Determinant Applications: Cramer's Rule, Inverse & More

Cramer's Rule

Ax = \mathbf{b}

where

A

is

n \times n

with

\det(A) \neq 0

, Cramer's rule expresses each component of the solution directly as a ratio of determinants:

Cramers Rule

x_i = \frac{\det(A_i)}{\det(A)}

Cramers rule →

where

A_i

is the matrix formed by replacing column

i

of

A

with the right-hand side vector

\mathbf{b}

. Every other column stays in place.

2×2 Example

For the system

Cramer's Rule

\begin{pmatrix} 3 & 2 \\ 1 & 5 \end{pmatrix} \begin{pmatrix} x_1 \\ x_2 \end{pmatrix} = \begin{pmatrix} 8 \\ 7 \end{pmatrix}

the coefficient determinant is

\det(A) = 3 \cdot 5 - 2 \cdot 1 = 13

. Replacing column

1

with

\mathbf{b}

:

Cramer's Rule

\det(A_1) = \det\begin{pmatrix} 8 & 2 \\ 7 & 5 \end{pmatrix} = 40 - 14 = 26

Replacing column

2

with

\mathbf{b}

:

Cramer's Rule

\det(A_2) = \det\begin{pmatrix} 3 & 8 \\ 1 & 7 \end{pmatrix} = 21 - 8 = 13

So

x_1 = 26/13 = 2

and

x_2 = 13/13 = 1

.

Pitfall: Replacing a row instead of a column

Putting

\mathbf{b}

into row

1

instead of column

1

gives

\det\begin{pmatrix} 8 & 7 \\ 1 & 5 \end{pmatrix} = 33

and

x_1 = 33/13

. With column

1

replaced,

\det(A_1) = 26

and

x_1 = 2

.

Both values can be checked in the original equations:

3 \cdot 2 + 2 \cdot 1 = 8

and

1 \cdot 2 + 5 \cdot 1 = 7

. The denominator

\det(A) = 13

is shared by both components, while each numerator comes from its own modified matrix, in which

\mathbf{b}

stands where the coefficients of that unknown stood.

3×3 Example

Cramer's Rule

\begin{pmatrix} 1 & 0 & 2 \\ -1 & 3 & 1 \\ 2 & 1 & 0 \end{pmatrix} \begin{pmatrix} x_1 \\ x_2 \\ x_3 \end{pmatrix} = \begin{pmatrix} 5 \\ 0 \\ 3 \end{pmatrix}

The coefficient determinant is

\det(A) = 1(0 - 1) - 0 + 2(-1 - 6) = -1 - 14 = -15

. The three modified determinants are:

Cramer's Rule

\det(A_1) = \det\begin{pmatrix} 5 & 0 & 2 \\ 0 & 3 & 1 \\ 3 & 1 & 0 \end{pmatrix} = 5(0 - 1) - 0 + 2(0 - 9) = -5 - 18 = -23

Cramer's Rule

\det(A_2) = \det\begin{pmatrix} 1 & 5 & 2 \\ -1 & 0 & 1 \\ 2 & 3 & 0 \end{pmatrix} = 1(0 - 3) - 5(0 - 2) + 2(-3 - 0) = -3 + 10 - 6 = 1

Cramer's Rule

\det(A_3) = \det\begin{pmatrix} 1 & 0 & 5 \\ -1 & 3 & 0 \\ 2 & 1 & 3 \end{pmatrix} = 1(9 - 0) - 0 + 5(-1 - 6) = 9 - 35 = -26

The solution is

x_1 = -23/(-15) = 23/15

,

x_2 = 1/(-15) = -1/15

,

x_3 = -26/(-15) = 26/15

.

Theoretical Significance

COROLLARY

Cramer's rule proves that each solution component is a rational function of the matrix entries and the right-hand side entries.

Fact: Maclaurin had the rule first

Colin Maclaurin's Treatise of Algebra, printed in 1748, two years after his death, already solves systems in 2 and 3 unknowns this way. Cramer's book of 1750 stated the rule for any number of unknowns, with a rule for the signs, and his name stuck.

This has consequences in pure algebra and in sensitivity analysis, where it shows how solutions respond to perturbations in the data. As a computational method, however, it requires

n + 1

determinant evaluations, making it far more expensive than Gaussian elimination for large systems.

The ratios in the rule were written down before determinants had a name or a notation. Early statements spelled out each numerator and each denominator as a sum of signed products of coefficients, six products for three unknowns, and left the reader to follow the pattern of signs to larger systems. The theory of determinants later grew out of exactly these sums.

Component	Column replaced by b in A	det(A_i)	x_i = det(A_i) / det(A)
x₁	column 1 → b = (5, 0, 3)	−23	23 / 15
x₂	column 2 → b = (5, 0, 3)	1	−1 / 15
x₃	column 3 → b = (5, 0, 3)	−26	26 / 15

One column swapped for the right-hand side

The column belonging to the unknown being solved for has been replaced by the constants, and the determinant of that altered matrix divided by the original gives the value. Each unknown costs its own determinant, which is why the rule is elegant for two or three and unusable beyond. Solve a full system with it on the Cramer's rule visualizer.

