Linear System Gaussian Elimination Matrices Elementary Matrices (EM) Using Gauss-Jordan Elimination to Find Inverse Types of Matrices

Made up of a system of linear equations 1. Locate the leftmost non-zero column. Bring a A 𝑚 × 𝑛 matrix can be generally written: An elementary matrix is a square matrix obtained If 𝑨 is invertible, then 𝑨 𝑰 − 𝐺𝐽𝐸 → 𝑰 𝑨−1 1. Square Matrix
nonzero entry to the top of the column if needed. from an identity matrix by a single ERO. Determinant A matrix is called a square matrix if it has the same
𝒂𝟏𝒙𝟏 + 𝒂𝟐𝒙𝟐 + ⋯ + 𝒂𝒏𝒙𝒏 = 𝒃, 𝑎1, … , 𝑎𝑛 ℝ
∈ 𝑎11 𝑎12 ⋯ 𝑎1𝑛 Types of Elementary Matrices number of rows and columns.
Let 𝑨 = 𝑎𝑖𝑗 be an 𝑛 𝗑 𝑛 matrix
Possible Solution Sets 2. For each row below the top row, add a suitable 𝑎21 𝑎22 ⋯ 𝑎2𝑛
𝑨= Case 1: Multiply a row by a non-zero constant
Consistent System Inconsistent System multiple of the top row to it so that the entry ⋮ ⋮ ⋱ ⋮ 𝒏 × 𝒏 matrix or a matrix of order 𝒏
Let 𝑴𝒊𝒋 be an (𝑛 − 1) 𝗑 (𝑛 − 1) matrix obtained
below 𝑎𝑚1 𝑎𝑚2 ⋯ 𝑎𝑚𝑛
Unique Solution from 𝑨 by deleting the 𝑖𝑡ℎ row and the j𝑡ℎ column.
No solutions the leading entry of the top row becomes zero. 𝑎11 𝑎12 ⋯ 𝑎1𝑛
Infinitely Many Solutions 𝑨 = 𝑎𝑖𝑗 𝑚𝗑𝑛, where 𝑎𝑖𝑗 is the 𝑖, 𝑗 𝑡ℎ entry of 𝑨 𝑎21 𝑎22 ⋯ 𝑎2𝑛
𝑎 𝑖𝑓 𝑛 = 1 ⋮ ⋮ ⋱ ⋮ Diagonal
Augmented Matrices 3. Repeat till REF is achieved det 𝑨 = |𝑨| = ቊ 11
where 𝑚 is the number of rows and 𝑛 the columns 𝑎11𝐴11 + ⋯ + 𝑎1𝑛𝐴1𝑛 𝑖𝑓 𝑛 > 1 𝑎𝑛1 𝑎𝑛2 ⋯ 𝑎𝑛𝑛 Entries
Given a system Gauss-Jordan Elimination
Equivalence of Matrices
4. Multiply a suitable constant to each row so Where 𝐴𝑖𝑗 = −1 𝑖+𝑗 det(𝑴𝒊𝒋) 2. Diagonal Matrix
Two matrices are equal if they have the same size
𝑎11𝑥1 + 𝑎12𝑥2 + ⋯ + 𝑎1𝑛𝑥𝑛 = 𝑏1 that all the leading entries becomes 1. 𝒄𝑹𝒊 This is called the 𝑖, 𝑗 -cofactor of 𝑨. A diagonal matrix is a square matrix with all the
and their corresponding entries are equal.
