A Comprehensive

Summary on

INTRODUCTORY
LINEAR
ALGEBRA
A Comprehensive Summary on Introductory Linear Algebra

A Comprehensive Summary on
Introductory Linear Algebra

mathvault.ca

Table of Contents 7.2.2 Procedures . . . . . . . . . 7

7.3 Properties on Matrices . . . . . . . 7

1 Matrix Terminologies 3 8 Elementary Matrices 7

8.1 Definition . . . . . . . . . . . . . . 7
2 Operations on Matrices 3 8.2 Facts . . . . . . . . . . . . . . . . . 8
2.1 Scalar Multiplication . . . . . . . . 3
2.2 Addition . . . . . . . . . . . . . . . 3 9 Diagonalization 8
2.3 Multiplication . . . . . . . . . . . . 3 9.1 Definitions . . . . . . . . . . . . . . 8
2.4 Transposition . . . . . . . . . . . . 3 9.2 Characteristic Polynomial . . . . . . 8
2.5 Trace . . . . . . . . . . . . . . . . . 4 9.3 Properties of Eigenvectors . . . . . . 8
9.4 Diagonalizable Matrices . . . . . . 8
3 Elementary Row Operations 4 9.4.1 Definition and Procedure . . 8
9.4.2 Properties of Diagonalizable
4 Row Reduction on Matrices 4 Matrices . . . . . . . . . . . 9
4.1 Preliminaries . . . . . . . . . . . . 4
4.2 Procedure . . . . . . . . . . . . . . 4 10 Basis and Related Topics 9
4.3 Comparison of Different Forms . . . 4 10.1 Definitions . . . . . . . . . . . . . . 9
10.2 Facts . . . . . . . . . . . . . . . . . 9
5 System of Linear Equations 4 10.3 Procedure for Basis Extraction . . . 10
5.1 Procedure for Solving a System . . . 5 10.4 Equivalences . . . . . . . . . . . . . 10
5.2 Facts . . . . . . . . . . . . . . . . . 5
11 Subspaces 10
6 Determinant 5 11.1 Definition . . . . . . . . . . . . . . 10
6.1 Definition . . . . . . . . . . . . . . 5 11.2 Examples of Subspace . . . . . . . . 10
6.2 Facts . . . . . . . . . . . . . . . . . 5 11.3 Standard Subspaces . . . . . . . . . 10
6.3 Properties on Matrix Rows and 11.3.1 Definitions . . . . . . . . . 10
Columns . . . . . . . . . . . . . . . 5 11.3.2 Bases of Standard Subspaces 11
6.4 Properties on Matrices . . . . . . . 6
12 Operations on Vectors 11
7 Inverse 6 12.1 Preliminaries . . . . . . . . . . . . 11
7.1 Invertibility . . . . . . . . . . . . . 6 12.2 Length . . . . . . . . . . . . . . . . 11
7.2 Procedures for Finding Inverses . . . 7 12.3 Dot Product . . . . . . . . . . . . . 11
7.2.1 Terminologies . . . . . . . . 7 12.3.1 Definition and Facts . . . . 11

Table of Contents 2
A Comprehensive Summary on Introductory Linear Algebra

12.3.2 Properties . . . . . . . . . . 12 • Scalar division can be defined simiarly i.e.,

12.4 Cross Product . . . . . . . . . . . . 12 A df 1

= A .
12.4.1 Definition and Facts . . . . 12 k k
12.4.2 Properties . . . . . . . . . . 12 • k1 (k2 A) = (k1 k2 )A
12.5 Projection-Related Operations . . . 12

13 2D/3D Vector Geometry 13 2.2 Addition

13.1 Equations . . . . . . . . . . . . . . 13
13.1.1 Lines . . . . . . . . . . . . . 13 • Requires two matrices of the same dimension.
13.1.2 Planes . . . . . . . . . . . . 13
13.2 Point vs. Point . . . . . . . . . . . . 13 • Preserves the dimension of the matrices.
13.3 Point vs. Line . . . . . . . . . . . . 13
• A+B =B+A
13.4 Point vs. Plane . . . . . . . . . . . . 14
13.5 Line vs. Line . . . . . . . . . . . . . 15 • (A + B) + C = A + (B + C)
13.6 Line vs. Plane . . . . . . . . . . . . 15
• k(A + B) = kA + kB
13.7 Plane vs. Plane . . . . . . . . . . . 16

