RRef Stanford
Uploaded by
mandcrutRRef Stanford
Uploaded by
mandcrut1
Reduced Row Echelon Matrices
Denition A matrix is in Reduced Row Echelon Form if: . (* stands for any 1) All its rows look like: 0 0 1 real number). In other words, if reading from left to right, the rst nonzero entry is a 1. We call the leading 1 a pivot 2) # of leading zeros in each row strictly increases from top to bottom. 3) The components of a column containing a pivot are zero (except for the pivot itself). We refer to such a column as a pivot column. If a column isnt a pivot column, then we call it a free column. For example: 0 1 2 0 4 0 0 0 1 2 0 0 0 0 0 is row reduced. 1 0 0 1 0 0 3 0 0 3 0 0 0 1 0 0 1 0 0 0 1 8 0 1 4 2 4 4 2 4
is too.
is NOT because the fourth column violates rule 3). 1 3 0 0 4 1 0 1 0 2 0 0 0 1 4 is NOT because the size of the second row is row, violating rule 2). 1 3 0 0 0 0 9 0 0 0 0 1 not strictly bigger than the rst 4 2 4
is NOT because the second row violates rule 1).
Why row-reduced matrices are our friends
Suppose we have an augmented matrix, [A| b ], and luckily for us A is rowreduced. Then there is an easy procedure to nd a parametrization for the solution space of the equation A = b . Here is how the procedure works: x Example 1 3 0 1 2 0 4 0 0 0 1 2 0 0 0 0 0 0 0 1
0) The non-augmented part of the matrix has 5 columns, so our solutions are x1 . of the form . . . x5 1) First we locate the pivots. There are only two of them here, the 1s in the 2nd and 4th column. Because of this, well call x2 and x4 the pivot variables, and the rest, x1 , x3 , and x5 , will be called free variables. 2) Next we write out the real-variable equations associated to the rows containing the pivots. In this case that means the rst and second rows. We get x2 + 2x3 + x5 = 3 x4 + 2x5 = 0 (Notice that each equation contains exactly one pivot variable. The remaining are free. Furthermore, the index of the pivot variable is the lowest index appearing in the equation, and the coecient in front of the pivot variable is 1.) 3) Solve each of the above equations for its lone pivot variable by subtracting the free variables to the other side. x2 = 3 2x3 x5 x4 = 2x5 4) Subsitute these equations to nd a parametrization of the space of solutions: x1 x1 x2 3 2x3 x5 x3 = x3 x4 2x5 x5 x5 And split according to variables and constants to easily see the parametrization 0 1 0 0 0 0 0 3 = 0 + x1 0 + x3 2 + x5 0 0 1 2 0 0 1 0 0 Key Observations/Facts: 1- Notice that the solution space is parametrized by 3 vectors, and that there were 2 pivots. 3+2=5, which is the number of columns in the non-augmented portion of the matrix. We will see later that this is not a coincidence. 2- Sometimes there is no solution. This happens IF AND ONLY IF there is a row in [A| b ] that looks like [00 0|4(=0)]. Remember, each row of an 2
augmented matrix corresponds to a real-valued equation, and a row like the one above translates to 0 = 0x1 + + 0xn = 4 which clearly has no solution. 3- If the number of pivots is equal to the number of columns (since were assuming A is row-reduced, this actually means A is the identity matrix), there is exactly ONE solution, and it obviously b . 4- If the number of pivots is less than the number of columns, and there are no rows like [00 0|4(=0)], then there is an innite number of solutions. (Try to use observation 1 to explain why). In summary: It is easy to nd all solutions to the equation [A| b ] when A is in Reduced Row Echelon Form. ******************************************************* Now were NOT going to assume A is in reduced row echelon form. So suppose were given any old matrix A and asked to nd the solution space for [A| b ]. Well, we might try to see if we can convert the problem to a related problem, [RRef (A)| b ], but where now RRef (A) is row reduced. By related we mean that the solution space of [A| b ] equals the solutions space of [RRef (A)| b ]. Surprise, surprise, there is a way, and the process goes by the name rowreducing or putting into row-echelon form or whatever. Im getting a little sleepy to explain how to row-reduce, but hopefully youre well on your way to guring that out on your owns. Of course, let me if you need help. Let me take this opportunity though to point out that until you row-reduce A, it is nigh impossible to tell much about the number of solutions (ie. number of pivots) or if any solutions exist at all. Yes, if there is a row in [A| b ] that looks like [00 0|4(=0)], there are no solutions. BUT, it is possible that at rst none of the rows resemble [00 0|4(=0)], and yet there are still no solutions. As an easy example: 1 1 row reduces to 1 0 2 0 3 0 4 1 2 2 3 3 4 5
and now we see there are no solutions, because [00 0|4(=0)] IS row reduced, and we make observation no. 2.