𝑎21𝑥1 + 𝑎22𝑥2 + ⋯ + 𝑎2𝑛𝑥𝑛 = 𝑏2 E.g. non-diagonal entries 0
⋮ Standard Determinants
5. Beginning with the last nonzero row and Basic Matrix Operations 1 0 0
𝑎𝑚1𝑥1 + 𝑎𝑚2𝑥1 + ⋯ + 𝑎𝑚𝑛𝑥𝑛 = 𝑏𝑚 0 𝒄 0 - Multiplying the second row by c Case 1: 2 X 2 Matrix
working upward, add a suitable multiple of each Let 𝑨 = 𝑎𝑖𝑗 𝑨 = 𝑎𝑖𝑗 is diagonal ↔ 𝑎𝑖𝑗 = 0 when 𝑖 ≠ 𝑗
𝑚𝗑𝑛, 𝑩 = 𝑏𝑖𝑗 𝑚𝗑𝑛 𝑛𝗑𝑛
row to the rows above to introduce zeros above 0 0 1
We can represent it as a Augmented Matrix Addition 𝑨 + 𝑩 = (𝑎𝑖𝑗 + 𝑏𝑖𝑗 )𝑚𝗑𝑛 𝑨 = 𝑎𝑑 − 𝑏𝑐
the leading entries till RREF is achieved. 𝑎11 0 ⋯ 0
Subtraction 𝑨 − 𝑩 = (𝑎𝑖𝑗 − 𝑏𝑖𝑗 )𝑚𝗑𝑛 Case 2: Interchange two rows
0 𝑎22 ⋯ 0
𝑎11 𝑎12 ⋯ 𝑎1𝑛 𝑏1 Solving Linear Systems Using GJE Case 2: 3 X 3 Matrix
Scalar Multiplication 𝑐𝑨 = (𝑐𝑎𝑖𝑗 )𝑚𝗑𝑛 ⋮ ⋮ ⋱ ⋮
𝑎21 𝑎22 ⋯ 𝑎2𝑛 𝑏2
0 0 ⋯ 𝑎𝑛𝑛
⋮ ⋮ 𝑎1𝑥 + 𝑏1𝑦 + 𝑐1𝑧 = 𝑑1
𝑎𝑚1 𝑎𝑚2 ⋯ 𝑎𝑚𝑛 𝑏𝑚 Advanced Matrix Operations
Given the system, ቐ𝑎2𝑥 + 𝑏2𝑦 + 𝑐2𝑧 = 𝑑2 3. Scalar Matrix
Matrix Multiplication
𝑎3𝑥 + 𝑏3𝑦 + 𝑐3𝑧 = 𝑑3 A scalar matrix is a diagonal matrix with all the
Elementary Row Operations (ERO) Let 𝑨 = 𝑎𝑖𝑗 𝑚𝗑𝒑 𝑎𝑛𝑑 𝑩 = 𝑏𝑖𝑗 𝒑𝗑𝑛 𝑨 = 𝑎 𝑒𝑖 − 𝑓ℎ − 𝑏 𝑑𝑖 − 𝑓𝑔 + 𝑐(𝑑ℎ − 𝑒𝑔) diagonal entries the same
Action Symbol Applying GJE can have 3 possible outcomes Case 3: Identity Matrix 0 𝑖𝑓 𝑖 ≠ 𝑗
𝑝 𝑨 = 𝑎𝑖𝑗 𝑛𝗑𝑛 is diagonal ↔ 𝑎𝑖𝑗 = ቊ
Possible Solutions/Outcomes 𝑹𝒊 ↔ 𝑹𝒋 𝑐 𝑖𝑓 𝑖 = 𝑗
Multiply a row by a nonzero constant 𝑐𝑅𝑖 𝑨𝑩 = ෍ 𝑎𝑖𝑘𝑏𝑘𝑗 E.g.
Case 1: Unique Solution If 𝑨 is a identity matrix, then its determinant is 1.
Interchange two rows 𝑅𝑖 ↔ 𝑅𝑗 𝑘=1 1 0 0
# of nonzero rows = # of variables Case 3: Triangular Matrix 𝑐 0 ⋯ 0
Add a multiple of one row to another 0 0 𝟏 - Interchanging second & third rows
𝑅𝑖 + 𝑐𝑅𝑗 • 𝑨𝑩 ≠ 𝑩𝑨 0 𝑐 ⋯ 0
0 𝟏 0 , 𝑐 𝑖𝑠 𝑎 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 ∈ ℝ
row 1 0 0𝛼 𝑨𝑩 is the pre-multiplication of 𝑨 to 𝑩 If 𝑨 is a triangular matrix, then 𝑨 is equal to the ⋮ ⋮ ⋱ ⋮
Row Equivalence 0 1 0 𝛽 , 𝛼, 𝛽, 𝛾 𝜖 ℝ 𝑩𝑨 is the post-multiplication of 𝑨 to 𝑩 product of the diagonal entries of 𝑨. 0 0 ⋯ 𝑐
Case 3: Add a multiple of one row to another row
0 0 1 𝛾
Two matrices are said to be row equivalent if one Case 4: At least 2 identical rows or columns
can be obtained from each other by EROs. • 𝑨𝑩 = 𝟎 does not imply 𝑨 = 𝟎 𝒐𝒓 𝑩 = 𝟎 4. Identity Matrix
Here, the system is consistent and has the unique An identity matrix is a scalar matrix with all the
Powers of Matrices
solution + The determinant of a square matrix with two diagonal entries = 1
If augmented matrices of two linear systems are 𝑥1 = 𝛼 Let 𝑨 be a square matrix and 𝑛 ∈ ℤ identical rows/columns is zero.
