If k is neither -5 nor 2, there's a single solution. In general, a system of linear equations has a single solution for some (and, equivalently, for any) right-hand side iff [=if and only if] its left-hand-side matrix is square and
1) its
rank equals its number of rows/columns, or, equivalently,
2) its
determinant is nonzero, or, equivalently,
3) it can be transformed into a unit matrix via the Gauss (row reduction) method.
That holds for any number of dimensions of the matrix.
It's worth noting that (computer science rant alert!):
1. In computers, determinants of big square matrices are almost always calculated via the Gauss method (because the determinant of an
upper triangular square matrix is the product of its diagonal elements) as the Gauss method is cubic in time (takes time roughly proportional to the cubed number of rows), while the method of calculating the determinant outlined in its definition is extremely slow as the number of rows grows.
2. Not only is swapping rows in the Gauss method a convenient option, but any computer scientist will tell you that, when using it to solve a linear system on a computer, it's
imperative to swap rows so that the principal row (the one which you're subtracting from the other with coefficients) has the number with the biggest available absolute value in the corresponding principal column, for the biggest numerical stability. That's because division by a number with a smaller abs. value introduces a bigger abs. computational error.
2a. It's of course not an issue when you do hw with 2x2-4x4 matrices on paper, swap rows as you wish then; in fact, on paper, it's easier to add or subtract the row with a 1 or -1 in the principal column - it saves the time and effort needed for division by coefficients.
Last edited by coon74; 09-14-2014 at 06:31 PM.