The rule is a statement about structure rather than a practical method; elimination beats it for anything you would actually solve.

Determinant Application Notation

Notation

Determinant Application Notation

Subscripts on this page rarely mean what subscripts usually mean. Four of them name a constructed matrix, a matrix size, a condition on pairs, and a point of evaluation.

The operator spelling

\det(A)

, the bar delimiters

|A|

and the

\Delta_x/\Delta

tradition of Cramer's own school belong to determinant properties. Minors, cofactors and the transpose hidden inside

\operatorname{adj}(A)

are set out at cofactor expansion, the

\hat{\mathbf{i}}\,\hat{\mathbf{j}}\,\hat{\mathbf{k}}

row and the symbolic-determinant device at the cross product, and

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

at the characteristic equation.

x_i = \frac{\det(A_i)}{\det(A)}

x sub i is the determinant of A sub i over the determinant of A

A_i

is not an entry, a row, a column or a member of a sequence. It is a whole matrix built for the occasion:

A

with its

i

-th column thrown away and

\mathbf{b}

put in its place, every other column untouched. The subscript names a construction the notation never describes.

CasesThe two subscripts in the formula do different jobs. On the left,

i

selects a component of the solution; on the right, the same

i

selects which column was replaced. They match because the rule pairs them, not because the symbols are the same kind of thing.

Also writtenCramer's own school writes

\Delta

for the coefficient determinant and

\Delta_{x}

,

\Delta_{y}

for the replaced ones, naming the columns by the unknown rather than by number.

Do not confuseReading

A_i

as a submatrix, in the way that deleting a row and column produces one for a minor. Nothing is deleted here:

A_i

is the same size as

A

, and the replacement is a substitution rather than a removal.

W(f_1, \dots, f_n)(x)

the Wronskian of f one through f n, evaluated at x

Two bracketed lists in a row, and they hold different kinds of thing. The first takes the functions and fixes which determinant is being formed; the second supplies the point at which that determinant is worked out.

W

consumes functions and returns a function, so the answer varies with

x

and a single number is never the whole object.

CasesInside the matrix, one symbol carries two indices with unrelated jobs: in

f_1^{(n-1)}

the subscript picks the function and the parenthesised superscript gives the order of differentiation. The parentheses are what keep it from reading as a power.

Also writtenShortened to

W(x)

once the functions are fixed by context, which hides the first list entirely and makes the object look like an ordinary function of one variable.

Do not confuseTesting the value at one convenient point and concluding dependence when it comes out zero. A single nonzero value proves independence; vanishing everywhere proves nothing unless the functions are already known to solve one linear differential equation.

\det(V) = \prod_{1 \leq i < j \leq n} (x_j - x_i)

the product over all pairs i less than j between one and n, of x j minus x i

What sits under the product sign is a condition, not a counter running from a start to a finish. It selects every pair of indices with

i

below

j

and multiplies one factor for each, which is

\binom{n}{2}

factors rather than

n

.

CasesThe condition also fixes the order inside each factor: the larger index comes first, so every difference is written

x_j - x_i

. Swapping them in one factor flips the sign of the whole product, which is why the inequality is part of the notation rather than a remark about it.

Also writtenSpelled out as

\prod_{i<j}

when the range

1

to

n

is clear, and occasionally as a nested double product, which obscures that each pair is counted once.

Do not confuseReading it as a single-index product and expecting

n

factors. For four nodes the formula has six, and the count is the first thing to check when a hand computation disagrees.

D_n = a_nD_{n-1} - b_nc_nD_{n-2}

D sub n equals a sub n times D sub n minus one, minus b sub n c sub n times D sub n minus two

The subscript on

D

is a matrix size.

D_n

is the determinant of the

n \times n

tridiagonal matrix,

D_{n-1}

the determinant of a smaller one, so the recurrence relates numbers coming from matrices of three different dimensions in a single line.

CasesThe subscripts on

a

,

b

and

c

mean something else in the same equation: they locate entries within one matrix. Two subscript conventions sit side by side, and only the letter tells them apart.

Also writtenWritten

\det(T_n)

when the matrices are named explicitly, which makes the size-indexing visible at the cost of a longer line.

Do not confuseTreating

D_{n-1}

and

D_{n-2}

as earlier values of one fixed matrix, in the way a sequence recurrence usually works. Each term here belongs to a different matrix, and the family has to be defined before the recurrence means anything.

Same glyph elsewhere

D

names a diagonal matrix at diagonalization and a scalar determinant here, and the letter is also the older rival to

\Delta

in the Cramer tradition.

All linear algebra symbols →↑ Determinants

The Inverse via the Adjugate

The adjugate identity

A \cdot \operatorname{adj}(A) = \det(A) \cdot I

immediately gives an explicit formula for the inverse when

\det(A) \neq 0

:

Inverse via Adjugate

A^{-1} = \frac{1}{\det(A)} \operatorname{adj}(A)

Learn more about this formula: Inverse via Adjugate →

Inverse via adjugate →

Every entry of

A^{-1}

is expressed as a cofactor of

A

divided by

\det(A)

.