row equivalent, then the two systems have the 0 𝑖𝑓 𝑖 ≠ 𝑗
ቐ𝑥2 = 𝛽 𝑰 𝑖𝑓 𝑛 = 0 Cofactor Expansion 𝑨 = 𝑰𝑛 = 𝑎𝑖𝑗 𝑛𝗑𝑛 is identity ↔ 𝑎𝑖𝑗 = ቊ
same set of solutions. 𝑨−𝑛 = 𝑨−1 𝑛 1 𝑖𝑓 𝑖 = 𝑗
𝑥3 = 𝛾 𝑨𝑛 = ቐ 𝑨𝑨 … 𝑨 𝑖𝑓 𝑛 ≥ 1
Find any row/column with the easiest numbers.
Row-Echelon Form Case 2: Infinitely Many Solutions 𝑛 times 𝑨𝑟 𝑨𝒔 = 𝑨𝑟+𝑠 1 0 ⋯ 0
𝑹𝒋 + 𝒄𝑹𝒊
# of nonzero rows ≠ # of variables E.g. 𝑎11 𝑎12 𝑎13 0 1 ⋯ 0
Leading Entry = First non-zero number Laws
# parameters = # of variables - # of nonzero rows 1 0 0 𝑨 = 𝑎21 𝑎22 𝑎23 ⋮ ⋮ ⋱ ⋮
1. 𝑨+𝑩 = 𝑩+𝑨 0 1 𝟑 - Add 3 X third row to second row 𝑎31 𝑎32 𝑎33 0 0 ⋯ 1
1. If there are any rows that consist entirely of
1 0 0𝛼 2. 𝑨+ 𝑩+𝑪 = 𝑨+𝑩 +𝑪 0 0 1
zeros, then they are grouped together at the 𝑎22 𝑎23
0 1 0 𝛽 , 𝛼, 𝛽 𝜖 ℝ 𝑀11 = 5. Zero Matrix
bottom of the matrix. 3. 𝑎 𝑨 + 𝑩 = 𝑎𝑨 + 𝑎𝑩 𝑎32 𝑎33
0 0 00 𝑨 A matrix is called a zero matrix if all entries = 0.
4. 𝑎 + 𝑏 𝑨 = 𝑎𝑨 + 𝑏𝑨 Properties of Invertible Matrices = 𝑎11 𝗑 (−1)1+1det 𝑀11
2. In any two successive rows that do not consist 𝑎12 𝑎13
Here, the system is consistent and has the general 5. 𝑎 𝑏𝑨 = 𝑎𝑏 𝑨 = 𝑏(𝑎𝑨) If 𝑨 is invertible: + 𝑎21 𝗑 (−1)2+1 det 𝑀21 𝑀21 = 0 0 ⋯ 0
entirely of zeros, the leading entry in the lower 𝑎32 𝑎33
solution 6. 𝑨 𝑩𝑪 = 𝑨𝑩 𝑪 + 𝑎31 𝗑 (−1)3+1 det 𝑀31 0 0 ⋯ 0
row occurs further to the right than the leasing 1. 𝑨𝒙 = 𝟎 has only the trivial solution. 𝑎 𝑎13 𝟎𝑛𝗑𝑛 =
7. 𝑨 𝑩 + 𝑩′ = 𝑨𝑩 + 𝑨𝑩′ 𝑀31 = 12 ⋮ ⋮ ⋱ ⋮
entry in the higher row. The leading entry need 𝑥1 = 𝛼 𝑎22 𝑎23
8. 𝑪 + 𝑪′ 𝑨 = 𝑪𝑨 + 𝑪′𝑨 0 0 ⋯ 0
not be in consecutive columns. ቐ𝑥2 = 𝛽 , 𝑠 𝑖𝑠 𝑎𝑛 𝑎𝑟𝑏𝑖𝑡𝑟𝑎𝑟𝑦 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 2. The RREF of 𝑨 is an identity matrix and the REF
9. 𝑎 𝑨𝑩 = 𝑎𝑨 𝑩 = 𝑨(𝑎𝑩) How does ERO Change Determinant?