14 Matrix Transformation 16
14.1 Preliminaries . . . . . . . . . . . . 16 2.3 Multiplication
14.2 Standard Matrix Transformations in 2D 16
• Requires two matrices with matching “inner
dimensions".
$ 1 Matrix Terminologies • Produces a matrix with the corresponding
“outer dimensions” (i.e., m × n times n × p →
• Diagonal matrix: A matrix whose non-zero en- m × p).
tries are found in the main diagonal only. • The ij entry of AB results from dot-multiplying
• Identity matrix: A diagonal, n × n matrix with the ith row of A with the jth column of B.
1 across the main diagonal. Usually denoted • (AB)C = A(BC), but AB 6= BA in general.
by I.
• (kA)B = k(AB) = A(kB)
• Upper-triangular matrix: A matrix whose non-
zero entries are found at or above the main • A(B +C) = AB +AC, (A+B)C = AC +BC
diagonal only.
• Lower-triangular matrix: A matrix whose non- 2.4 Transposition
zero entries are found at or below the main
diagonal only. • Reverses the dimension of the matrix.
• (AT )ij = Aji

$ 2 Operations on Matrices • ith row of AT corresponds to ith column of A.

• j th column of AT corresponds to j th row of A.
2.1 Scalar Multiplication • (AT )T = A, (kA)T = k(AT )

• Preserves the dimension of the matrix. • (A + B)T = AT + B T , (AB)T = B T AT

2 Operations on Matrices 3
A Comprehensive Summary on Introductory Linear Algebra

2.5 Trace 4.2 Procedure

Given a n × n matrix A, the trace of A — or Tr(A) To search for the ith leading number in a column,
for short — is the sum of all the entries in A’s main check the column from the ith entry and onwards:
diagonal.
• If all said entries are zero, search for the ith
• Tr(kA) = k Tr(A) leading number in the next column instead.

• Tr(A + B) = Tr(A) + Tr(B) • If not, use elementary row operations to:

• Tr(AT ) = Tr(A), Tr(AB) = Tr(BA) • Create a leading number in the ith entry
of the column.
• Reduce all entries beneath it to zero.
$ 3 Elementary Row Opera- and proceed to search for the (i + 1)th leading
tions number in the next column.
Once all the columns are handled, the matrix would
The three elementary row operations are: be in a ladder form, where
• Row Multiplication (Ri → kRi , k 6= 0) • Dividing each non-zero row with its leading
number would put the matrix into a row-
• Row Swapping (Ri ↔ Rj ) echelon form.
• Row Absorption (Ri → Ri + kRj ) • From here, further reducing all non-leading-
one entries in each column to zero would put
Note that each elementary row operation can be
the matrix into the reduced row-echelon form.
reversed by another elementary row operation of
the same type.
4.3 Comparison of Different
Forms
$ 4 Row Reduction on Ma-
trices • A ladder form is similar to a row-echelon form,
except that a non-zero row needs not start with
1.
4.1 Preliminaries
• In a row-echelon form, all entries beneath a
At the most fundamental level, to perform row reduc- leading one is 0. In a reduced row-echelon
tion on a matrix is to alternate between the following form, all entries above and beneath a leading
two processes: one is 0.

1. Finding a leading number (i.e., a small, non- • While a matrix can have several ladder forms
zero number) in a column — whenever appli- and row-echelon forms, it can only have one
cable. reduced row-echelon form.

2. Nullify all other entries in the same column.

By running through all the columns this way from
$ 5 System of Linear Equa-
left to right, the final matrix — also known as the tions
reduced matrix — can then be obtained.

5 System of Linear Equations 4

A Comprehensive Summary on Introductory Linear Algebra

5.1 Procedure for Solving a Sys- • A homogeneous system — a system whose con-
tem stant terms are all zero — has at least the trivial
solution.
To solve a system of linear equations with m equa-
tions and n variables using matrices, proceed as
follows: $ 6 Determinant
1. Convert the equations into an augmented ma-
trix — with the m × n coefficient matrix on the 6.1 Definition
left and the m × 1 constant matrix on the right.
Given an n × n matrix A, the determinant of A —
2. Reduce the augmented matrix into a ladder or det(A) for short — is a scalar quantity which can
form, or — if needed — a row-echelon form be defined recursively:
or a reduced row-echelon form. Once there,  
three scenarios ensue: • For a 2 × 2 matrix
a b
:
c d
• If an inconsistent row — a row with zero
everywhere except the last entry — pops
 