The 2×2 Case

For

A = \begin{pmatrix} a & b \\ c & d \end{pmatrix}

, the cofactor matrix is

\begin{pmatrix} d & -c \\ -b & a \end{pmatrix}

, and transposing gives

\operatorname{adj}(A) = \begin{pmatrix} d & -b \\ -c & a \end{pmatrix}

. The inverse is

The Inverse via the Adjugate

A^{-1} = \frac{1}{ad - bc} \begin{pmatrix} d & -b \\ -c & a \end{pmatrix}

This is the familiar swap-the-diagonal, negate-the-off-diagonal formula that appears in every introductory linear algebra course.

3×3 Worked Example

For

A = \begin{pmatrix} 1 & 2 & 0 \\ 0 & 1 & 3 \\ 2 & 0 & 1 \end{pmatrix}

, first compute

\det(A)

by expanding along the first row:

The Inverse via the Adjugate

\det(A) = 1(1 - 0) - 2(0 - 6) + 0 = 1 + 12 = 13

The nine cofactors are:

The Inverse via the Adjugate

C_{11} = +(1 \cdot 1 - 3 \cdot 0) = 1, \quad C_{12} = -(0 \cdot 1 - 3 \cdot 2) = 6, \quad C_{13} = +(0 \cdot 0 - 1 \cdot 2) = -2

The Inverse via the Adjugate

C_{21} = -(2 \cdot 1 - 0 \cdot 0) = -2, \quad C_{22} = +(1 \cdot 1 - 0 \cdot 2) = 1, \quad C_{23} = -(1 \cdot 0 - 2 \cdot 2) = 4

The Inverse via the Adjugate

C_{31} = +(2 \cdot 3 - 0 \cdot 1) = 6, \quad C_{32} = -(1 \cdot 3 - 0 \cdot 0) = -3, \quad C_{33} = +(1 \cdot 1 - 2 \cdot 0) = 1

The adjugate is the transpose of the cofactor matrix:

The Inverse via the Adjugate

\operatorname{adj}(A) = \begin{pmatrix} 1 & -2 & 6 \\ 6 & 1 & -3 \\ -2 & 4 & 1 \end{pmatrix}

So

A^{-1} = \frac{1}{13} \begin{pmatrix} 1 & -2 & 6 \\ 6 & 1 & -3 \\ -2 & 4 & 1 \end{pmatrix}

.

Pitfall: Forgetting to transpose the cofactor matrix

Dividing the cofactor matrix by

13

without transposing it gives a candidate whose product with

A

has

(1,2)

entry

\frac{1}{13}(1 \cdot 6 + 2 \cdot 1 + 0 \cdot (-3)) = \frac{8}{13}

instead of

0

. Transposed first, the cofactors give

\operatorname{adj}(A)

above.

Verification:

A \cdot A^{-1}

should produce the identity. The

(1,1)

entry is

\frac{1}{13}(1 \cdot 1 + 2 \cdot 6 + 0 \cdot (-2)) = \frac{13}{13} = 1

. The

(1,2)

entry is

\frac{1}{13}(1 \cdot (-2) + 2 \cdot 1 + 0 \cdot 4) = \frac{0}{13} = 0

. The remaining entries check out similarly.

Practical Assessment

The adjugate formula writes every entry of the inverse as an explicit ratio of cofactors and the determinant. This is valuable for symbolic work — it shows exactly how each entry of

A^{-1}

depends on the entries of

A

. For numerical computation on matrices larger than

3 \times 3

, row reduction is vastly more efficient.

The cofactor array, before transposing

Transposing this array and dividing by the determinant produces the inverse, which puts the invertibility condition in plain view: every step works until the division, and the division fails precisely when the determinant is zero. Follow it through to the finished inverse on the matrix inverse visualizer.

The same warning applies here: exact, illuminating, and far too expensive to use beyond small matrices.

The Cross Product as a Determinant

The cross product of two vectors

\mathbf{a} = (a_1, a_2, a_3)

and

\mathbf{b} = (b_1, b_2, b_3)

in

\mathbb{R}^3

can be computed as a symbolic

3 \times 3

determinant:

The Cross Product as a Determinant

\mathbf{a} \times \mathbf{b} = \det\begin{pmatrix} \hat{\mathbf{i}} & \hat{\mathbf{j}} & \hat{\mathbf{k}} \\ a_1 & a_2 & a_3 \\ b_1 & b_2 & b_3 \end{pmatrix}

Expanding along the first row using the cofactor formula:

The Cross Product as a Determinant

\mathbf{a} \times \mathbf{b} = \hat{\mathbf{i}}(a_2 b_3 - a_3 b_2) - \hat{\mathbf{j}}(a_1 b_3 - a_3 b_1) + \hat{\mathbf{k}}(a_1 b_2 - a_2 b_1)

Each component of the resulting vector is a

2 \times 2

minor — the sub-determinant obtained by deleting the appropriate row and column from the lower two rows.

This is a formal rather than literal use of the determinant. The first row contains basis vectors, not numbers, so the "determinant" is not a scalar but a vector. The cofactor expansion still applies mechanically, and the alternating signs

+, -, +

produce the correct cross product components.