𝑥3 = 𝑠 of 𝑨 has no zero row. 6. Symmetric Matrix
3. Columns with leading entries are pivot 10. 𝑨𝑚𝗑𝑛𝟎𝑛𝘹𝑞 = 𝟎𝑚𝘹𝑞 𝑎𝑛𝑑 𝟎𝑝𝗑𝑚𝑨𝑚𝗑𝑛 = 𝟎𝑝𝗑𝑛 𝑨 –ERO→ 𝑩 Change in Determinant
Case 3: No Solution A symmetric matrix is a square matrix with all
columns. 11. 𝑨𝑚𝗑𝑛𝑰𝑛 = 𝑰𝑚𝑨𝑚𝗑𝑛 = 𝑨𝑚𝗑𝑛 3. 𝑨 can be expressed as a product of elementary 𝑐𝑅𝑖 det 𝑩 = 𝑐 ∙ det(𝑨)
entries reflected along the diagonal
1 0 0𝛼 matrices. 𝑅𝑖 ↔ 𝑅𝑗 det 𝑩 = −det(𝑨)
0 1 0 𝛽 , 𝛼, 𝛽, 𝛾 𝜖 ℝ Transpose of a Matrix Inverse of a Matrix 𝑅𝑖 + 𝑐𝑅𝑗 det 𝐵 = det(𝐴) 𝑨 = 𝑎𝑖𝑗 is symmetric ↔ 𝑎𝑖𝑗 = 𝑎𝑗𝑖 for all 𝑖, 𝑗
𝑛𝗑𝑛
0 0 0 𝛾 The transpose of a The inverse of a 𝑛 × 𝑛 4. The determinant of 𝑨 is not 0. Properties of Determinants
𝑚 × 𝑛 matrix 𝑨, matrix 𝑨, is 𝑨-1 Invertible Matrices as a Product of EMs det 𝑨𝑩 = det 𝑨 det(𝑩) 𝑎11 𝛽 ⋯ 𝛾
Here, the system is inconsistent (0 = 𝛾), and has Suppose rref(𝑨) = 𝑰 𝛽 𝑎22 ⋯ 𝛿
denoted by 𝑨T (or 𝑨t ), det 𝑎𝑨 = 𝑎 𝑛 det 𝑨 , 𝑛 is the order of 𝑨
no solution as the last column is a pivot column. 𝑨 is invertible if ∃ 𝑩 | ⋮ ⋮ ⋱ ⋮
𝑨𝑩 = 𝑰𝑛 & 𝐁𝐀 = 𝑰𝑛 1 𝛾 𝛿 ⋯ 𝑎𝑛𝑛
Reduced Row-Echelon Form Homogenous Liner System is the matrix obtained by There exists elementary matrices 𝑬𝟏𝑬𝟐, … , 𝑬𝒌 such det 𝑨−1 =
that 𝑬𝒌 ⋯ 𝑬𝟐𝑬𝟏𝑨 = 𝑰 det(𝑨)
1. The leading entry of every nonzero row is 1. interchanging the rows 𝑨 is singular if for all 𝑩,
𝑎11𝑥1 + 𝑎12𝑥2 + ⋯ + 𝑎1𝑛𝑥𝑛 = 0 and columns of 𝑨. *PRE-MULTIPLICATION 1 7. Upper/Lower Triangular Matrix
𝑨𝑩 ≠ 𝑰𝑛 and 𝐁𝐀 ≠ 𝑰𝑛 𝑨−1 = adj(𝑨)
𝑎21𝑥1 + 𝑎22𝑥2 + ⋯ + 𝑎2𝑛𝑥𝑛 = 0 det 𝑨 An upper/lower triangular matrix is a square
2. Each column that contains a leading entry has Given the HLS
⋮ Laws of Transposition Laws of Inversion Hence, 𝑨 = 𝑬𝟏−1𝑬𝟐−1 ⋯ 𝑬𝒌−1𝑰 = 𝑬𝟏−1𝑬𝟐−1 ⋯ 𝑬𝒌−1 matrix with all the lower/upper half entries = 0