a b df
det = ad − bc
up during the row reduction process, then c d
the original system is unsolvable.
• For a n×n matrix A, Caij — the cofactor of the
• If the reduced matrix has n leading num- ij entry of A — is defined to be the signed de-
bers and no inconsistent row exists, then terminant of the matrix resulted from removing
the system has a unique solution. the ith row and the j th column of A.
• If the reduced matrix has less than n lead- • For a general n × n matrix A (with n ≥ 3),
ing numbers and no inconsistent row ex- the determinant can be defined as the cofactor
ists, then the system has infinitely many expansion along the first row of A:
solutions, in which case: df
det(A) = a11 Ca11 + · · · + a1n Ca1n
• When converted back into equation
form, the system will have less than
n leading variables. 6.2 Facts
• By turning the non-leading variables
Guveb an n × n matrix A, det(A) can be obtained
into parameters and applying back
by cofactor-expanding along any row or any column
substitution, we can then find the gen-
of A. As a result:
eral solution to the system — along
with the basic vectors that generate • If A has a row of zeros or a column of zeros,
them. then det(A) = 0.
• If A is upper or lower-triangular, then det(A)
5.2 Facts is the product of its main-diagonal entries.

• For a n-variable system with infinitely many

solutions: 6.3 Properties on Matrix Rows
and Columns
# of parameters = # of non-leading variables
= n − # of leading variables In what follows:

6 Determinant 5
A Comprehensive Summary on Introductory Linear Algebra

• All matrices presented are assumed to be n × n

.. ..
   
matrices. . .
Ri  Ri 
   
• Ri and Rj are assumed to be n-entry row vec-  
 ..
 = det  ... 
  
tors. det 
 .
  
R + kR  R 
 j i  j
• Ci and Cj are assumed to be n-entry column .. ..
vectors. . .

Row/Column Multiplication  
det · · · Ci · · · Cj + kCi · · · =
 
 .  . det · · · Ci · · · Cj · · ·
.. ..
det kRi  = k det Ri 
   
.. .. 6.4 Properties on Matrices
. .

Given a n × n matrix A:
   
det · · · kCi · · · = k det · · · Ci · · ·

• If A has a duplicate row or a duplicate column,

Row/Column Addition then det(A) = 0.
 ..  .
..
 . 
.. • det(kA) = k n det(A)
.
det Ri + Rj  = det Ri  + det Rj  • det(AT ) = det(A)
     
.. .. ..
. . . 1
• det(A−1 ) =
det(A)

n−1
•


det Adj(A) = det(A)
   
det · · · Ci + Cj · · · = det · · · Ci · · ·
 
+ det · · · Cj · · ·
In addition, if B is also a n × n matrix, then:
det(AB) = det(A) det(B)
Row/Column Swapping
In particular:
.. ..
m
   
. . det(Am ) = det(A) (where m ∈ N)
Rj   Ri 
   
 .   . 
 ..  = − det  .. 
det    
R 
 i
R 
 j $ 7 Inverse
.. ..
. .
7.1 Invertibility
 
det · · · Cj · · · Ci · · · = Given a n × n matrix A and the n × n identity matrix
 
− det · · · Ci · · · Cj · · · I, A is said to be invertible if and only if there is a
n × n matrix B such that:
Row/Column Absorption AB = I and BA = I

7 Inverse 6
A Comprehensive Summary on Introductory Linear Algebra

In which case, since B is the only matrix with such In particular, in the case of a 2 × 2 matrix:
properties, it is referred to as the inverse of A — or  
A−1 for short. d −b
−1
−c a

a b
The following claims are all equivalent: c d
=
ad − bc
• A is invertible.
• det(A) 6= 0 • Row reduction method: By writing A and I
alongside each other and carry out row reduc-
• The equation Ax = 0 (where x and 0 are n- tion until A is reduced to the identity matrix,
entry column vectors denoting the variable vec- the original I would be reduced to A−1 as well:
tor and the zero vector, respectively) has only
the trivial solution.    Row Reduction   −1 
A  I −−−−−−−→ I  A
• The reduced row-echelon form of A is I.
• The equation Ax = b has a unique solution
for each n-entry column vector b. 7.3 Properties on Matrices
• The rows of A are linearly independent. Given an invertible n × n matrix A:
• The columns of A are linearly independent. −1
• (A−1 ) =A
A−1
7.2 Procedures for Finding In- • (kA) −1
= (where k 6= 0)
k
verses
−1 T
• AT = (A−1 )
Ï 7.2.1 Terminologies
In addition, if B is also an invertible n × n matrix,
Given a n × n matrix A: then:
(AB)−1 = B −1 A−1
• The cofactor matrix of A is the n × n matrix
whose ij entry is the cofactor of aij . In particular:
m
• Adj(A), the adjoint of A, is the transpose of (Am )−1 = (A−1 ) (where m ∈ N)
the cofactor matrix of A.

