=0 whi h is, at x = 1/2. This is a maximum

be ause d2 f (x)/dx2 = −c whi h is negative. This point x=1/2 is alled the
riti al point of the fun tion possessing only a single maximum in a given intervel
( 0 < x < 1) in this ase.
98 CHAPTER 5. SIMULATIONS

3. After some iterations of the map , it often onverges to some xed value alled
an 'attra tor'. Any further iteration of will yield the same value. If x∗n is su h a
value,

x∗n = cx∗n (1 − x∗n ) (3)

x∗n = 1 − 1/c

Condition for stability of attra tor

In the map if xn < 0, then iterations will lead xn+1 to − ∞. If xn = 0 then xn+1 is zero
always. For x = 1/c, xn+1 = 1 − 1/c = x∗n . The range 0 < xn ≤ 1/c is alled the 'basin
of attra tion' of x∗n . A value xn approa hes x∗n if su essive iterations bring it loser to
x∗n .

 xn+1 − x∗n 

 xn − x∗  < 1
 
n

 f (xn ) − x∗n 

 f (xn−1 ) − x∗  < 1
 
n

In the limit f (xn−1 ) − x∗n → 0

 
 df (xn ) 

 dxn 
 < 1
xn =x∗n
|c(1 − 2x∗n )| = |2 − c| < 1

This is possible only if 1 < c < 3.

When c = 3 the attra tor bifur ates to two xed points x∗1 and x∗2 in su h a way
that

x∗2 = f (x∗1 )
x∗1 = f (x∗2 )
x∗2 = f [f (x∗2 )]
= c2 x∗2 (1 − x∗2 )[1 − cx∗2 (1 − x∗2 )]

Ea h x2 is said to be a xed point of period 2. In general, if xp is a xed point of

period p , xp repeats after a set of p iterations of f. That is

f (p) (xp ) = f (f (f...p − times...(xp ) = xp

This bifur ation of the attra tor at c = 3 is alled pit hfork bifur ation due to its
shape.

|f (f (xn ))| ≤ 1
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 99
√
This requires c ≥ 1 + 6 = 3.449489743. Then ea h bran h of xed points bifur ates
into two separate bran hes. The points on these bran hes will be of period 4.

If c is further in reased, further bran hing o urs. Fixed points of period p give rise
to 2p bran hes. It is found that for c = 3.5699....., an innite number of bifur ations
o ur. In logisti map, xed points never repeat. The band of xed points forms a
ontinuum. Complete haos begins from this point. Thus bifur ation is the route to
haos for logisti equation.

Logistic Map
1.0

0.8

0.6
Population

0.4

0.2

0.01.0 1.5 2.0 2.5 3.0 3.5 4.0

Control Parameter
100 CHAPTER 5. SIMULATIONS

Program

from pylab import*

def f( , x): return * x * (1 - x)
i, f,x0,n,g =2.9,3.655, 0.1,50,1000
i, f=input("range of ontrol parameter = ")
s=( f- i)/1000.0
L ,Lx = [℄,[℄
for in arange( i, f, s):
x = x0
for i in range(g): x = f( , x)
p = 0
while p < n:
x = f( , x)
L .append( )
Lx.append(x)
p += 1
plot(L , Lx, ".")
xlabel("Control Parameter")
ylabel("Population")
show()

5.1.5 Driven LCR ir uit

C L

v(t)

Prin iple:

If a voltage v(t) is given to a series LCR- ir uit, Kir ho's voltage law gives

di q
L + Ri + = V (t)
dt C
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 101

Sin e urrent is rate of ow of harge

d2 q dq q
L + R + − V (t) = 0
dt2 dt C
This se ond order dierential equation an be split into two rst order oupled equa-
tions and solved simultaneously.

dq(t) di(t) R q V (t)

= i(t), = − i(t) − +
dt dt L LC L
These oupled equations an be solved numeri ally using Runge-Kutta fourth order
method. The general formula for oupled dierential equations is given below.