zeros everywhere else, and this is a pivot column. 𝑎𝑚1𝑥1 + 𝑎𝑚2𝑥1 + ⋯ + 𝑎𝑚𝑛𝑥𝑛 = 0 𝑨𝑇 𝑇 = 𝑨 𝑨𝑩 = 𝑨𝑩′ → 𝑩 = 𝑩′
𝑨 + 𝑩 𝑇 = 𝑨𝑇 + 𝑩𝑇 𝑨−1 −1 = 𝑨 Matrices that are a product of EMs are invertible. 𝑨 = 𝑎𝑖𝑗 𝑛𝗑𝑛 is UT ↔ 𝑎𝑖𝑗 = 0 for all 𝑖 > 𝑗
1. A HLS has either only the trivial solution or Ways to find Inverse of Matrices 𝑨 = 𝑎𝑖𝑗 𝑛𝗑𝑛 is LT ↔ 𝑎𝑖𝑗 = 0 for all 𝑗 > 𝑖
𝑨𝑩 𝑇 = 𝑩𝑇𝑨𝑇 𝑎𝑨 −1 = (1/𝑎)𝑨−1
infinitely many solutions in addition to the trivial
det 𝑨𝑇 = det(𝑨) 𝑨𝑩 −1 = 𝑩−1𝑨−1 1. Gauss-Jordan Elimination of 𝑨 𝑰 → 𝑰 𝑨−1
solution. 𝑎11 𝑎12 ⋯ 𝑎1𝑛 𝑎11 0 ⋯ 0
- Look at Top Right
𝑨𝑇 −1 = (𝑨−1)𝑇 0 𝑎22 ⋯ 𝑎2𝑛 𝑎21 𝑎22 ⋯ 0
2. A HLS with more unknowns than equations has 𝑨−1 = 𝑩, 𝑩−1 = 𝑨, 𝑩𝑨 = 𝑰 ⋮ ⋮ ⋱ ⋮ ⋮ ⋮ ⋱ ⋮
2. Using adjoint and determinant
infinitely many solutions. 0 0 ⋯ 𝑎𝑛𝑛 𝑎𝑛1 𝑎𝑛2 ⋯ 𝑎𝑛𝑛
det(𝑨) ≠ 𝟎 - Look at next page
 Adjoint/Adjugate Solving Linear Systems Using Cramer’s Rule Vectors Linear Combination Subspaces
Let Example of using Cramer’s Rule to solve a linear A vector is a directed line segment. A vector 𝒗 is a linear combination of vectors (𝒖𝟏, 𝒖𝟐 , … , 𝒖𝒏) if 𝒗 = 𝑎𝒖𝟏 + 𝑏𝒖𝟐 + ⋯ + 𝑛𝒖𝒏 A subspace must satisfy the following conditions
𝑎11 𝑎12 ⋯ 𝑎1𝑛 system. Determine if a Vector is a Linear Combination
𝑎21 𝑎22 ⋯ 𝑎2𝑛 Vectors have 1) Direction and 2) Magnitude Let 𝑆 = (𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌) be a subspace of 𝐑𝑛, then
𝑨=
⋮ ⋮ ⋱ ⋮ Vectors are equal iff their Directions and 𝒗 = 𝑎𝒖𝟏 + 𝑏𝒖𝟐 + ⋯ + 𝑛𝒖𝑛
𝑎𝑛1 𝑎𝑛2 ⋯ 𝑎𝑛𝑛 Magnitudes are equal. 1. 𝟎 is in span 𝑆
𝑎 𝒖𝟏𝟏, 𝒖𝟏𝟐, … , 𝒖𝟏𝒏 + b 𝒖𝟐𝟏, 𝒖𝟐𝟐, … , 𝒖𝟐𝒏 + ⋯ + n 𝒖𝒏𝟏 , 𝒖𝒏𝟐, … , 𝒖𝒏𝒏 = 𝒗𝟏, 𝒗𝟐, … , 𝒗𝒏
Vector Operations 2. For any 𝒗𝟏, 𝒗𝟐, … , 𝒗𝒓 in span 𝑆 and 𝒄𝟏, 𝒄𝟐, … , 𝒄𝒓 in 𝐑,
Then
Equality 𝒖 = 𝒗 iff 𝒖𝒊 = 𝒗𝒊 ∀ 𝑖 𝒖𝟏𝟏 𝒖𝟐𝟏 𝒖𝒏𝟏 𝒗𝟏 𝑎𝒖𝟏𝟏 + 𝑏𝒖𝟐𝟏 + ⋯ + 𝑛𝒖𝒏𝟏 = 𝒗𝟏 𝒄𝟏𝒗𝟏 + 𝒄𝟐𝒗𝟐 + ⋯ + 𝒄𝒓𝒗𝒓 is in span(S).