$ 8 Elementary Matrices
Ï 7.2.2 Procedures
Given an invertible n × n matrix A, two methods 8.1 Definition
for finding A−1 exist:
An n × n elementary matrix is a matrix obtainable
• Adjoint method: Since A Adj(A) = det(A)I by performing one elementary row operation on
and det(A) 6= 0, A−1 can be determined using the n × n identity matrix I.
the following formula:
As a result, three types of elementary matrices exist:
−1 Adj(A)
A = • Those resulted from row multiplication
det(A)

8 Elementary Matrices 7
A Comprehensive Summary on Introductory Linear Algebra

• Those resulted from row swapping 9.2 Characteristic Polynomial

• Those resulted from row absorption
Given an n × n matrix A, the following claims are
equivalent:
8.2 Facts
• λ is an eigenvalue of A.
Given a matrix A with n rows: • The equation Ax = λx has a non-zero solution.

• Performing an elementary row operation on • The equation (λI − A)x = 0 has a non-zero
A is equivalent to left-multiplying A with the solution.
n × n elementary matrix associated with the • det(λI − A) = 0
said operation.
• Since each elementary row operation is re- In other words:
versible, each elementary matrix is invertible
• If we define det(xI − A) as the characteristic
In particular, if A is an n × n invertible matrix, then: polynomial of A, then its roots are precisely
the eigenvalues of A.
• A−1 can be conceived as the series of ele- • Once an eigenvalue λ is determined, its
mentary row operations leading A to I (i.e., eigenspace and basic eigenvectors can also be
A−1 = En . . . E1 ). found by solving the equation (λI − A)x = 0.
• Similarly, A can be conceived as the series of
elementary row operations leading I to A (i.e.,
A = E1−1 . . . En−1 ). 9.3 Properties of Eigenvectors
More schematically:
• Since eigenspaces of distinct eigenvalues are
E1 ··· En — apart from the zero vector — disjoint from
A

I each other, it follows that every eigenvector is

−1 ... −1
E1 En associated with a unique eigenvalue.
• The collection of all basic eigenvectors (from
the distinct eigenspaces) forms a linearly inde-
$ 9 Diagonalization pendent set.
• As a result, an n × n matrix can have at most
9.1 Definitions n basic eigenvectors.
Given an n × n matrix A:
9.4 Diagonalizable Matrices
• A number λ is called an eigenvalue of A if and
only if the equation Ax = λx has a non-zero
solution. In which case: Ï 9.4.1 Definition and Procedure
• The set of all the solutions is called the Given a n × n matrix A, A is said to be diagonal-
eigenspace of A with eigenvalue λ. izable if and only if there exists an n × n diagonal
matrix D and an n × n invertible matrix P such that:
• Each non-zero solution is called an eigen-
vector of A with eigenvalue λ. P −1 AP = D

9 Diagonalization 8
A Comprehensive Summary on Introductory Linear Algebra

In fact, it can be shown that: 10.1 Definitions

A is diagonalizable ⇐⇒ A has n linearly
Given a set of vectors v, v1 , . . . , vn from a vector
independent eigenvectors.
space V :
For example: df
• A linear combination of v1 , . . . , vn = a vector
• If A has n eigenvalues, then it is automatically of the form k1 v1 +· · ·+kn vn (for some numbers
diagonalizable. k1 , . . . , kn )
df
• If A has less than n eigenvalues, but neverthe- • Span(v1 , . . . , vn ) = the set of all linear com-
less possesses n basic eigenvectors, then it is binations of v1 , . . . , vn
still diagonalizable.
(In other words, to show that v is in the span of
In which case, if we let: v1 , . . . , vn is to show that v can be expressed
as a linear combination of v1 , . . . , vn .)
• v1 , . . . , vn to be the n basic eigenvectors asso-
ciated with the eigenvalues λ1 , . . . , λn , respec- • v1 , . . . , vn is a spanning set of V (equiv.,
df
tively. v1 , . . . , vn span V ) ⇐⇒ Span(v1 , . . . , vn ) =
V.
• P to be the n × n matrix v1 . . . vn .
 