Let
dy dz
= f (x, y, z), = g(x, y, z)
dx dx
be two oupled equations with initial onditions y(x0) = y0, z(x0) = z0. Then the
values y(x + δx), z(x + δx) are given by a 4-step determination of slopes at x0, x0 +
δx/2, x0+δx/2, x0+δx su essively as follows. let δx = h whi h is the usual onvention.

k1=h f(x0,y0,z0)
m1=h g(x0,y0,z0)
k2=h f(x0+h/2,y0+k1/2,z0+m1/2)
m2=h g(x0+h/2,y0+k1/2,z0+m1/2)
k3=h f(x0+h/2,y0+k2/2,z0+m2/2)
m3=h g(x0+h/2,y0+k2/2,z0+m2/2)
k4=h f(x0+h,y0+k3,z0+m3)
m4=h g(x0+h,y0+k3,z0+m3)
y(x0+h)=y0+(k1+2 k2+2 k3+k4)/6
z(x0+h)=z0+(m1+2 m2+2 m3+m4)/6

Using the pair (y(x0 + h), z(x0 + h) the values at x0 + 2h (y(x0 + 2h), z(x0 + 2h) are
found using the above formula. The pro ess is repeated till we get the value of (y, z)
at the desired x.

Program:

#x=time, y= harge, z= urrent

#The following fun tion solves the
# oupled equations y'=f(x,y,z) and z'=g(x,y,z)
from pylab import*
102 CHAPTER 5. SIMULATIONS

def f(x,y,z):return z
def g(x,y,z):return (-0.2*z-y+sin(4*x))#L=1H,C=1F, omega=4/s,R=0.2ohm.

#Runge-Kutta fourth order fun tion

def rk4solution(f,g,x,y,z,h,n):
result=[[℄,[℄,[℄℄
for i in range(n):
k1=h*f(x,y,z)
m1=h*g(x,y,z)
k2=h*f(x+h/2,y+k1/2,z+m1/2)
m2=h*g(x+h/2,y+k1/2,z+m1/2)
k3=h*f(x+h/2,y+k2/2,z+m2/2)
m3=h*g(x+h/2,y+k2/2,z+m2/2)
k4=h*f(x+h,y+k3,z+m3)
m4=h*g(x+h,y+k3,z+m3)
x,y,z=x+h,y+(k1+2*k2+2*k3+k4)/6,z+(m1+2*m2+2*m3+m4)/6
result[0℄.append(x)
result[1℄.append(y)
result[2℄.append(z)
return result

s=rk4solution(f,g,0.,0.,0.,0.05,1000)
figure(1)
subplot(1,2,1)
xlabel('time')
ylabel(' harge')
title('Charge variation')
grid(True)
plot(s[0℄,s[1℄)
subplot(1,2,2)
xlabel('time')
ylabel('Current')
title('Current variation')
grid(True)
plot(s[0℄,s[2℄)
figure(2)
xlabel('Charge')
ylabel('Current')
title('Current-Charge plot')
grid(True)
plot(s[1℄,s[2℄)
show()
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 103

0.3
Charge variation Current variation

0.4
0.2

0.2
0.1
Current
charge

0.0 0.0

−0.1
−0.2

−0.2
−0.4

−0.3
0 10 20 30 40 50 0 10 20 30 40 50
time time

The graphs show some initial distortions but later starts os illating with the im-
pressed frequen y ω . It an be explained if we look at the analyti solution of the
104 CHAPTER 5. SIMULATIONS

dierential equation

d2 q dq q
L 2 + R + − V0 sin (ωt)(t) = 0
dt dt C

d2 q dq
2
+ 2γ + ω02 q − v0 sin (ωt)(t) = 0
t dt
where 2γ = R/L, ω02 = 1/LC, v0 = V (0)/L

q(t) = Ae−γt sin (ω ′ t + φ) + B sin (ωt + ψ)

where A is real and positive amplitude, φ, ψ are additional phases ω ′ = ω02 − γ 2 and
p

v0
B=p
(ω02 − ω 2)2 + 4γ 2 ω 2

The rst term of the solution represents damped os illations while the se ond term gives
the for ed os illations. They interfere to give some distorted waveforms. Gradually the
amplitude of damped os illation redu es to zero and the LCR- ir uit starts os illating
with impressed frequen y.

5.1.6 Standing waves

Standing waves are produ ed when a travelling wave gets ree ted and superimpose
with the original wave.