Worked Example

For

\mathbf{a} = (2, -1, 3)

and

\mathbf{b} = (4, 0, -2)

:

The Cross Product as a Determinant

\mathbf{a} \times \mathbf{b} = \det\begin{pmatrix} \hat{\mathbf{i}} & \hat{\mathbf{j}} & \hat{\mathbf{k}} \\ 2 & -1 & 3 \\ 4 & 0 & -2 \end{pmatrix}

The Cross Product as a Determinant

= \hat{\mathbf{i}}((-1)(-2) - (3)(0)) - \hat{\mathbf{j}}((2)(-2) - (3)(4)) + \hat{\mathbf{k}}((2)(0) - (-1)(4))

The Cross Product as a Determinant

= \hat{\mathbf{i}}(2) - \hat{\mathbf{j}}(-16) + \hat{\mathbf{k}}(4) = (2, 16, 4)

Pitfall: Dropping the minus sign on the middle component

Leaving out the minus in front of

\hat{\mathbf{j}}

turns the result into

(2, -16, 4)

, whose dot product with

\mathbf{a}

is

4 + 16 + 12 = 32

, not

0

. With the sign,

\mathbf{a} \times \mathbf{b} = (2, 16, 4)

.

The magnitude is

|\mathbf{a} \times \mathbf{b}| = \sqrt{4 + 256 + 16} = \sqrt{276} = 2\sqrt{69}

. This equals the area of the parallelogram spanned by

\mathbf{a}

and

\mathbf{b}

, connecting the cross product back to the geometric interpretation of the determinant as an area measure.

The cross product written as a determinant

Basis vectors occupy the first row and the two operands fill the other two, so the cross product is a cofactor expansion wearing different clothes. The alternating sign board is what puts the minus on the middle component, and the whole thing vanishing means the two rows were proportional — that is, the vectors were parallel. Expand it yourself on the cross product visualizer.

Seen this way the cross product is not a new operation at all but a determinant with a basis row bolted on.

The Characteristic Polynomial

For an

n \times n

matrix

A

, the characteristic polynomial is defined as

Characteristic Polynomial

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

Learn more about this formula: Characteristic Polynomial →

Characteristic polynomial →

This is a polynomial of degree

n

in the variable

\lambda

. Its roots are the eigenvalues of

A

— the scalars

\lambda

for which the matrix

A - \lambda I

becomes singular.

2×2 Example

For

A = \begin{pmatrix} 4 & 1 \\ 2 & 3 \end{pmatrix}

:

The Characteristic Polynomial

A - \lambda I = \begin{pmatrix} 4 - \lambda & 1 \\ 2 & 3 - \lambda \end{pmatrix}

The Characteristic Polynomial

p(\lambda) = (4 - \lambda)(3 - \lambda) - 2 = \lambda^2 - 7\lambda + 10 = (\lambda - 2)(\lambda - 5)

The eigenvalues are

\lambda = 2

and

\lambda = 5

.

Pitfall: Subtracting lambda from every entry

Taking

\lambda

off every entry gives

\det\begin{pmatrix} 4 - \lambda & 1 - \lambda \\ 2 - \lambda & 3 - \lambda \end{pmatrix} = 10 - 4\lambda

, a polynomial of degree

1

. Since

\lambda I

is diagonal, only the diagonal changes, and

p(\lambda) = \lambda^2 - 7\lambda + 10

.

Two quick checks come from the matrix itself. The eigenvalues add up to

2 + 5 = 7

, which is the sum

4 + 3

of the diagonal entries of

A

, and they multiply to

2 \cdot 5 = 10

, which is

\det(A) = 4 \cdot 3 - 1 \cdot 2

. Both facts hold for every square matrix, as the two identities at the end of this section show.

3×3 Example

For

A = \begin{pmatrix} 2 & 1 & 0 \\ 0 & 3 & 1 \\ 0 & 0 & 1 \end{pmatrix}

:

This is upper triangular, so

A - \lambda I

is also upper triangular with diagonal entries

2 - \lambda

,

3 - \lambda

,

1 - \lambda

. The determinant of a triangular matrix is the product of its diagonal entries:

The Characteristic Polynomial

p(\lambda) = (2 - \lambda)(3 - \lambda)(1 - \lambda)

The eigenvalues are

\lambda = 1, 2, 3

— they sit directly on the diagonal, which is always the case for triangular matrices.

Two Identities

Setting

\lambda = 0

in the characteristic polynomial gives

p(0) = \det(A)

, which means the constant term of the characteristic polynomial is the determinant. Since the roots of

p

are the eigenvalues

\lambda_1, \dots, \lambda_n

, this yields

Determinant Product of Eigenvalues

\det(A) = \lambda_1 \, \lambda_2 \cdots \lambda_n

Determinant product of eigenvalues →

The determinant equals the product of all eigenvalues, counted with algebraic multiplicity. A second identity connects the coefficient of

\lambda^{n-1}

to the trace:

The Characteristic Polynomial

\lambda_1 + \lambda_2 + \cdots + \lambda_n = \operatorname{tr}(A)

Together, these two identities link the determinant and trace to the eigenvalue spectrum of the matrix.