adj 𝑨 =
𝐴11 𝐴12 ⋯ 𝐴1𝑛 𝑇 𝐴11 𝐴21 ⋯ 𝐴𝑛1 Addition 𝒖 + 𝒗 = (𝑢1 + 𝑣1, 𝑢2 + 𝑣2, … , 𝑢𝑛 + 𝑣𝑛) 𝒖𝟏𝟐 𝒖𝟐𝟐 𝒖𝒏𝟐 𝒗𝟐 𝑎𝒖𝟏2 + 𝑏𝒖𝟐𝟐 + ⋯ + 𝑛𝒖𝒏𝟐 = 𝒗𝟐
𝑎 +𝑏 +⋯+𝑛 = → * {𝟎} is zero space and is a subspace too
𝐴21 𝐴22 ⋯ 𝐴2𝑛 𝐴12 𝐴22 ⋯ 𝐴𝑛2 Scalar X 𝑐𝒖 = (𝑐𝑢1, 𝑐𝑢2, … , 𝑐𝑢𝑛) ⋮ ⋮ ⋮ ⋮ ⋮
= 𝒖𝟏𝒏 𝒖𝟐𝒏 𝒖𝒏𝒏 𝒗𝒏 𝑎𝒖𝟏𝑛 + 𝑏𝒖𝟐𝒏 + ⋯ + 𝑛𝒖𝒏𝒏 = 𝒗𝒏 Determining if 𝑉 is a Subspace
⋮ ⋮ ⋱ ⋮ ⋮ ⋮ ⋱ ⋮ Negative −𝒖 = (−𝑢1, −𝑢2, … , −𝑢𝑛)
𝐴𝑛1 𝐴𝑛2 ⋯ 𝐴𝑛𝑛 𝐴1𝑛 𝐴2𝑛 ⋯ 𝐴𝑛𝑛 1) Can we write 𝑉 as a span, or a linear combination of some
Minus 𝒖 − 𝒗 = (𝑢1 − 𝑣1, 𝑢2 − 𝑣2, … , 𝑢𝑛 − 𝑣𝑛) 𝒖𝟏𝟏 𝒖𝟐𝟏 ⋯ 𝒖𝒏𝟏 𝒗𝟏 vectors
where 𝐴𝑖𝑗 = −1 𝑖+𝑗 det 𝑴𝒊𝒋 , the 𝑖, 𝑗 -cofactor Laws 𝒖𝟏𝟐 𝒖𝟐𝟐 ⋯ 𝒖𝒏𝟐 𝒗𝟐 2) Does 𝑉 contain the origin?
will the matrix to use GJE on
of 𝑨 𝒖+𝒗 = 𝒗+𝒖 ⋮ ⋮
***DON’T FORGET THE TRANSPOSE*** 𝒖𝟏𝒏 𝒖𝟐𝒏 ⋯ 𝒖𝒏𝒏 𝒗𝒏 Both must be YES. To show 𝑉 is not a subspace, at least one has
𝒖+ 𝒗+𝒘 = 𝒖+𝒗 +𝒘
Calculating Adjoint 𝒖+𝟎 = 𝟎+𝒖 to be NO.