df
• v1 , . . . , vn are linearly independent ⇐⇒ the
• D to be the n × n diagonal matrix with
equation x1 v1 + . . . + xn vn = 0 has only the
λ1 , . . . , λn in the main diagonal.
trivial solution.
then it would follow that: (In other words, the zero vector in V can be
expressed as a linear combination of v1 , . . . , vn
P −1 AP = D or A = P DP −1 in a unique way.)
df
• v1 , . . . , vn is a basis of V ⇐⇒ v1 , . . . , vn span
Ï 9.4.2 Properties of Diagonaliz- V and are linearly independent.
able Matrices
Given a n × n diagonalizable matrix A with eigen- 10.2 Facts
values λ1 , . . . , λn (with repeating multiplicities):
Given a series of vectors v1 , . . . , vn from a vector
• Am = P Dm P −1 (where m ∈ N) space V :

(Note: This formula can be used to compute • v1 , . . . , vn are linearly dependent ⇐⇒ one
any power of A quickly.) of the vector vi can be expressed as a linear
combination of the other vectors.
• Tr(A) = λ1 + · · · + λn
• v1 , . . . , vn is a basis of V =⇒ every vector in
• det(A) = λ1 × · · · × λn
V can be expressed as a linear combination of
V in a unique way.

$ 10 Basis and Related Top- In general, given two sets A and B:

ics If A is a linearly independent set in V and
B is a spanning set of V , then |A| ≤ |B|.

10 Basis and Related Topics 9

A Comprehensive Summary on Introductory Linear Algebra

In particular • v1 , . . . ,vn are linearly

 independent ⇐⇒ the
matrix v1 . . . vn can be reduced to a lad-
• If A and B are both basis of V , then |A| = |B|. der form (equiv., a row-echelon form or a re-
• In other words, any basis of V will have the duced row-echelon form) with n leading num-
same number of vectors. This number is known bers.
as the dimension of V — or dim(V ) for short. • Span(u1 , . . . , um ) = Span(v1 , . . . , vn ) ⇐⇒
As a result, given a series of vectors v1 , . . . , vn in each ui can be expressed as a linear combi-
V , we have that: nation of v1 , . . . , vn , and each vi can be ex-
pressed as a linear combination of u1 , . . . , um .
• n < dim(V ) =⇒ v1 , . . . , vn does not span V .
• n > dim(V ) =⇒ v1 , . . . , vn are not linearly
independent. $ 11 Subspaces
• n = dim(V ) and v1 , . . . , vn are linearly inde-
pendent =⇒ v1 , . . . , vn is a basis of V . 11.1 Definition
Given a subset S of a vector space V , S is called a
10.3 Procedure for Basis Ex- subspace of V if and only if all of following three
traction conditions hold:

Given a series of vectors v1 , . . . , vn from a vector 1. 0V ∈ S.

space V , a basis of Span(v1 , . . . , vn ) can be deter- 2. For all v1 , v2 ∈ S, v1 + v2 ∈ S.
mined as follows:
  3. For all v ∈ S and each number k, kv ∈ S.
v1
 .. 
1. Create the matrix  . .
vn 11.2 Examples of Subspace
2. Reduce the matrix to a ladder form (equiv., a
Given a vector space V and a series of vectors
row-echelon form, or a reduced row-echelon
v1 , . . . , vn in V , some examples of subspace in-
form). Once there:
clude:
• The non-zero rows of the reduced matrix
will form a basis of Span(v1 , . . . , vn ). • The trivial subspaces (i.e., {0v } and V )

• The number of those non-zero rows will • Span(v1 , . . . , vn )

be the dimension of Span(v1 , . . . , vn ). • A line through the origin (in R2 or R3 )
• A plane through the origin (in R3 )
10.4 Equivalences
Given a series of vectors v, v1 , . . . , vn , u1 , . . . , um 11.3 Standard Subspaces
from a vector space V :

• v is a linear combination of v1 , . . . , vn ⇐⇒
Ï 11.3.1 Definitions
the augmented matrix v1 . . . vn v is

Given a m × n matrix A:
solvable.