A determinant turns the search for eigenvalues into finding the roots of one polynomial.

The Wronskian

The Wronskian extends the determinant's role as a linear independence test from vectors to functions. Given

n

functions

f_1, f_2, \dots, f_n

, each differentiable at least

n - 1

times, the Wronskian is

The Wronskian

W(f_1, \dots, f_n)(x) = \det\begin{pmatrix} f_1(x) & f_2(x) & \cdots & f_n(x) \\ f_1'(x) & f_2'(x) & \cdots & f_n'(x) \\ \vdots & \vdots & \ddots & \vdots \\ f_1^{(n-1)}(x) & f_2^{(n-1)}(x) & \cdots & f_n^{(n-1)}(x) \end{pmatrix}

Each column corresponds to one function, and each row raises the order of differentiation by one. The result is a function of

x

, not a constant.

The Independence Test

THEOREMWronskian Test

If

W(f_1, \dots, f_n)(x_0) \neq 0

at some point

x_0

, then the functions

f_1, \dots, f_n

are linearly independent on any interval containing

x_0

.

The logic mirrors the matrix case: a nonzero determinant means the "columns" — here the function-derivative profiles — are linearly independent.

The converse requires care. A Wronskian that vanishes everywhere does not automatically imply dependence unless the functions are known to be solutions of a single linear ordinary differential equation. Without that structural assumption, counterexamples exist.

Worked Example

Take

f_1 = e^x

,

f_2 = e^{2x}

,

f_3 = e^{3x}

. The Wronskian matrix is

The Wronskian

\begin{pmatrix} e^x & e^{2x} & e^{3x} \\ e^x & 2e^{2x} & 3e^{3x} \\ e^x & 4e^{2x} & 9e^{3x} \end{pmatrix}

Factoring

e^x

from column

1

,

e^{2x}

from column

2

, and

e^{3x}

from column

3

:

The Wronskian

W = e^{6x} \det\begin{pmatrix} 1 & 1 & 1 \\ 1 & 2 & 3 \\ 1 & 4 & 9 \end{pmatrix}

The remaining matrix is a Vandermonde matrix with nodes

1, 2, 3

. Its determinant is

(2 - 1)(3 - 1)(3 - 2) = 1 \cdot 2 \cdot 1 = 2

. So

W = 2e^{6x}

, which is nonzero for all

x

, confirming that

e^x, e^{2x}, e^{3x}

are linearly independent.

Context

The Wronskian arises most naturally in the theory of linear ordinary differential equations, where it determines whether a proposed set of solutions forms a fundamental system. Abel's identity gives a differential equation for the Wronskian itself, relating its evolution to the coefficient in the ODE. These developments belong to differential equations rather than linear algebra, but the underlying mechanism — testing independence via a determinant — is purely algebraic.

Vandermonde and Structured Determinants

Certain matrices with patterned entries have determinants that admit elegant closed-form expressions. The most important of these is the Vandermonde matrix.

The Vandermonde Determinant

An

n \times n

Vandermonde matrix is built from

n

distinct nodes

x_1, x_2, \dots, x_n

:

Vandermonde and Structured Determinants

V = \begin{pmatrix} 1 & x_1 & x_1^2 & \cdots & x_1^{n-1} \\ 1 & x_2 & x_2^2 & \cdots & x_2^{n-1} \\ \vdots & \vdots & \vdots & \ddots & \vdots \\ 1 & x_n & x_n^2 & \cdots & x_n^{n-1} \end{pmatrix}

Its determinant has the closed form

Vandermonde Determinant

\det(V) = \prod_{1 \leq i < j \leq n} (x_j - x_i)

Vandermonde determinant →

The product runs over all pairs with

j > i

, so it contains

\binom{n}{2}

factors. Each factor is a difference between two nodes.

3×3 Verification

For nodes

x_1 = 1

,

x_2 = 2

,

x_3 = 4

:

Vandermonde and Structured Determinants

V = \begin{pmatrix} 1 & 1 & 1 \\ 1 & 2 & 4 \\ 1 & 4 & 16 \end{pmatrix}

Direct expansion:

\det(V) = 1(32 - 16) - 1(16 - 4) + 1(4 - 2) = 16 - 12 + 2 = 6

.

The product formula:

(x_2 - x_1)(x_3 - x_1)(x_3 - x_2) = (2 - 1)(4 - 1)(4 - 2) = 1 \cdot 3 \cdot 2 = 6

.

Why It Matters

COROLLARY

The Vandermonde determinant is nonzero precisely when all nodes are distinct.

Fact:

b^2 - 4c

is a squared Vandermonde

For

x^2 + bx + c

with roots

r_1, r_2

, the discriminant

b^2 - 4c

equals

(r_2 - r_1)^2

, the square of the

2 \times 2

Vandermonde determinant of the roots. For every polynomial with leading coefficient

1

, the discriminant is the squared Vandermonde of its roots, which is why it is zero exactly at a repeated root.

This guarantees that a polynomial of degree at most

n - 1

is uniquely determined by its values at

n

distinct points — the theoretical foundation of polynomial interpolation. It also appears in the theory of symmetric polynomials and in the derivation of various discrete orthogonality relations.

