Chapter 2 Introduction to matrices Auteur : Loïc AUDREN Formation : EENG UE : Refresher Version : 1.0 Date : 05/09/2018 1 Pôle Numérique 1. Structure and operations Definitions A matrix with n rows and p colums in (where is or or ) is a rectangular table having np elements of . We denote this matrix like that : is the term of the matrix M at the 𝑡ℎ row and at the 𝑡ℎ column. The set of the matrices with n rows and p columns with corefficients in The expression × is called the dimension of the matrix 2 Pôle Numérique , is noted , . . 23/09/2019 Mini Quizz 1 Dimension of : 3 Pôle Numérique 2x3 3x2 6 6 -6 12 23/09/2019 Definitions If , the matrix is a square matrix (an order n matrix) 2 x 2 matrix = order 2 matrix it belongs to the set 3x 3matrix = order 3 matrix If n = 1, the matrix is a row matrix. If p = 1, the matrix is a column matrix. 4 Pôle Numérique 23/09/2019 Definitions A diagonal matrix is a square matrix where any coefficients is a diagonal matrix An upper triangular matrix (resp. lower) is a square matrix where all the elements below (resp. above) the diagonal are null : (resp. . Upper triangular matrix Lower triangular matrix 5 Pôle Numérique 23/09/2019 Definitions The null matrix , is the matrix where all elements are null : , for all and . , The identity matrix is the square matrix where all elements are null : but the diagonal is made of 1 : identity matrix 6 Pôle Numérique 23/09/2019 Mini Quizz 2 Is it an upper or lower triangular matrix ? Upper Lower None Is the null matrix always a square matrix? Yes No Is the identity matrix always a square matrix? Yes No Is the diagonal matrix always a square matrix? Yes No 7 Pôle Numérique 23/09/2019 Basic operations addition of two matrices : if then , and , multiplication by a scalar : if then , and , 8 Pôle Numérique 23/09/2019 Basic operations The addition of two matrices is commutative : The null matrix is neutral for the matrix addition : The opposite matrix of The multiplication by a scalar is distributive with respect to the matrix addition : is denoted and is such that : 9 Pôle Numérique , , 23/09/2019 Matrix multiplication if , and we have with and then the product A.B is well defined , , 10 Pôle Numérique 23/09/2019 Mini Quizz 3 Work out the following products 11 Pôle Numérique 23/09/2019 Transposition The matrix multiplication is NOT commutative : The identity matrix The matrix multiplication is associative : The matrix multiplication is distributive with respect to addition : multiplication to the left multiplication to the right for , (in general !). is neutral for the matrix multiplication : , Dividing by a matrix does not make sense , but we can multiply by its inverse… 12 Pôle Numérique 23/09/2019 Transposition For with , , we denote the transpose of , In other words it corresponds to : swapping rows and columns of A. Elements of the transpose matrix have in the original matrix, a symmetric position with respect to the diagonal. 13 Pôle Numérique 23/09/2019 For The transpose of an upper triangular matrix is a lower triangular matrix (and inversely) The transpose of a diagonal matrix is itself ! 14 Pôle Numérique 23/09/2019 2. Square matrices and inversion Definitions and properties of square matrices For is symmetric if is antisymmetric if is nilpotent if there exists If If and are diagonal matrices, then AB is a diagonal matrix too. is a diagonal matrix and , then is still a diagonal matrix and The sum or product of two diagonal matrices is a diagonal matrix. The sum or product of two upper triangular matrices is an upper triangular matrix. The sum or product of two lower triangular matrices is a lower triangular matrix. 15 Pôle Numérique 23/09/2019 Definitions and properties of square matrices Theorem : (Binomial expansion for matrices) For Ex: if they commute, then if they don’t: 16 Pôle Numérique 23/09/2019 Definitions and properties of square matrices The trace of a matrix is the sum of its diagonal elements : (even if do not commute !) Ex: Express the traces of 17 Pôle Numérique 23/09/2019 Inverse of a matrix is invertible such that is then called the inverse of A and is written If is not invertible, then is said to be singular or degenerate. The inverse of a matrix is unique. does not exist. a diagonal matrix is invertible all its coefficients are non zero then and they are invertible then is invertible, then such that are invertible 18 Pôle Numérique 23/09/2019 How to inverse a matrix: the Gauss-Jordan elimination For we denote the ith-row of A. Then the elementary operations on the rows are the following : Swapping two rows Multiplying a row by a non-zero number : Linear combination of two rows : Theorem: Any invertible matrix can be transformed into by successive elementary operations. And performing exactely the same sequence of elementary operations on the identity matrix will lead to . Two matrices are row equivalent one may be obtained from the other one via elementary row operations. 