If the system is consistent, then 𝒗 is a linear combination of (𝒖𝟏, 𝒖𝟐, … , 𝒖𝒏)
𝒖 + (−𝒖) = 𝟎 The solution of this system will be the values of the coefficients, (𝑎, 𝑏, … , 𝑛) Examples of not a subspace
𝑎(𝑏𝒖) = 𝑎𝑏 𝒖
Spans and Linear Combinations
𝑎(𝒖 + 𝒗) = 𝑎𝒖 + 𝑎𝒗
(𝑎 + 𝑏)𝒖 = 𝑎𝒖 + 𝑏𝒖 The vectors 𝑆 = (𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌) span 𝐑𝑛 iff all vectors in the space are a linear
1𝒖 = 𝒖 combination of (𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌)

Fails Condition 1
Using Vectors to Represent Solution Sets The linear combination of any vectors in 𝐑𝑛 will be in 𝐑𝑛.
Case 1: Line in R2
Implicit: 𝑥, 𝑦 𝑎𝑥 + 𝑏𝑦 = 𝑐 If 𝑘 < 𝑛, then 𝑆 cannot span 𝐑𝑛 → Cardinality of 𝑆 cannot be less than 𝑛
𝑐 − 𝑏𝑡 Spans and Subsets
, 𝑡 | 𝑡 𝑖𝑛 𝐑 𝑖𝑓 𝑎 ≠ 0 Let 𝑆1 = {𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌} and 𝑆1 = {𝒗𝟏, 𝒗𝟐, … , 𝒗𝒎} be subsets of vectors in 𝐑𝑛.
𝑎
Explicit: 𝑐 − 𝑎𝑡
𝑡, | 𝑡 𝑖𝑛 𝐑 𝑖𝑓 𝑏 ≠ 0 Then span 𝑆1 ⊆ span 𝑆2 iff each 𝒖𝒊 is a linear combination of 𝒗𝟏, 𝒗𝟐, … , 𝒗𝒎.
𝑏
And span 𝑆2 ⊆ span 𝑆1 iff each 𝒗𝒊 is a linear combination of𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌.
Case 2: Plane in R3
* span 𝑆1 = span 𝑆2 iff span 𝑆1 ⊆ span 𝑆2 and span 𝑆2 ⊆ span 𝑆1
Implicit: 𝑥, 𝑦, 𝑧 𝑎𝑥 + 𝑏𝑦 + 𝑐𝑧 = 𝑑
𝑑 − 𝑏𝑠 − 𝑐𝑡
, 𝑠, 𝑡 | 𝑠, 𝑡 𝑖𝑛 𝐑 𝑖𝑓 𝑎 ≠ 0 Example:

Fails Condition 2
𝑎 To show span{𝒖𝟏, 𝒖𝟐 , 𝒖𝟑} ⊆ span{𝒗𝟏, 𝒗𝟐}, write all 𝒖𝒊 as a linear combination of all 𝒗
𝑑 − 𝑎𝑠 − 𝑐𝑡
𝑠, , 𝑡 | 𝑠, 𝑡 𝑖𝑛 𝐑 𝑖𝑓 𝑏 ≠ 0
Explicit: 𝑏
𝑑 − 𝑎𝑠 − 𝑏𝑡
𝑠, 𝑡, | 𝑠, 𝑡 𝑖𝑛 𝐑 𝑖𝑓 𝑐 ≠ 0
𝑏

Case 3: Line in R3
𝑎0, 𝑏0, 𝑐0 + 𝑡(𝑎, 𝑏, 𝑐) 𝑡 𝑖𝑛 𝐑
𝑎0, 𝑏0, 𝑐0 is a point on the line If this system is consistent, then span{𝒖𝟏, 𝒖𝟐, 𝒖𝟑} ⊆ span{𝒗𝟏, 𝒗𝟐}
Explicit: (𝑎, 𝑏, 𝑐) is the direction of the line Subspaces in 𝐑𝑛
where 𝑎0, 𝑏0, 𝑐0, 𝑎, 𝑏, 𝑐 are real constants, Redundant Vectors All the subspaces of 𝐑2 All the subspaces of 𝐑3
and 𝑎, 𝑏, 𝑐 are not all zero If span 𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌+𝟏 = span{𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌 }, that is, removing 𝒖𝒌+𝟏 has no effect on - {𝟎}
- {𝟎}
Example: the span of the subset, then 𝒖𝒌+𝟏 is a redundant/useless vector. - Lines through the origin
- Lines through the origin
Types of Spans - Planes containing the origin
- 𝐑2
The solution of - 𝐑3
Case 1: A single vector in 𝐑2
Subspace as a solution set of HLS
The span of a single vector 𝒖 = (𝒖𝟏, 𝒖𝟐) in 𝐑2 is a line given by The solution set (solution space) of a homogeneous system of
span 𝒖 = 𝑐𝒖 | 𝑐 𝑖𝑛 𝐑 = 𝑥, 𝑦 𝑎𝑥 + 𝑏𝑦 = 0} linear equations in 𝒏 variables is a subspace of 𝐑𝑛.