Exchanging two nodes exchanges two rows of

V

, so the determinant changes sign. Its square does not change at all:

\det(V)^2

stays the same under every reordering of the nodes, which makes it a symmetric polynomial in them. When the nodes are the roots of a polynomial, any symmetric polynomial in the roots can be rewritten in terms of the coefficients, without solving for the roots.

Other Structured Determinants

Several other matrix families have known determinant formulas. Circulant matrices, built from cyclic shifts of a single row, have determinants expressible through the discrete Fourier transform: if the first row is

(c_0, c_1, \dots, c_{n-1})

, then

\det(C) = \prod_{k=0}^{n-1} p(\omega^k)

where

p(x) = c_0 + c_1 x + \cdots + c_{n-1} x^{n-1}

and

\omega = e^{2\pi i/n}

is a primitive

n

-th root of unity.

Hilbert matrices, with entries

H_{ij} = \frac{1}{i + j - 1}

, have a closed-form determinant involving products of factorials. These matrices are notoriously ill-conditioned — their condition numbers grow rapidly with

n

, reflecting extreme sensitivity to perturbation; their determinants also shrink rapidly, but a small determinant alone does not measure conditioning.

Tridiagonal matrices, with nonzero entries only on the main diagonal and the two adjacent diagonals, have determinants satisfying a three-term recurrence: if

D_n

denotes the determinant of the

n \times n

tridiagonal matrix, then

D_n = a_n D_{n-1} - b_n c_n D_{n-2}

, where

a_n

is the

n

-th diagonal entry and

b_n, c_n

are the adjacent off-diagonal entries. This recurrence allows

O(n)

computation, much faster than general methods.

Each of these families illustrates the same principle: when a matrix has special structure, its determinant often has a formula that exploits that structure directly, bypassing both cofactor expansion and row reduction.

Family	Structure	Determinant formula	Key property / use
Vandermonde	entry V_ij = x_i^j−1; columns are 1, x, x², …, xⁿ⁻¹	∏_{i < j} (x_j − x_i)	nonzero iff nodes are distinct → unique polynomial interpolation
Circulant	each row is a cyclic shift of the first row (c₀, c₁, …, c_n−1)	∏_k=0ⁿ⁻¹ p(ω^k), where p(x) = Σ c_j x^j and ω = e^2πi/n	diagonalized by the discrete Fourier transform
Hilbert	entry H_ij = 1 / (i + j − 1)	closed form involving products of factorials; shrinks rapidly with n	notoriously ill-conditioned; benchmark for numerical sensitivity
Tridiagonal	nonzero entries only on the main diagonal and the two adjacent diagonals	three-term recurrence D_n = a_n D_n−1 − b_n c_n D_n−2	O(n) computation — much faster than general methods

Summary: Determinant Applications at a Glance

The six applications above span linear systems, inverses, vector geometry, eigenvalue analysis, function-space independence, and structured matrix families. The table below collects each one alongside what the determinant produces, the key formula that drives it, and the situation in which it is the right tool to reach for.

The six uses below divide along a line that is easy to miss when they are listed together. In the first group the determinant produces a number that something else could have produced faster; in the second it produces a decision that nothing else produces at all. Reading them as one undifferentiated list of applications tends to leave the impression that determinants are a computational tool, which is close to the opposite of the truth.

Download

PNGRaster · 2×PDFPrint quality

Determinants · applications

Six places a determinant does the work

Three of these produce a value and three produce a verdict. The split matters because the first group is almost never the fastest route to that value — while the second group has no competitor at all.

6applications

Producing a value — correct, rarely fastest3

1

Cramer’s rule§ 1

\det(A) \neq 0

;

A_i

has column

i

replaced by

\mathbf{b}

x_i = \dfrac{\det(A_i)}{\det(A)}

Each component as a ratio of determinants. Costs

n{+}1

determinants against one elimination, so it is a formula rather than a method — valuable because it exhibits the solution as an explicit function of the entries, which is what sensitivity analysis needs.

2

Adjugate inverse§ 2

\det(A) \neq 0

A^{-1} = \dfrac{1}{\det A}\operatorname{adj}(A)

Every entry of the inverse in closed form. Practical at

2 \times 2

and defensible at

3 \times 3

; beyond that the cofactor count makes it unusable numerically. Its real work is theoretical — it proves the inverse entries are rational functions of

A

.

3

Structured determinants§ 6

the matrix has exploitable structure

Vandermonde:

\prod_{i<j}(x_j - x_i)

Closed forms that skip the computation entirely. The Vandermonde value is nonzero exactly when the nodes are distinct, which is the statement that polynomial interpolation through distinct points has a unique solution.

Producing a verdict — nothing else does it as well3

4

Characteristic polynomial§ 4

roots are the eigenvalues

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

The determinant is what converts "this homogeneous system has a nontrivial solution" into a polynomial equation in

\lambda

. Without it there is no equation to solve, which is why eigenvalue theory begins here rather than with eigenvectors.

