The determinant of a square matrix is a number that provides a lot of useful information about the matrix.
Its definition is unfortunately not very intuitive. It is derived from abstract principles, laid out with the aim of satisfying a certain mathematical need. Therefore, before giving a definition of determinant, we explain what the mathematical need is.
Consider the linear space of all real vectors. The space can be represented as a Cartesian plane. A vector can be represented as a point in the plane, whose - and -coordinate are the first and second entry of respectively.
Let be a set of vectors such that the corresponding points in the Cartesian plane form a region whose area can be measured and is equal to .
Now, take a matrix and define the set
In other words, we use the square matrix to linearly transform of all the vectors in the set . The resulting linear transformations are in the new set .
In the Cartesian space, the points of form a region whose area is equal to .
We ask the following question: can we derive the area from and ?
It turns out that there is a number, called the determinant of and denoted by , that satisfiesthat is, when we multiply the area of the original region by the determinant, we obtain the area of the transformed region.
Importantly, the number depends only on and not on and its area. The linear transformation defined by transforms any region in a new region whose area is times the area of the original region.
When we add one dimension and consider the space of real vectors, we make obvious changes: we are no longer in the Cartesian plane, but in three-dimensional space; the matrix used to perform linear transformations is a matrix; the determinant is the scaling factor that allows to compute the volumes of the linearly transformed regions.
In -dimensional spaces (for ), there are generalizations of the concept of volume, and linear transformations are defined by matrices, but the determinant keeps playing the same role: it acts as a scaling factor for the volume.
We have discussed volumes and how they are scaled by linear transformations, but we have omitted an important detail: when discussing determinants, we deal with signed volumes, that is, volumes that can be either positive or negative, depending on the orientation in space of the region whose volume is being measured. For this reason, also determinants can be positive or negative, depending on whether the linear transformation preserves or reverses the orientation of shapes.
As it can be difficult to get an intuitive grasp of the concept of signed volume, we are going to illustrate it with the following plots.
The Cartesian plane in the upper left corner represents the original linear space. A set of points is depicted in blue (the circular arrow). The other five planes represent sets of points that have been obtained by applying different linear transformations to with different matrices :
rotation does not affect the area of the transformed region; the determinant of the matrix used to rotate the shape is ;
resizing obviously affects the area of the region; in the figure, the determinant of the matrix used to shrink the shape is less than ;
shearing is a deformation that can affect the area of the deformed region; in the case displayed here, it shrinks the area; its determinant is less than ;
reflection, like in a mirror, does not change the size of the circular arrow, but it changes its orientation; after being reflected, the direction of the arrow is clockwise, unlike the direction of the original, which is counterclockwise; the determinant of is , because the transformation changes the orientation of the shape but not its magnitude;
projection of the shape on a 1-dimensional subspace transforms it in a segment whose area is ; the determinant of the matrix used to perform the projection is .
Many more linear transformations can be performed, for example, by combining the elementary transformations shown in the plot above. However, the principle remains the same: if the linear transformation associated to the matrix does not change the orientation of the circular arrow, then ; if it changes the orientation, then if the arrow is flattened (it loses one dimension), then .
In the next section we are going to provide a definition of determinant that actually provides a way of calculating the determinant from the elements of . That definition has been derived by mathematicians who took the following steps:
they found some simple properties that are satisfied by the volumes of linearly transformed regions (e.g., if a linear transformation doubles the upper and lower edge of a rectangle and leaves its left and right edge unchanged, then the area of the rectangle doubles; as a consequence, the scaling factor of the linear transformation must be );
they imposed these properties as axioms that the determinant should satisfy;
they proved that a number satisfying the axioms exists and is unique;
they found a formula for calculating it, that can be used as its definition.
For more details about this axiomatic approach, you can refer to the beautiful treatment by Schneider and Barker (1989).
We are now ready to provide a formal definition of determinant.
Definition Let be a matrix. Let be the set of all possible permutations of the first natural numbers The determinant of , denoted by or by , is
A permutation is an ordering of . The elements of the permutation are denoted by . The number is either or depending on the parity of the permutation (even or odd).
The product is over entries of the matrix . For each row , we choose the entry located in column . Note that there is exactly one chosen entry in each column and row.
The sum is over the set of all possible permutations .
Let us apply the definition to the case of a matrix .
There are two possible permutations of the set of the first two natural numbers:
There are no inversions in , so its parity is even and
There is one inversion in , so its parity is odd and
Having established these facts, we can compute the determinant of :
Example Define the matrixIts determinant is
Let us now tackle the case of a matrix .
There are six possible permutations of the set of the first three natural numbers. We report them below, together with the number of inversions and their sign (deriving them is left as an exercise - revise the lecture on the sign of a permutation if you find any difficulties):
Below you can find some exercises with explained solutions.
Find the determinant of the matrix
We can apply the formula for the determinant of a matrix found above
The matrix used to perform the reflection in the lower left panel of the figure above is Prove that .
Schneider, H., and Barker, G. P. (1989) Matrices and linear algebra, Dover Publications.
Most of the learning materials found on this website are now available in a traditional textbook format.