Case 2: A single vector in 𝐑3 Case 1: Only the trivial solution
Using Adjoint to Find Inverse
This will be the zero space, {𝟎}.
The span of a single vector 𝒖 = (𝒖𝟏, 𝒖𝟐, 𝒖𝟑) in 𝐑3 is a line given by
span 𝒖 = 𝑐𝒖 | 𝑐 𝑖𝑛 𝐑 = 𝑐𝒖𝟏, 𝑐𝒖𝟐, 𝑐𝒖𝟑| 𝑐 𝑖𝑛 𝐑 Case 2: Infinitely many solutions

Types of Sets of Vectors in a Space Case 3: 2 vectors in 𝐑2 Given 𝒙𝟏 , 𝒙𝟐, … , 𝒙𝒏 are the variables to the system,
Term Definition The solution space will be
If vectors 𝒖 and 𝒗 are not parallel and linearly independent, then this is the plane 𝐑2 itself
The set of all 𝑛-vectors of real numbers is called the Euclidean 𝑛-space, denoted by 𝐑𝑛.
Space Case 4: 2 vectors in 𝐑3
𝒖 in 𝐑𝑛 iff 𝒖 = (𝒖𝟏, 𝒖𝟐, … , 𝒖𝒏) for some 𝒖𝟏, 𝒖𝟐, … , 𝒖𝒏 in 𝐑.
Subset A collection of vectors
The span of vectors 𝒖 = (𝒖𝟏, 𝒖𝟐, 𝒖𝟑) and 𝒗 = (𝒗𝟏, 𝒗𝟐, 𝒗𝟑) is a plane given by
Let 𝑆 = {𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌} be a subset of 𝐑𝑛.
span 𝒖, 𝒗 = 𝑠𝒖 + 𝑡𝒗| 𝑠, 𝑡 𝑖𝑛 𝐑 = 𝑥, 𝑦, 𝑧 𝑎𝑥 + 𝑏𝑦 + 𝑐𝑧 = 0
The subspace 𝑉 = 𝑐1𝒖𝟏 + 𝑐2𝒖𝟐 + ⋯ , 𝑐𝑘𝒖𝒌 𝑐1, 𝑐2, … , 𝑐𝑘 𝑖𝑛 𝐑} of 𝐑𝑛 is called the space
Span spanned by 𝑆 (or by 𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌). Lines and Planes in 𝐑3
𝟎 must be in in span(𝑆) for it to be a span.
To get a line in 𝐑3, we will need one point and the span of one direction vector
𝑉 = span 𝑆 𝑉 = span{𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌} 𝑆 spans 𝑉 𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌 spans 𝑉 𝐿 = 𝒙 + 𝒘 𝒘 ϵ 𝐿0} = 𝒙 + 𝒘 𝒘 ϵ span{𝒖}} = 𝒙 + 𝑡𝒖 𝑡 ϵ ℝ}.
Let 𝑆 = {𝒖𝟏, 𝒖𝟐, … , 𝒖𝒌} be a subset of 𝐑𝑛.
Subspace 𝑉 is a subspace of 𝐑𝑛 if 𝑉 = span 𝑆 To get a plane in 𝐑3, we will need one point and the span of 2 direction vectors
𝑉 is the subspace spanned by 𝑆 𝑆 spans the subspace 𝑉 𝐿 = 𝒙 + 𝒘 𝒘 ϵ 𝑃0} = 𝒙 + 𝒘 𝒘 ϵ span{𝒖, 𝒗}} = 𝒙 + 𝑠𝒖 + 𝑡𝒗 𝑠, 𝑡 ϵ ℝ}.