11 Subspaces 10
A Comprehensive Summary on Introductory Linear Algebra

• The row space of A — or Row(A) for short — 12.1 Preliminaries

is the span of the rows of A.
Given a number k and two vectors u = (u1 , . . . , un )
• The column space of A — or Col(A) for short
and v = (v1 , . . . , vn ) in Rn :
— is the span of the columns of A.
• The null space of A — or Null(A) for short — •
df
u + v = (u1 + v1 , . . . , un + vn )
is the set of all solutions to the homogeneous df
system Ax = 0. • kv = (kv1 , . . . , kvn )
df
• In particular, if A is n × n matrix and λ is • 0 = (0, . . . , 0)
an eigenvalue of A, then the eigenspace
| {z }
n times
df
of A (with eigenvalue λ) = Null(λI − A). In the case where u and v are non-zero vectors:
df
u and v are parallel ⇐⇒ u = kv for
Ï 11.3.2 Bases of Standard Sub- some number k.
spaces
Given a m×n matrix A, when A is reduced to a lad- 12.2 Length
der form (equiv., a row-echelon form or a reduced
row-echelon form) A0 : Given a vector v = (v1 , . . . , vn ) in Rn , the length of
v — or |v| for short — is defined as follows:
• The non-zero rows of A0 form a basis of q
df
Row(A). |v| = v12 + . . . + vn2
• The columns of A corresponding to the leading Note that:
columns of A0 form a basis of Col(A).
• |v| = 0 ⇐⇒ v = 0
• The basic vectors in the general solution of
the augmented matrix A0 0 form a basis of •

|kv| = |k||v|
Null(A), where:
df
# of basic vectors = dim(Null(A)) = Nullity(A) 12.3 Dot Product
Since in A0 , the number of non-zero rows is equal
to the number of leading numbers, we have that:
Ï 12.3.1 Definition and Facts
df
dim(Row(A)) = dim(Col(A)) = Rank(A) Given two vectors u = (u1 , . . . , un ) and v =
(v1 , . . . , vn ) in Rn :
Furthermore, since in the homogeneous system as-
df
sociated with A0 , the number of leading variables u · v = u1 v1 + · · · + un vn
and non-leading variables add up to n, we also have
that: In the case where u and v are non-zero vectors in
R3 (or in R2 ), we have that:
Rank(A) + Nullity(A) = n
u · v = |u| |v| cos θ
df
(θ = the angle between u and v)
$ 12 Operations on Vectors
From which it follows that:

12 Operations on Vectors 11
A Comprehensive Summary on Introductory Linear Algebra

• u · v = 0 ⇐⇒ u and v are perpendicular. • |u×v| is the area of the parallelogram spanned

by u and v.
• u · v > 0 ⇐⇒ u and v form an acute angle.
• u · v < 0 ⇐⇒ u and v form an obtuse angle.
Ï 12.4.2 Properties
Ï 12.3.2 Properties Given a number k and three vectors u, v, w in R3 ,
we have that:
Given a number k and three vectors u, v, w in Rn ,
we have that: • 0×v =v×0=0

• u · u = |u|2 • v×v =0

• u·0=0·u=0 • u × v = −(v × u)

• u·v =v·u • (ku) × v = k(u × v) = u × (kv)

• (ku) · v = k(u · v) = u · (kv) • u × (v + w) = u × v + u × w

• u · (v + w) = u · v + u · w • (u + v) × w = u × w + v × w
• (u + v) · w = u · w + v · w • (u × v) · u = 0, (u × v) · v = 0
 
u
• u · (v × w) = det v 
12.4 Cross Product

w

Ï 12.4.1 Definition and Facts

12.5 Projection-Related Opera-
Given two vectors u = (u1 , u2 , u3 ) and v =
(v1 , v2 , v3 ) in R3 : tions

u×v =
df Given a vector v and a non-zero directional vector
       d in R3 (or in R2 ):
u u3 u1 u3 u1 u2
det 2 , − det , det
v2 v3 v1 v3 v1 v2 df
• projd v = the projection of v onto d
In the case where u and v are non-zero vectors, we df
• oprojd v = the orthogonal projection of v
have that: onto d
df
|u × v| = |u| |v| sin θ • refld v = the reflection of v about d
df
(θ = the angle between u and v) oprojd v
projd v
From which it follows that: refld v v
d
u × v = 0 ⇐⇒ u and v are parallel.