Prin iple

Let y1 = f (x − vt) be the forward wave and y2 = f (x + vt) be the ree ted wave. Then
y = y1 + y2 will be the omposite wave. If it is a harmoni wave of spatial frequen y
k = 2π/λ and angular frequen y ω = 2πν , it an be represented as y1 = A sin (kx − ωt)
and y2 = A sin (kx + ωt). Hen e superposed wave is given by

y = A sin (kx − ωt) + A sin (kx + ωt)

It an be seen that the superposed pattern onsists of points of zero amplitude alled
nodes . Energy is not transmitted through nodes. Hen e the name stationary waves.
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 105

Stationary waves for k=2,omega=1

2.0

1.5

1.0

0.5

0.0
x

−0.5

−1.0

−1.5

−2.0
0 2 4 6 8 10
displacement y

Program

from pylab import*

k,omega,=2.,1.
t=x=linspa e(0,10,100)
y=[sin(k*x-omega*i)+sin(k*x+omega*i) for i in t℄
xlabel('displa ement y')
ylabel('x')
title('Stationary waves for k=2,omega=1')
plot(x,y)
show()

5.1.7 Simulation of radioa tivity

Prin iple

Radioa tive de ay is an inherently non-deterministi pro ess that an be simulated

very naturally using the Monte Carlo method. The observation that the mean-life is a
hara teristi of the nu leus leads to the assumption that the probability P of any one
parti le de aying per unit time in a radioa tive sample is a onstant. Suppose that the
probability of any given atom de aying over a time interval ∆t is given by λ, where
0 < λ < 1. Then the history of a single atom an be simulated by hoosing a sequen e
106 CHAPTER 5. SIMULATIONS

of random numbers xk , k = 1, .. uniformly distributed on (0, 1). The atom survives

until the rst o uran e of xk < λ. This approa h an be used to simulate an ensemble
of N atoms. Let ∆N be the number of parti les that de ay in some small time interval
∆t. Then the de ay probability per parti le, ∆N/N , is proportional to the length of
the time interval over whi h we observe the parti le.

∆N(t)
= −λ∆t, ∆N(t) = −λ∆tN(t)
N(t)

This is a nite-dieren e equation in whi h ∆N(t) and ∆t are experimental observ-

ables. Hen e it annot be integrated the way one solves a dierential equation. But
numeri al or algebrai solutions are possible. Be ause the de ay pro ess is random, an
exa t value for ∆N(t) annot be predi ted. ∆N(t) may be taken as the average num-
ber of de ays when observations are made of many identi al systems of N radioa tive
parti les.

Algorithm
1: Read N ,D ,(initial number of parent and daughter atoms), maximum no of time
intervals m and de ay onstant λ
2: T ←0
3: while N > 0 and T < m do
4: NU ← N
5: for i = 1 to NU do
6: x ← rand()
7: if 0 < x ≤ lambda then
8: N ←N −1
9: end if
10: end for
11: T ←T +1
12: end while
13: End

Program:

from pylab import*

N,M,Lambda,T=100,10,0.21,0
NU=[N℄
while N>0 and T<M:
for i in range(N):
if random()<= Lambda:N-=1
5.1. A COMPUTATIONAL APPROACH TO PHYSICS 107

NU+=[N℄
T+=1
T=arange(T+1)
plot(T,NU)
plot(T,Y)
Y=NU[0℄*exp(-Lambda*T)
legend(['Simulated','Exponential'℄)
xlabel('time')
ylabel('No.of unde ayed atoms')
title('Radioa tive de ay')
grid(True)

It an be seen that as N in reases, the de ay graph oin ides with the exponential urve.
dN
Hen e the dierential equation = −λdt is only a large-number approximation of
N
∆N
the dieren e equation = −λ∆t
N

