Matrices (Further Pure Mathematics 2: Topic 2.2)

Welcome to the Matrices chapter of Further Pure Mathematics 2! If you tackled matrices in FP1, you learned about basic operations and 2D transformations. Now, we dive deeper into the power of matrices, specifically focusing on solving complex 3D systems and introducing the game-changing concepts of eigenvalues and eigenvectors.

Why is this important? Matrices are the language of modern data science, physics (quantum mechanics!), and engineering. Understanding eigenvalues, for instance, helps describe natural frequencies, system stability, and principal axes—essentially, how complex systems behave under stress or change. Let's conquer this powerful topic!

1. Solving Systems of Linear Equations (3 Unknowns)

In FP2, you will deal with solving three simultaneous linear equations in three variables, usually represented by planes in 3D space. We use matrices to formulate and solve these problems efficiently.

Formulating the Matrix Equation

Consider the system:

\(a_{11}x + a_{12}y + a_{13}z = b_1\)
\(a_{21}x + a_{22}y + a_{23}z = b_2\)
\(a_{31}x + a_{32}y + a_{33}z = b_3\)

This system can be written simply as:

\(A\mathbf{x} = \mathbf{b}\)

Where \(A\) is the coefficient matrix, \(\mathbf{x}\) is the variable vector, and \(\mathbf{b}\) is the constant vector:

\( A = \begin{pmatrix} a_{11} & a_{12} & a_{13} \\ a_{21} & a_{22} & a_{23} \\ a_{31} & a_{32} & a_{33} \end{pmatrix}, \quad \mathbf{x} = \begin{pmatrix} x \\ y \\ z \end{pmatrix}, \quad \mathbf{b} = \begin{pmatrix} b_1 \\ b_2 \\ b_3 \end{pmatrix} \)

Consistency and Geometric Interpretation

The number of solutions depends entirely on the determinant of \(A\), denoted \(\det A\), and whether the system is consistent (has solutions) or inconsistent (has no solutions).

Case 1: Unique Solution (The "Nice" Case)

If the coefficient matrix \(A\) is non-singular:

  • Condition: \(\det A \neq 0\).
  • Solution: There is a unique solution, given by \(\mathbf{x} = A^{-1}\mathbf{b}\).
  • Geometric Interpretation: The three planes meet at a single, common point.
Case 2: Singular Matrix (\(\det A = 0\))

If \(A\) is singular, there are either infinite solutions or no solutions. We must now check the consistency of the system (i.e., whether \(\mathbf{b}\) is related correctly to \(A\)). This is often done by performing row reduction (Gaussian elimination) or using vector methods to see if the equations are dependent.

A) Infinite Solutions (Consistent System)

If \(\det A = 0\) and the system is consistent:

  • Geometric Interpretation:
    1. The three planes intersect in a common line. (Most common scenario for infinite solutions).
    2. The three planes are identical (or two are identical and the third passes through that plane).
  • Solving: When solving a consistent singular system, one variable (or sometimes two) will be arbitrary, leading to solutions expressed in terms of a parameter (e.g., \(x=k, y=2k+1\)).
B) No Solution (Inconsistent System)

If \(\det A = 0\) and the system is inconsistent:

  • Geometric Interpretation: The planes have no common intersection point or line. They often form a triangular prism (two planes parallel, or three planes intersecting pairwise but not all three at once), or they are three parallel distinct planes.
  • Detection: During row reduction, you will arrive at a contradiction, such as \(0 = k\) where \(k\) is a non-zero number.

Key Takeaway for Linear Systems: The determinant tells you if a unique solution exists (\(\det A \neq 0\)). If \(\det A = 0\), you must investigate consistency to determine if there are zero or infinite solutions, and interpret the result geometrically (point, line, or no common intersection).

2. Eigenvalues and Eigenvectors

Eigen-theory is central to advanced matrix work. It helps us find special vectors that are only scaled, not rotated, by a matrix transformation.