In the case where u and v are non-parallel vectors:

In addition:
• u × v is a vector perpendicular to both u and v·d
v. • projd v = d
d·d
12 Operations on Vectors 12
A Comprehensive Summary on Introductory Linear Algebra

• oprojd v = v − projd v • Point-Normal Form: (X − P ) · n = 0

(since projd v + oprojd v = v) df
( X = (x, y, z) being the variable vector )
• refld v = 2 projd v − v • Standard Form: ax + by + cz = k
(since v + refld v = 2 projd v) (where (a, b, c) = n and k = ax0 + by0 + cz0 )

$ 13 2D/3D Vector Geome- 13.2 Point vs. Point

try
Given two points P and Q in R2 (or in R3 ):
13.1 Equations −→
• PQ = Q − P
−→
Ï 13.1.1 Lines • The distance between P and Q = |P Q|

Given a point P = (x0 , y0 .z0 ) and a directional

vector d = (dx , dy , dz ) in R3 , the line spanned by 13.3 Point vs. Line
d passing through P can be expressed in various
forms: Given a point Q and a line ` : P + dt in R2 (or R3 ):
• Point-Direction Form: P + dt df
Q is on ` ⇐⇒ Q = P + dt for some number t.
(t being a scalar parameter)
In the case where Q is not on `, the distance be-
tween Q and ` can be determined in three ways:

x = x 0 + d x t

• Component Form: y = y0 + dy t
 • Orthogonal Projection Approach
z = z0 + dz t

df −→
x − x0 y − y0 z − z0 1. Compute v = oprojd QP . Once there:
• Symmetric Form: = =
dx dy dz • |v| would give the distance between
(provided that dx , dy , dz 6= 0) Q and `.
• Q + v would give the point on ` clos-
Note that each of the forms above have an analogue
est to Q.
in R2 as well.
` Q+v d
Ï 13.1.2 Planes v P

Given a point P = (x0 , y0 , z0 ), two non-parallel Q

directional vectors d1 , d2 and an associated nor-
mal vector n in R3 , the plane spanned by d1 and • Dot Product Approach
d2 passing through P can be expressed in various
1. Let X = P + dt be the point on ` where
forms:
the shortest distance occurs.
• Point-Direction Form: P + d1 s + d2 t 2. By plugging X (in the above form) into
−−→
(s, t being scalar parameters) the equation QX · d = 0, we can solve

13 2D/3D Vector Geometry 13

A Comprehensive Summary on Introductory Linear Algebra

for the missing value of t — and hence • |v| would give the distance between
determine the coordinates of X as well. Q and P.
−−→
3. Once there, |QX| would give the distance • Q + v would give the point on P clos-
between Q and `. est to Q.

• Cross Product Approach n

The distance between Q and ` is calculated as Q+v P
−→
the height of the parallelogram spanned by QP v
and d: Q
Area
−→
z }| {
|QP × d| • Dot Product Approach (for P : P + d1 s + d2 t)
|d|
|{z}
Base 1. Let X = P + d1 s + d2 t be the point on
P where the shortest distance occurs.

13.4 Point vs. Plane 2. By plugging X (in the above form) into
the system:
Given a point Q = (x1 , y1 , z1 ) and a plane P in R3 : (−−→
QX · d1 = 0
−−→
• If P is in the point-direction form P + d1 s + QX · d2 = 0
d2 t:
we can solve for the missing values of s
df
Q is on P ⇐⇒ Q = P + d1 s + d2 t and t — and hence determine the coordi-
for some numbers s and t nates of X as well.
−−→
• If P is in the standard form ax + by + cz = k: 3. Once there, |QX| would give the distance
df between Q and P.
Q is on P ⇐⇒ ax1 + by1 + cz1 = k

In the case where Q is not on P, the distance be- • Intersection Point Approach (for P in standard
tween Q and P can be determined in three ways: form)

1. Let X be the point on P closest to Q.

• Projection Approach −−→
Since QX is parallel to n, X must be of
df −→ the form Q + nt for some number t.
1. Compute v = projn QP . Note that:
2. By plugging X (in the above form) into
• If P is in point-direction form, the
the equation of P, we can solve for the
cross product of the directional vec-
missing value of t — and hence determine
tors can be used as n.
the coordinates of X as well.
• If P is in standard form, then any −−→
point on the plane can be used as 3. Once there, |QX| would give the distance
P. between Q and P.

2. Once there:

13 2D/3D Vector Geometry 14

A Comprehensive Summary on Introductory Linear Algebra

13.5 Line vs. Line 2. By plugging X1 and X2 (in the above

forms) into the following system:
In R2 , a pair of two lines falls into exactly one of (−−−→
following categories: X1 X2 · d1 = 0
−−−→
X1 X2 · d2 = 0
• Parallel intersecting
we can solve for the missing values of s
• Parallel non-intersecting and t — and hence determine the coordi-
nates of X1 and X2 as well.
• Non-parallel intersecting
3. Once there, the distance between `1 and
−−−→
In contrast, a pair of two lines in R3 falls into exactly `2 is simply |X1 X2 |.
one of the following four categories:
13.6 Line vs. Plane
• Parallel intersecting (i.e., overlapping)
• Parallel non-intersecting In R3 , a line ` and a plane P must fall into exactly
one of the following categories:
• Non-parallel intersecting
• Parallel intersecting (i.e., overlapping)
• Non-parallel non-intersecting (i.e. skew)
• Parallel non-intersecting
Given two lines `1 : P1 + d1 s and `2 : P2 + d2 t: • Non-parallel intersecting