Radioactive decay
100
Simulated
Exponential
80
No.of undecayed atoms

60

40

20

0
0 2 4 6 8 10
time
Index

mth root, 25 deation, 73

Des artes Rule, 70
Fa tor theorem, 70 det, 45
Irrational Root Theorem, 70 determinant, 45
Rational Root Theorem, 70 di tionaries, 2
algorithm, 63 di tionary, 5, 12
aliasing, 82 dieren e, 10
angular frequen y, 104 dieren e operator, 87
append, 6 dierential operator, 87
array, 29 dis rete transform, 82
axis(), 50 domain, 61
dump, 16
bar, 50
barh, 50 eig, 46
Bessel's Fun tion, 59 eigenvalue, 46
binomial oe ients, 20 eigenve tor, 46
bise tion method, 25 eigh, 46
boundary onditions, 85 eigvals, 46
eigvalsh, 46
entigrade, 25 elif, 14
lose, 16 else, 14
math, 22 expe tation, 78
o-domain, 61 expe ted value, 78
ombination, 19 extend, 6
Complex Root Theorem, 70
on atenation, 6 fa torial, 19
onstru tor, 10 Fahrenheit, 25
ontour, 50 Fast Fourier Transform, 84
ontourf, 50 Fibona i series, 18
ount, 6 gure(), 50
ross, 39 le, 16
ross produ t, 40 Fourier transforms, 82
ubi polynomial, 76 frozenset, 5
ubi spline, 66 fun tion, 15

D.Lehmer, 96 Gamma fun tion, 56

Danielson-Lan zos Lemma, 85 Gaussian fun tion, 56

108
INDEX 109

genfromtxt, 35 multidimensional, 29
grid(), 50 multipli ative ongruential re ursive method,
96
Hailstorm number, 21 mutable, 5
harmoni wave, 104
hermitian matrix, 46 n-Zero theorem, 70
homogeneous, 29 ndarray, 29
Newton-Raphson method, 24, 25
if, 14 node, 104
image, 61 non-singular, 45
immutable, 5, 10, 12 norm, 45
importan e sampling, 81 numeri , 2
imshow, 54 Numpy, 49
imshow(), 1 numpy, 15
index, 7, 39 Nyquist riti al frequen y, 82
inner produ t, 39 Nyquist sampled, 82
input, 1 Nyquist-Shannon Sampling Theorem, 81
insert, 7
interpolate, 66 ogrid[℄, 51
interse tion, 10 one-to-one, 62
inv, 45 one-to-one mapping, 61
inverse fun tion, 61 onto mapping, 61
invertible matri es, 45 open, 16
irrational roots, 74 operand, 3
operator, 3
key, 12 orthogonality, 22
os, 15
Laguerre's Method, 73 outer, 40
law of large numbers, 78 outer produ t, 40
legend(), 50
Legendre Fun tion, 57 palindrome, 23
Legurre's pro edure, 76 parametri plot, 53
linalg, 45 Pas al's triangle, 20
linear algebra, 45 Permutation, 19
linear equation, 45 pi kle, 16
list, 2, 5 Pie harts, 53
load, 16 pinv, 45
logisti map, 97 plot(), 49
loglog plot, 51 Polar oordinate, 52
polar plot, 51
magi numbers, 96 pop, 7
math, 15 preimage, 61
Matplotlib, 49 prime number, 23
module, 15 pseudo-random numbers, 95
Monte Carlo, 77, 95 pseudoinverse, 45
110 INDEX

pylab, 15 String, 4, 5
string, 2
quadrati equation, 22 swapping method, 64
quadrati formula, 77 symmetri dieren e, 10
random, 15 text(), 50
random variable, 78 the horizontal line test, 62
rational number, 52 title(), 50
ravel, 33 transpose, 33
raw_input, 1 triangular sequen e, 20
read, 16 tuple, 5, 12
relaxation method, 87 tuples, 2
Remainder theorem, 70
remove, 7 unary operation, 42
reshape, 33 union, 10
reshape fun tion, 34 unit matrix, 30
resize, 33
reverse, 8 varian e, 80
rhodonea, 52 write, 16
Robert May, 97
rose urve, 52 xlabel(), 50
sampling, 80 ylabel(), 50
sampling theorem, 81
Sande-Tukey algorithm, 84
saveg(), 1
savetxt, 35
s atter plot, 51
semilogx, 51
semilogy, 51
shooting method, 85
show(), 1, 49
simple harmoni os illator, 90
simulation, 81
sinusoid, 52
skiprows, 35
sli e, 39
solve, 45
sort, 8
sorted, 8
spatial frequen y, 104
square matrix, 46
square root, 24
standard deviation, 79
standing waves, 104