The determinant is a single number computed from a square matrix that reveals essential information about the linear transformation it represents: whether the transformation can be undone (invertibility), and by what factor it scales area, volume, or higher-dimensional volume. A determinant of zero signals that the transformation collapses space into a lower dimension, losing information irreversibly.
Imagine a matrix as a machine that stretches, squishes, and rotates space. The determinant is a single number that tells you exactly how much that machine scales area (in 2D) or volume (in 3D). If the determinant is 2, everything doubles in area. If it's 0.5, everything shrinks to half its area. And if the determinant is exactly zero, the machine squashes space flat — turning a 2D square into a 1D line, or a 3D cube into a flat 2D sheet — permanently losing information in the process.
For the matrix [[a,b],[c,d]], the determinant is simply ad − bc. This tiny formula tells you the exact area-scaling factor of that transformation, including whether it also flips orientation (a negative determinant means the transformation includes a mirror-image flip).
For 3×3 matrices, the formula (via "cofactor expansion" along the first row) is:
for matrix [[a,b,c],[d,e,f],[g,h,i]]. Larger matrices use a recursive generalisation of this same cofactor expansion process.
The determinant of a matrix equals the product of all its eigenvalues (counted with multiplicity). This is why a zero eigenvalue always forces a zero determinant — a zero eigenvalue means some direction in space gets collapsed entirely to a point, exactly the "squashing" behaviour a zero determinant signals.
For a system of linear equations Ax = b with A invertible, each individual unknown xᵢ can be computed as a ratio of two determinants: xᵢ = det(Aᵢ)/det(A), where Aᵢ is A with its i-th column replaced by b. Though elegant, this method (Cramer, 1750) is computationally inefficient for large systems compared to Gaussian elimination, and is mainly of theoretical and pedagogical rather than practical computational interest today.
Remarkably, determinants were studied and used well before the modern matrix concept was formalised. Seki Takakazu in Japan (1683) and Gottfried Leibniz in Europe (1693, in a letter to L'Hôpital) independently developed early determinant-like concepts for solving systems of linear equations, decades before Cayley's systematic matrix theory (1850s). The term "determinant" itself was coined by Gauss (1801) in a related but distinct context (studying quadratic forms), and later adopted in its modern sense by Cauchy (1812), who also proved the crucial multiplicative property det(AB)=det(A)det(B).
For a general n×n matrix, the determinant can be defined precisely via the Leibniz formula, summing over all n! possible permutations σ of {1,...,n}:
where sgn(σ) is +1 or −1 depending on whether the permutation σ is "even" or "odd" (determined by the minimum number of pairwise element swaps needed to produce it). This formula, while elegant and theoretically foundational, has n! terms and becomes computationally impractical for even moderately large matrices — which is why Gaussian-elimination-based methods are used in actual numerical computation.
More abstractly, the determinant can be characterised as the unique function (up to scalar multiple) that is linear in each row separately, changes sign when any two rows are swapped ("alternating"), and assigns the value 1 to the identity matrix. This characterisation reveals the determinant's deep connection to the theory of exterior algebras and differential forms in advanced differential geometry, where determinants appear naturally as the coefficients relating volume elements under a change of coordinates.
When changing variables in a multivariable integral (for instance, converting from Cartesian to polar coordinates), the determinant of the matrix of partial derivatives — the Jacobian determinant — appears as the correct scaling factor needed to properly account for how the coordinate transformation distorts area or volume. This is the essential rigorous justification behind substitution techniques in multivariable calculus, generalising the simpler single-variable u-substitution rule.