• `1 and `2 are parallel ⇐⇒ d1 and d2 are In what follows, we assume that a plane is always
parallel. converted into standard form for easier analysis.
More specifically, if ` is in the form (x(t), y(t), z(t))
• `1 and `2 are intersecting ⇐⇒ the equation with directional vector d and P is in the form ax +
P1 + d1 s = P2 + d2 t is solvable for some num- by + cz = k with n = (a, b, c), then:
bers s and t.
(in the case where such a (s, t) pair exists • ` and P are parallel ⇐⇒ d · n = 0
and is unique, the coordinates of the intersec- • ` and P are intersecting ⇐⇒ the equation
tion point can be determined by, say, back- ax(t) + by(t) + cz(t) = k is solvable for some
substituting the value of s into P1 + d1 s.) t. Moreover:

In the case where `1 and `2 don’t intersect, the dis- • If the equation holds for all t, then ` and
tance between them can be determined as follows: P are overlapping.
• If the equation holds for a single t, then
• If `1 and `2 are parallel, then the distance be-
` and P intersect at a unique point —
tween them is simply the distance between `1
whose coordinates can be determined
and any point on `2 .
by back-substituting the value of t into
• If `1 and `2 are non-parallel, then: ` : (x(t), y(t), z(t)).

1. The shortest distance must occur between In the case where ` and P are non-intersecting
a point X1 = P1 + d1 s on `1 , and a point (hence parallel), the distance between ` and P is
X2 = P2 + d2 t on `2 . simply the distance between P and any point on `.

13 2D/3D Vector Geometry 15

A Comprehensive Summary on Introductory Linear Algebra

13.7 Plane vs. Plane • f (v1 + v2 ) = f (v1 ) + f (v2 ) for all n-entry
vectors v1 and v2 .
In R3 , any pair of planes must fall into exactly one
• f (kv) = kf (v) for all numbers k and n-entry
of the following categories:
vectors v.
• Parallel intersecting (i.e., overlapping)
In fact, any function from Rn to Rm with these two
• Parallel non-intersecting properties must be a matrix transformation as well.
• Non-parallel intersecting

In what follows, we assume that a plane is always

14.2 Standard Matrix Transfor-
converted into standard form for easier analysis. mations in 2D
More specifically, given P1 : a1 x + b1 y + c1 z = k1
If f is a linear transformation on 2D vectors, then
and P2 : a2 x + b2 y + c2 z = k2 :
f must be a matrix transformation induced by the
• P1 and P2 are parallel ⇐⇒ (a1 , b1 , c1 ) = 2 × 2 matrix
(a2 , b2 , c2 )   
1
 
0
f f
• P1 and P2 are intersecting ⇐⇒ the system 0 1
(
a1 x + b 1 y + c 1 z = k 1 The following matrices operate on 2D vectors based
a2 x + b 2 y + c 2 z = k 2 on the line ` : y = mx:

is solvable for some x, y and z. In which case: Matrix Eigenvectors

(1, m) (λ = 1)
 
• If the solution set is generated by one pa- Projection 1 m
rameter, then P1 and P2 intersect at a line. (onto `) m m2 (m, −1) (λ = 0)
1 + m2
• If not, then P and Q are overlapping.
(1, m) (λ = 0)
 
Orthogonal m2 −m
In the case where P1 and P2 are non-intersecting Projection −m 1 (m, −1) (λ = 1)
(hence parallel), the distance between P1 and P2 is (onto `) 1 + m2
simply the distance between P1 and any point on
(1, m) (λ = 1)
 
P2 . Reflection 1 − m2 2m
(about `) 2m m2 − 1 (m, −1) (λ = −1)
1 + m2
$ 14 Matrix Transformation
The following matrix operates on 2D vectors by
14.1 Preliminaries applying a counter-clockwise rotation with angle
θ:  
cos θ − sin θ
A function f from Rn to Rm is a matrix transforma-
sin θ cos θ
tion if and only and there exists a m × n matrix M
such that:
f (v) = M v

In which case, f is also a linear transformation in

the sense that:

14 Matrix Transformation 16