5

Cross product§ 3

\mathbb{R}^3

only

\mathbf{a} \times \mathbf{b} = \det\begin{bmatrix}\hat{\imath} & \hat{\jmath} & \hat{k} \\ \mathbf{a} \\ \mathbf{b}\end{bmatrix}

A mnemonic that is also an explanation. Expanding along the first row reproduces the component formula, and the determinant being zero when two rows match is exactly why

\mathbf{a} \times \mathbf{a} = \mathbf{0}

— see cross product.

6

Wronskian§ 5

W \neq 0

somewhere

\Rightarrow

independent

W(x) = \det\bigl[f_i^{(j)}(x)\bigr]

One-directional, and the direction matters. A nonzero value at a single point proves independence;

W \equiv 0

proves nothing in general —

x^2

and

x|x|

are independent with

W

identically zero. The converse holds only for solutions of one linear ODE.

Where a determinant computes something, elimination usually computes it faster; where a determinant decides something, nothing else decides it as cleanly. That is the honest summary of the determinant’s role: a poor algorithm and an excellent criterion.

Six places a determinant does the work·/linear-algebra/determinants/applicationsLearn Math Class

Cramer's rule and the adjugate inverse are the clearest cases. Both are correct, both are taught, and both are slower than elimination by a margin that grows with

n

— Cramer's rule needs

n{+}1

determinants where elimination needs one pass. What they offer instead is a closed form: the solution written as an explicit function of the matrix entries, which is what makes them useful for symbolic work and for asking how the answer responds when an entry changes.

The second group has no such competitor. The characteristic polynomial exists because the determinant turns a statement about nontrivial solutions into a polynomial equation, and without that step there is nothing to solve for. The Wronskian tests something no elimination can reach, since its entries are functions rather than numbers. These are the applications that justify the determinant rather than merely using it.

Common Pitfalls

Four slips account for many wrong answers when determinants are put to work. Each is marked with a box where it happens in the sections above. Here each is worked through in full, with the wrong and the right working side by side.

Replacing a row instead of a column

In Cramer's rule,

A_i

is

A

with column

i

replaced by

\mathbf{b}

. The vector

\mathbf{b}

is written into row

i

instead, the way the equations are read off one line at a time.

Wrong working: for the

2 \times 2

example, with

\mathbf{b} = (8, 7)

in row

1

and then in row

2

,

\det\begin{pmatrix} 8 & 7 \\ 1 & 5 \end{pmatrix} = 40 - 7 = 33, \quad \det\begin{pmatrix} 3 & 2 \\ 8 & 7 \end{pmatrix} = 21 - 16 = 5

so

x_1 = 33/13

and

x_2 = 5/13

.

Right working: with

\mathbf{b}

in column

1

and then in column

2

,

\det\begin{pmatrix} 8 & 2 \\ 7 & 5 \end{pmatrix} = 26, \quad \det\begin{pmatrix} 3 & 8 \\ 1 & 7 \end{pmatrix} = 13

so

x_1 = 2

and

x_2 = 1

.

Why it happens: each equation is a row of

A

, and since

\det(A^T) = \det(A)

, rows and columns feel interchangeable. They are not here: replacing rows of

A

is replacing columns of

A^T

, so the wrong working solves the different system

A^T \mathbf{x} = \mathbf{b}

.

How to check: substitute into the first equation.

3 \cdot \frac{33}{13} + 2 \cdot \frac{5}{13} = \frac{109}{13}

, not

8

, while

3 \cdot 2 + 2 \cdot 1 = 8

.

Forgetting to transpose the cofactor matrix

The cofactors are arranged with

C_{ij}

in row

i

and column

j

, and this cofactor matrix is divided by

\det(A)

directly. The adjugate is its transpose:

C_{ij}

belongs in row

j

and column

i

.

Wrong working: for the

3 \times 3

example, the candidate

\frac{1}{13}\begin{pmatrix} 1 & 6 & -2 \\ -2 & 1 & 4 \\ 6 & -3 & 1 \end{pmatrix}

multiplied by

A

gives

\frac{1}{13}\begin{pmatrix} -3 & 8 & 6 \\ 16 & -8 & 7 \\ 8 & 9 & -3 \end{pmatrix}

which is not

I

.

Right working: transpose first.

\operatorname{adj}(A) = \begin{pmatrix} 1 & -2 & 6 \\ 6 & 1 & -3 \\ -2 & 4 & 1 \end{pmatrix}, \quad A \operatorname{adj}(A) = 13I

so

A^{-1} = \frac{1}{13}\operatorname{adj}(A)

.

Why it happens: each cofactor is computed in its natural position, and the transpose is a separate step that is easy to skip. For a symmetric matrix the cofactor matrix is symmetric too, so the slip goes unnoticed in many small examples.

How to check: compute one off-diagonal entry of

A

times the candidate. Row

1

of

A

times column

2

of the untransposed candidate is

\frac{1}{13}(1 \cdot 6 + 2 \cdot 1 + 0 \cdot (-3)) = \frac{8}{13}

, where

0

is needed.

Dropping the minus sign on the middle component

The first-row expansion of the cross product determinant carries the signs

+, -, +

. The middle

2 \times 2

minor is written down as the

\hat{\mathbf{j}}

component without its minus sign.