19 Pôle Numérique 23/09/2019 How to inverse a matrix: the Gauss-Jordan elimination Step1: we swap with in order to have the coefficient Step2: using linear combinations , we cancel out all the coefficients . Step3: we repeat the these steps with the coefficients on the next columns. So using these elementary operations, we can put A under upper triangular form, and perform at the same time the exact same operations from the identity matrix. Step 4 : if the coefficients onto the diagonal are non zeros, then we use them (as in step 2 and 3), starting from the last row with , to cancel out coefficients above the diagonal. This step should turn the upper triangular matrix into a diagonal matrix. Step5 : we divide row by their corresponding diagonal coefficients : This step turns the diagonal matrix into the identity matrix. Step6: we repeat these steps (or are made in parallel) onto the identity matrix in order to get the inverse matrix . 20 Pôle Numérique 23/09/2019 Mini Quizz 4 Using the Gauss-Jordan elimination method, express the inverse of: 21 Pôle Numérique 23/09/2019 3.Cramer systems Definitions A matrix is in row echelon form if: • All non-zero rows are above any rows of all zeros • The leading coefficient (= pivot) of a non-zero row is always strictly to the right of the leading coefficient of the row above it. A matrix is in reduced row echelon form (also called row canonical form) if it satisfies the following conditions: • It is in row echelon form. • Every leading coefficient is 1 and is the only nonzero entry in its column. 22 Pôle Numérique 23/09/2019 Definitions An augmented matrix is obtained by appending the columns of two given matrices. The augmented matrix of A and B is denoted (A|B) . Ex: Augmented matrices are used in linear algebra to • represent systems of linear equations, • quickly perform and keep track of elementary row operations and transformations into equivalent systems. 23 Pôle Numérique 23/09/2019 Cramer system Using the Gauss-Jordan elimination algorithm, we can solve a linear system of equations of the kind: where being the unknown. , If the corresponding matrix is invertible, performing linear combinations on rows of the system using the Gauss-Jordan elimination algorithm, we can get the system in a row echelon form which in that case would be an upper triangular matrix. This provides one of the unknowns (the last one), and substituting it in the previous row, you obtain an other unknown. Cascading up to the top will provide all the unknowns. In this case, we say that it is a Cramer sytem. 24 Pôle Numérique 23/09/2019 Cramer system Ex: Let’s have the linear system of simultaneous equations with Using the Gauss-Jordan elimination algorithm is an upper triangular matrix / in row echelon form So the initial system is a Cramer system. 25 Pôle Numérique 23/09/2019 General case If the corresponding matrix is not invertible (which will be the case if you have more unknowns than equations for instance), there is no single solution: - either there is no solution at all ( the system is said to be inconsistent/incompatible), - either there is an infinity of solutions (where some unknowns depend on others). In that second case, we can express the solutions using the reduced row echelon form. In the case of an incompatible system, performing the Gaussian elimination on the augmented matrix, leads to a last column with non zero pivot(s). 