Analogy: Imagine a transformer machine (the matrix \(A\)). It usually squishes, rotates, and stretches any vector (\(\mathbf{x}\)) you put in. An eigenvector is like a perfectly shaped stick that, when put into the machine, comes out pointing in exactly the same direction—it was only stretched or shrunk. The eigenvalue is the scale factor of that stretch or shrink.

The Eigenvalue Equation

For a square matrix \(A\), a scalar \(\lambda\) (the eigenvalue) and a non-zero vector \(\mathbf{e}\) (the eigenvector) satisfy the equation:

\(A\mathbf{e} = \lambda\mathbf{e}\)

To find the non-zero eigenvectors, we rearrange the equation:

\(A\mathbf{e} - \lambda\mathbf{e} = \mathbf{0}\)
\((A - \lambda I)\mathbf{e} = \mathbf{0}\)

Where \(I\) is the identity matrix.

Step-by-Step: Finding Eigenvalues (\(\lambda\))

For a non-trivial solution (i.e., \(\mathbf{e} \neq \mathbf{0}\)) to exist, the matrix \((A - \lambda I)\) must be singular. Therefore, its determinant must be zero.

1. Form the Characteristic Equation:

\(\det(A - \lambda I) = 0\)

2. Solve for \(\lambda\): Solving this polynomial equation gives the eigenvalues.

Did you know? If \(A\) is a 2x2 matrix, the characteristic equation is quadratic. If \(A\) is 3x3, it is cubic. Don't worry if this seems tricky at first; solving cubic equations derived from matrices often results in integer roots that are easy to spot or factorise!

Step-by-Step: Finding Eigenvectors (\(\mathbf{e}\))

1. Substitute \(\lambda\) back: For each eigenvalue \(\lambda\) found, substitute it back into the equation \((A - \lambda I)\mathbf{e} = \mathbf{0}\).

2. Solve the system: Since \(\det(A - \lambda I) = 0\), this system is singular and will have infinite solutions. By solving the resulting simultaneous equations, you find the unique direction of the eigenvector \(\mathbf{e}\). (The components of \(\mathbf{e}\) will be given relative to each other, often involving an arbitrary constant, \(k\)).

Common Mistake to Avoid: An eigenvector is defined by its direction only. When you solve for \(\mathbf{e}\), you must express the vector as simply as possible (often setting one component to 1 or solving for all components in terms of \(k\)).

3. Diagonalisation of a Matrix

Diagonalisation is a technique that simplifies matrices, allowing us to perform operations like calculating high powers easily.

Syllabus Restriction: Diagonalisation via this method is restricted to cases where the eigenvalues are real and distinct.

The Diagonal Form (\(D\)) and Diagonalisation Formula

If a matrix \(A\) is diagonalisable, it can be written as:

\(A = Q D Q^{-1}\)

Where:

  • \(D\) (The Diagonal Matrix): Contains the eigenvalues \(\lambda_i\) of \(A\) along its main diagonal. The off-diagonal entries are zero.

    \(D = \begin{pmatrix} \lambda_1 & 0 & 0 \\ 0 & \lambda_2 & 0 \\ 0 & 0 & \lambda_3 \end{pmatrix}\)

  • \(Q\) (The Modal Matrix): Is formed by placing the corresponding eigenvectors \(\mathbf{e}_i\) as its columns. Crucially, the order of eigenvectors in \(Q\) must match the order of eigenvalues in \(D\).

