Module 1: Mathematical methods · Lesson 1.8 · about 45 min
Matrices, determinants and eigenvalues
Matrix algebra, determinants, inverse and rank, then eigenvalues and eigenvectors with the trace and determinant shortcuts, Hermitian, unitary and orthogonal matrices, Cayley-Hamilton and functions of matrices, the way JAM tests them.
Module 1 · about 45 minutes. By the end you will be able to find eigenvalues of small matrices in your head, use the trace and determinant to skip whole calculations, recognise Hermitian, unitary and orthogonal matrices and what their eigenvalues must be, and compute powers and exponentials of matrices the quick way.
1. Matrices as machines
A matrix takes a vector and returns another:
Multiplying two matrices means doing one machine after the other, which is why order matters: in general . The rules that follow from “one after the other”:
- and : undoing reverses the order, like taking off socks and shoes.
- , and for an matrix (every one of the rows gets the factor ).
- The trace (sum of the diagonal) obeys .
2. Determinant, inverse and rank
The determinant measures how much the machine scales areas (in 2D) or volumes (in 3D). If it is zero, the machine squashes space flat, and nothing can undo that: the matrix has no inverse (it is singular).
For the inverse: swap the diagonal, negate the off-diagonal, divide by the determinant. Do not swap the off-diagonal entries (that gives the transpose of the right answer, a favourite wrong option).
Useful facts for and bigger:
- A triangular matrix has determinant equal to the product of its diagonal.
- Swapping two rows flips the sign; adding a multiple of one row to another changes nothing.
- The rank is the number of independent rows. A square matrix has full rank exactly when its determinant is not zero. A system has a unique solution exactly when .
3. Eigenvalues and eigenvectors
Most vectors change direction when a matrix acts on them. A few special ones do not: they come out along the same line, just stretched by a number .
Watch it happen. The grey arrow is a unit vector going round the circle; the coloured arrow is . They line up only along the dashed eigenvector lines:
How to find them. means has a non-zero solution, so must be singular:
For a matrix this is always
and for any size:
These two lines are the biggest time-savers in the whole topic. “Sum of eigenvalues” is just the trace; “product” is just the determinant; “is 0 an eigenvalue?” is just “is the determinant 0?”.
- Eigenvector v
- a direction the matrix does not turn, only stretches (or flips)
- Eigenvalue λ
- the stretch factor along that direction; negative means flipped
- tr A
- the sum of all eigenvalues
- det A
- the product of all eigenvalues; zero means some direction is squashed to nothing
- Physical meaning
- Normal modes of coupled oscillators, principal axes of a spinning body, and the allowed energies of a quantum system are all eigenvalue problems.
Worked example. Find the eigenvalues and eigenvectors of .
- Trace 4, determinant 3, so : or .
- For : gives , so .
- For : , so .
- Check: (trace) and (determinant) ✓. The two eigenvectors are perpendicular, as they always are for a real symmetric matrix.
Two special cases to spot at once: a triangular matrix has its eigenvalues on the diagonal, and a block-diagonal matrix has the eigenvalues of its blocks.
Not every matrix has enough eigenvectors
A repeated eigenvalue may come with only one eigenvector. The shear has eigenvalue 1 twice, but has rank 1, so there is only one eigenvector, . Such a matrix cannot be diagonalised. Count eigenvectors for an eigenvalue as .
When there are independent eigenvectors, put them as the columns of ; then is diagonal with the eigenvalues on its diagonal. Matrices related this way (similar matrices) share eigenvalues, trace and determinant, but not eigenvectors.
4. The special matrices of physics
| Matrix | Definition | Eigenvalues | Eigenvectors |
|---|---|---|---|
| real symmetric | real | can be chosen orthonormal | |
| Hermitian | ( = conjugate transpose) | real | orthogonal for different |
| real antisymmetric | or purely imaginary | ||
| orthogonal | modulus 1 () | ||
| unitary | modulus 1 () | orthogonal | |
| projection | 0 or 1 |
Why Hermitian eigenvalues are real, in one line: if , then , so . That is why quantum mechanics represents measurable quantities by Hermitian matrices: measurements give real numbers.
A rotation by in the plane has eigenvalues : modulus 1, but complex, which is why the rotation in the figure has no real eigenvector. A rotation in 3D has eigenvalues (the 1 belongs to the axis), so its trace is .
The Pauli matrices are the standard Hermitian, unitary, traceless examples:
Each has eigenvalues , each squares to , and (cyclic). For a real vector , , so has eigenvalues .
“The eigenvalues of A + B are the eigenvalues of A plus those of B.” Only if and share eigenvectors (for example, if they commute). , and each have eigenvalues , but has eigenvalues , not . What always adds is the trace: .
5. Functions of a matrix and Cayley-Hamilton
If , then , , and for any polynomial or power series :
So the eigenvalues of are , with the same eigenvectors. This turns many “compute this big matrix” questions into arithmetic:
Worked example. A matrix has eigenvalues 1, 2, 3. Find and .
- Eigenvalues of : . Determinant: .
- Eigenvalues of : . Trace: .
Two famous consequences: for every square matrix, and when (the rotation matrix comes out of the exponential, just as ).
Cayley-Hamilton theorem: every matrix satisfies its own characteristic equation. For a matrix,
so any power of can be reduced to a combination of and . For (trace 5, det ): , and then .
Two more shortcuts for powers: a nilpotent part stops the series early ( because ), and a diagonalisable matrix gives .
Physics example: normal modes. Two equal masses between three equal springs obey . The eigenvectors and are the two normal modes (moving together, moving opposite); the eigenvalues 1 and 3 give and . See it move in fibub’s coupled oscillators tool, and check any matrix with the matrix calculator or the matrix transform tool.
- , , . A matrix is invertible exactly when its determinant is non-zero.
- Eigenvalues solve ; for , . Their sum is the trace and their product the determinant.
- Triangular: eigenvalues on the diagonal. A repeated eigenvalue may have only one eigenvector (then no diagonalisation).
- Hermitian and real symmetric: real eigenvalues, orthogonal eigenvectors. Unitary and orthogonal: eigenvalues of modulus 1. Antisymmetric: 0 or imaginary. Projection: 0 or 1.
- The eigenvalues of are ; . Cayley-Hamilton: for .
- Eigenvalues of a sum do not add (unless the matrices commute); traces always do.
Practise: exam questions on this lesson
They start easy and climb to top-rank level. Check each answer, then read the full solution. More on the matrices practice page.
Before this: lesson 1.7, differential equations. Next: lesson 1.9, complex numbers. Eigenvalues come back in coupled oscillations, rigid-body rotation and all of quantum mechanics. For GATE and CSIR NET, continue with Hermitian and unitary matrices and diagonalisation in the GATE Physics course.
Check yourself
Sources
- M. L. Boas, Mathematical Methods in the Physical Sciences, 3rd ed. (2006), chapter 3
- G. Strang, Introduction to Linear Algebra, 5th ed. (2016), chapters 5 and 6
- G. B. Arfken, H. J. Weber and F. E. Harris, Mathematical Methods for Physicists, 7th ed. (2013), chapter 6
- IIT JAM Physics syllabus (Mathematical Methods section), published by the organising IIT each year
Your ticks stay in this browser. Save them as a file on the course home.
Updated 2026-11-22. Not affiliated with IIT or the JAM organising institute. Exam pattern as published for recent years: always check the official JAM website before the exam. All questions are original.