26 Pôle Numérique 23/09/2019 General case The number of nonzero rows in the row echelon matrix R is called the rank of R and also the rank of A. Case 1: Unique solution (Cramer system) If r=m=n then the system is consistent and there is exactly one solution, which can be found by back substitution. Case 2: No solution If r < m (meaning that R actually has at least one row of all 0s) and at least one of the numbers is not zero, then the system RX=F is inconsistent (so is AX=B) Gauss-Jordan elimination r 0 0 m 27 Pôle Numérique Case 3: Infinitely many solutions If r < m and all of the numbers are null, then the system is consistent and we obtain any of these solutions by choosing values of arbitrarily. Then we solve the rth equation for (in terms of those arbitrary values), then the (r-1)st equation for , and so on up the line. 23/09/2019 General case Case 1: unique solution with Using the Gauss-Jordan elimination algorithm is an upper triangular matrix / in row echelon form So there is a unique solution. 28 Pôle Numérique 23/09/2019 General case Case 2: no solution with Using the Gauss-Jordan elimination algorithm : impossible ! So there is no solution. 29 Pôle Numérique 23/09/2019 General case Case 3: infinitely many solutions with Using the Gauss-Jordan elimination algorithm : true whatever the values of . So we fix one of them and express the others with respect to that one. So there is an infinite number of solutions. 30 Pôle Numérique 23/09/2019 4.Determinant Definition and properties The determinant of a square matrix from calulcations on the coefficients. This determinant is denoted Theorem: is a real number obtained is invertible (if is invertible) 31 Pôle Numérique 23/09/2019 Expression of the determinant For a 2x2 matrix For a 3x3 matrix Sarrus’rule 32 Pôle Numérique 23/09/2019 Mini Quizz 5 Work out the determinant of the following matrices using the Sarrus rule: 33 Pôle Numérique 23/09/2019 Definition of the determinant of a matrix Effects of elementary operations (expressed on rows, similar on columns) on the determinant: • (swapping two rows) changes the sign of the determinant • (multiplying a row by a scalar number) multiplies the determinant by the same number • (combining two rows) doesn’t change the determinant The determinant of a square matrix choosing the row i to develop, we have is given by the following formula: where is the determinant of the square matrix obtained from column j . by ignoring row i and In a similar way we can express the determinant by developing with respect to a column j: 34 Pôle Numérique 23/09/2019 Definition of the determinant of a matrix Ex: Find the following determinants: (Sarrus’rule) OR 35 Pôle Numérique 23/09/2019 Mini Quizz 6 Work out the determinant of the following matrices using the last method: 36 Pôle Numérique 23/09/2019 Calculation of the determinant using a faster method There exists an other method to work out the determinant. It is based on the transformation of the matrix in its triangular form using the elementary operations: • (swapping two rows) changes the sign of the determinant • (multiplying a row by a scalar number) multiplies the determinant by the same number • (combining two rows) doesn’t change the determinant So we apply the Gauss-Jordan elimination method, in order to turn the matrix triangular. Once the matrix is triangular, the determinant becomes very easy to express: So with 37 Pôle Numérique number of row switches 23/09/2019 Calculation of the determinant using a faster method Ex: Find the following determinants: 38 Pôle Numérique 23/09/2019 Mini Quizz 7 Work out the determinant of the following matrices using the fast method: 39 Pôle Numérique 23/09/2019 Matrix inversion using the co-matrix and the determinant If is invertible, its inverse with and is given by being the determinant of the matrix ignoring the row i and the column j. is called the co-matrix of , and are called the co-factors. is also called the adjoint (or adjugate) matrix, and is written 40 Pôle Numérique . 23/09/2019 Matrix inversion using the co-matrix and the determinant For 2x2 matrix, For 3x3 matrix, 41 Pôle Numérique 23/09/2019 Mini Quizz 8 Using the determinant and the adjoint matrix, express the inverse of: 42 Pôle Numérique 23/09/2019 ! BONUS ! The Cramer’s rule is invertible Then when solving the equation the system has a unique solution with The solutions are with being the square matrix formed by substituting the k-column of 43 Pôle Numérique by . 23/09/2019 The Cramer’s rule Ex1: Use Cramer’s rule to solve the following: 44 Pôle Numérique 23/09/2019 The Cramer’s rule Ex2: Use Cramer’s rule to solve the following: ……. 45 Pôle Numérique 23/09/2019 Thanks for your attention Loïc AUDREN [email protected] 46 Pôle Numérique