Wrong working: for

\mathbf{a} = (2, -1, 3)

and

\mathbf{b} = (4, 0, -2)

, the three minors are

2

,

-16

and

4

, and they are copied straight into

(2, -16, 4)

.

Right working: the middle minor is subtracted,

-(-16) = 16

, so

\mathbf{a} \times \mathbf{b} = (2, 16, 4)

.

Why it happens: the other two components keep their minors unchanged, and nothing in the middle minor itself shows the sign. The minus belongs to the position in the expansion, not to the entries.

How to check:

\mathbf{a} \times \mathbf{b}

is perpendicular to both vectors. For

(2, 16, 4)

, the dot products are

4 - 16 + 12 = 0

with

\mathbf{a}

and

8 + 0 - 8 = 0

with

\mathbf{b}

. For

(2, -16, 4)

, the dot product with

\mathbf{a}

is

4 + 16 + 12 = 32

.

Subtracting lambda from every entry

A - \lambda I

is formed by subtracting

\lambda

from every entry of

A

. But

\lambda I

has

\lambda

only on the diagonal and zeros elsewhere, so only the diagonal entries change.

Wrong working: for

A = \begin{pmatrix} 4 & 1 \\ 2 & 3 \end{pmatrix}

,

\det\begin{pmatrix} 4 - \lambda & 1 - \lambda \\ 2 - \lambda & 3 - \lambda \end{pmatrix} = (\lambda^2 - 7\lambda + 12) - (\lambda^2 - 3\lambda + 2) = 10 - 4\lambda

a polynomial of degree

1

with the single root

\lambda = \frac{5}{2}

.

Right working:

\det\begin{pmatrix} 4 - \lambda & 1 \\ 2 & 3 - \lambda \end{pmatrix} = \lambda^2 - 7\lambda + 10 = (\lambda - 2)(\lambda - 5)

with the eigenvalues

2

and

5

.

Why it happens:

A - \lambda

looks like a number subtracted from a matrix, which is not defined, and it is read as subtraction from every entry. The

I

is what places

\lambda

on the diagonal only.

How to check: the characteristic polynomial of an

n \times n

matrix has degree

n

, here

2

. The eigenvalues must also add up to the trace,

4 + 3 = 7

, and multiply to

\det(A) = 10

; the single value

\frac{5}{2}

fails both.

Determinant Applications FAQ

In Cramer's rule, is $A_i$ a smaller matrix?

+

No, it is exactly the same size as

A

. The subscript marks a substitution rather than a deletion: column

i

has been replaced by the constant vector, and everything else is untouched. That distinguishes it from a minor, where a row and a column really are removed and the result genuinely shrinks.Read more →

Does a Wronskian of zero prove that functions are dependent?

+

No, and this is the usual overstatement. The implication runs one way only: a single nonzero value proves independence. Vanishing everywhere proves nothing on its own, and only becomes conclusive when the functions are already known to solve one linear differential equation. Testing one convenient point settles nothing either way.Read more →

How many factors does a Vandermonde determinant have?

+

One for every pair of nodes, so

\binom{n}{2}

rather than

n

. Four nodes give six factors, not four, and miscounting is the first thing to check when a hand computation disagrees. The ordering matters too: each difference is written with the larger index first, since swapping one factor flips the sign of the whole product.Read more →

Applications of Determinants

Solving Problems with Determinants

Cramer's Rule

2×2 Example

3×3 Example

Theoretical Significance

Determinant Application Notation

Determinant Application Notation

The Inverse via the Adjugate

The 2×2 Case

3×3 Worked Example

Practical Assessment

The Cross Product as a Determinant

Worked Example

The Characteristic Polynomial

2×2 Example

3×3 Example

Two Identities

The Wronskian

The Independence Test

Worked Example

Context

Vandermonde and Structured Determinants

The Vandermonde Determinant

3×3 Verification

Why It Matters

Other Structured Determinants

Summary: Determinant Applications at a Glance

Six places a determinant does the work

Common Pitfalls

Determinant Applications FAQ

In Cramer's rule, is $A_i$ a smaller matrix?

Does a Wronskian of zero prove that functions are dependent?

How many factors does a Vandermonde determinant have?

Applications of Determinants

Solving Problems with Determinants

Cramer's Rule

2×2 Example

3×3 Example

Theoretical Significance

Determinant Application Notation

Determinant Application Notation

The Inverse via the Adjugate

The 2×2 Case

3×3 Worked Example

Practical Assessment

The Cross Product as a Determinant

Worked Example

The Characteristic Polynomial

2×2 Example

3×3 Example

Two Identities

The Wronskian

The Independence Test

Worked Example

Context

Vandermonde and Structured Determinants

The Vandermonde Determinant

3×3 Verification

Why It Matters

Other Structured Determinants

Summary: Determinant Applications at a Glance

Six places a determinant does the work

Common Pitfalls

Determinant Applications FAQ

In Cramer's rule, is AiA_iAi​ a smaller matrix?

Does a Wronskian of zero prove that functions are dependent?

How many factors does a Vandermonde determinant have?

In Cramer's rule, is $A_i$ a smaller matrix?