    \(Q = \begin{pmatrix} \mathbf{e}_1 & \mathbf{e}_2 & \mathbf{e}_3 \end{pmatrix}\)

Using Diagonalisation to Calculate Powers

Calculating high powers of a standard 3x3 matrix (e.g., \(A^{100}\)) is tedious. Using the diagonal form, it becomes simple:

\(A^n = (Q D Q^{-1})(Q D Q^{-1}) \dots (Q D Q^{-1})\)

Since the middle terms cancel (\(Q^{-1}Q = I\)), this simplifies drastically:

\(A^n = Q D^n Q^{-1}\)

The calculation of \(D^n\) is trivial:

\( D^n = \begin{pmatrix} \lambda_1^n & 0 & 0 \\ 0 & \lambda_2^n & 0 \\ 0 & 0 & \lambda_3^n \end{pmatrix} \)

Step-by-Step for \(A^n\):

  1. Find eigenvalues (\(\lambda_i\)).
  2. Find corresponding eigenvectors (\(\mathbf{e}_i\)).
  3. Construct \(D\) and \(Q\).
  4. Calculate \(Q^{-1}\) (using techniques learned in FP1, extended to 3x3).
  5. Calculate \(D^n\).
  6. Compute \(A^n = Q D^n Q^{-1}\).

Key Takeaway for Diagonalisation: Diagonalisation is an essential tool for computing high powers of a matrix, turning a marathon calculation into a straightforward matrix multiplication sequence using the easy-to-power diagonal matrix \(D\).

4. The Cayley-Hamilton Theorem

This theorem provides a powerful shortcut, especially for finding the inverse or small powers of a matrix, without needing to calculate all eigenvectors.

The Principle

The theorem states that every square matrix satisfies its own characteristic equation.

If the characteristic equation is \(P(\lambda) = \det(A - \lambda I) = 0\), then replacing \(\lambda\) with the matrix \(A\) (and the constant term with the constant times the Identity matrix \(I\)) results in the zero matrix:

\(P(A) = \mathbf{0}\)

Example: If the characteristic equation for a 2x2 matrix is \(\lambda^2 - 4\lambda + 3 = 0\), then the theorem states:

\(A^2 - 4A + 3I = \mathbf{0}\)

Application 1: Finding Powers of \(A\)

We can use the characteristic equation to express higher powers of \(A\) in terms of lower powers.

From the example above, we can rearrange to find \(A^2\):

\(A^2 = 4A - 3I\)

To find \(A^3\), multiply by \(A\):

\(A^3 = 4A^2 - 3IA = 4A^2 - 3A\)

Then substitute the expression for \(A^2\) back in:

\(A^3 = 4(4A - 3I) - 3A = 16A - 12I - 3A = 13A - 12I\)

This method works rapidly for finding successive powers of \(A\).

Application 2: Finding the Inverse \(A^{-1}\)

We can use the theorem to derive an expression for the inverse \(A^{-1}\) without using the traditional adjoint method (especially useful for 3x3 matrices).

Start with the characteristic equation expression and multiply by \(A^{-1}\):

\(A^2 - 4A + 3I = \mathbf{0}\)
(Multiply by \(A^{-1}\))
\(A^{-1}(A^2) - 4A^{-1}(A) + 3A^{-1}(I) = A^{-1}(\mathbf{0})\)
\(A - 4I + 3A^{-1} = \mathbf{0}\)

Rearrange to isolate \(A^{-1}\):

\(3A^{-1} = 4I - A\)
\(A^{-1} = \frac{1}{3}(4I - A)\)

This allows you to calculate the inverse using only simple scalar multiplication and matrix subtraction.

Quick Review Box: FP2 Matrix Essentials

  • Linear Systems: If \(\det A \neq 0\), unique solution. If \(\det A = 0\), check consistency (zero or infinite solutions, interpreted geometrically).
  • Eigen-Theory: Find \(\lambda\) using \(\det(A - \lambda I) = 0\). Find \(\mathbf{e}\) using \((A - \lambda I)\mathbf{e} = \mathbf{0}\).
  • Diagonalisation: \(A^n = Q D^n Q^{-1}\) (only for real, distinct eigenvalues).
  • Cayley-Hamilton: \(P(A) = \mathbf{0}\) (Use this for finding \(A^n\) or \(A^{-1}\) through substitution).