Welcome to Extra Pure Matrices!
In your Core Pure studies, you learned how matrices can transform space—flipping, rotating, and stretching shapes. In this Extra Pure chapter, we are going deep "under the hood." You will learn how to find the special directions where a transformation simply stretches space without turning it. These are called Eigenvectors. We will also discover clever shortcuts for calculating massive powers of matrices and explore a beautiful result called the Cayley-Hamilton Theorem.
Don't worry if this seems a bit abstract at first! We’ll use plenty of analogies to keep things grounded. Think of this chapter as learning the "DNA" of a matrix.
1. Eigenvalues and Eigenvectors
When you multiply a vector by a matrix, the vector usually changes both its length and its direction. However, for most matrices, there are special vectors that only change their length. Their direction stays exactly the same (or reverses).
What are they?
- Eigenvector: A non-zero vector \(\mathbf{v}\) that, when multiplied by a matrix \(\mathbf{M}\), results in a multiple of itself.
- Eigenvalue: The scale factor \(\lambda\) (lambda) by which the eigenvector is stretched.
The fundamental equation is: \( \mathbf{M}\mathbf{v} = \lambda\mathbf{v} \)
The Characteristic Equation
To find these special values, we use the Characteristic Equation:
\(\det(\mathbf{M} - \lambda\mathbf{I}) = 0\)
Where \(\mathbf{I}\) is the Identity Matrix. Solving this equation gives you a polynomial in \(\lambda\). For a \(2 \times 2\) matrix, it’s a quadratic; for a \(3 \times 3\), it’s a cubic.
Step-by-Step Process:
- Subtract \(\lambda\) from the main diagonal elements of matrix \(\mathbf{M}\).
- Find the determinant of this new matrix and set it to zero.
- Solve for \(\lambda\) to find the eigenvalues.
- For each eigenvalue, substitute it back into \((\mathbf{M} - \lambda\mathbf{I})\mathbf{v} = \mathbf{0}\) to find the eigenvectors.
Quick Review: An eigenvector is like a "fixed track" that the transformation slides things along. The eigenvalue tells you if you are stretching (\(\lambda > 1\)), squashing (\(0 < \lambda < 1\)), or reflecting (\(\lambda\) is negative).
2. Diagonalisation
Diagonal matrices (where all numbers are zero except on the leading diagonal) are the "easiest" matrices to work with. Diagonalisation is the process of turning a complicated matrix \(\mathbf{M}\) into a simple diagonal matrix \(\mathbf{D}\).
How it works
If a matrix \(\mathbf{M}\) has distinct real eigenvalues, we can write it in the form:
\( \mathbf{M} = \mathbf{P}\mathbf{D}\mathbf{P}^{-1} \)
- \(\mathbf{D}\) is the diagonal matrix containing the eigenvalues on the main diagonal.
- \(\mathbf{P}\) is the modal matrix, where each column is an eigenvector corresponding to the eigenvalues in \(\mathbf{D}\).
Analogy: Imagine you have a set of instructions written in a difficult language (\(\mathbf{M}\)). Diagonalisation is like translating those instructions into a very simple language (\(\mathbf{D}\)), doing the work, and then translating back (\(\mathbf{P}\) and \(\mathbf{P}^{-1}\)).
Key Takeaway: To diagonalise, you just need the eigenvalues and their corresponding eigenvectors. Make sure the order of eigenvectors in \(\mathbf{P}\) matches the order of eigenvalues in \(\mathbf{D}\)!
3. Powers of Matrices
Calculating \(\mathbf{M}^{10}\) by hand would be a nightmare. Diagonalisation gives us a massive shortcut.
Because \( \mathbf{M} = \mathbf{P}\mathbf{D}\mathbf{P}^{-1} \), when we multiply \(\mathbf{M}\) by itself, the "inner" \(\mathbf{P}^{-1}\) and \(\mathbf{P}\) cancel out out (since \(\mathbf{P}^{-1}\mathbf{P} = \mathbf{I}\)).
This leads to the beautiful formula:
\( \mathbf{M}^n = \mathbf{P}\mathbf{D}^n\mathbf{P}^{-1} \)
Since \(\mathbf{D}\) is diagonal, \(\mathbf{D}^n\) is just a matrix where you raise each individual eigenvalue to the power of \(n\). Easy!
Common Mistake: Don't forget that matrix multiplication is not commutative. You must keep the order as \(\mathbf{P}\), then \(\mathbf{D}^n\), then \(\mathbf{P}^{-1}\). Swapping them will give the wrong answer!
4. The Cayley-Hamilton Theorem
This is one of the most surprising results in matrix algebra. It states that every square matrix satisfies its own characteristic equation.
If your characteristic equation is \(\lambda^2 - 5\lambda + 6 = 0\), then the matrix \(\mathbf{M}\) will satisfy:
\( \mathbf{M}^2 - 5\mathbf{M} + 6\mathbf{I} = \mathbf{0} \)
Why is this useful?
- Finding Inverses: You can rearrange the equation to find \(\mathbf{M}^{-1}\) without using the traditional determinant/adjugate method. Simply multiply the whole equation by \(\mathbf{M}^{-1}\).
- High Powers: You can express high powers like \(\mathbf{M}^3\) in terms of lower powers like \(\mathbf{M}\) and \(\mathbf{I}\).
Did you know? Even though it looks like you are just "replacing \(\lambda\) with \(\mathbf{M}\)," you must remember to turn the constant term (like the \(6\) above) into a matrix by multiplying it by the identity matrix \(\mathbf{I}\).
5. Geometric Significance
Eigenvalues and eigenvectors tell us a lot about 2-D and 3-D transformations.
In 2-D Transformations:
- An eigenvector represents an invariant line through the origin. Any point on this line stays on this line after the transformation.
- If \(\lambda = 1\), every point on that line is an invariant point (it doesn't move at all).
In 3-D Transformations:
- If a \(3 \times 3\) matrix represents a reflection in a plane, the plane itself is made of eigenvectors with \(\lambda = 1\). The vector normal (perpendicular) to the plane is an eigenvector with eigenvalue \(\lambda = -1\).
- If a matrix represents a rotation about an axis, the axis of rotation is an eigenvector with \(\lambda = 1\) (because points on the axis don't move).
Summary Tip: When asked to find the "invariant line" or "axis of rotation," you are really being asked to find the eigenvectors!
Quick Review Box
1. Characteristic Equation: \(\det(\mathbf{M} - \lambda\mathbf{I}) = 0\)
2. Eigenvector Condition: \(\mathbf{M}\mathbf{v} = \lambda\mathbf{v}\)
3. Diagonalisation: \(\mathbf{M} = \mathbf{P}\mathbf{D}\mathbf{P}^{-1}\)
4. Matrix Powers: \(\mathbf{M}^n = \mathbf{P}\mathbf{D}^n\mathbf{P}^{-1}\)
5. Cayley-Hamilton: A matrix satisfies its own polynomial equation.