The Determinant as an Alternating Form

The top exterior power $\Lambda^n V$ underlies the modern definition of orientation in differential geometry: an orientation of an $n$-manifold is a consistent choice of generator of $\Lambda^n(T_pM)^*$ at each point, and the determinant of the Jacobian of a coordinate change is exactly the scalar by which the top form transforms — a direct descendant of this construction. In differential forms and Stokes' theorem, the top-degree form $dx_1 \wedge \cdots \wedge dx_n$ plays the role of $\omega_0$, and pullback of forms under a diffeomorphism generalizes the operator action $T^*\omega$ studied here. In multivariable calculus, the change-of-variables formula for integrals features the absolute value of the Jacobian determinant precisely because it is the scalar factor by which the volume form is rescaled. In physics, the Jacobian determinant of a canonical transformation appearing in Liouville's theorem (conservation of phase-space volume in Hamiltonian mechanics) is again this same scalar, now acting on $\Lambda^{2n}$ of phase space. In algebraic geometry and representation theory, $\Lambda^n V$ for $V$ an $n$-dimensional representation is the determinant representation, a one-dimensional representation whose character is exactly $\det T$ on group elements $T$; this is a fundamental building block in the representation theory of $GL_n$ and appears throughout invariant theory. In numerical linear algebra and computer graphics, the sign and magnitude of $\det T$ for a transformation matrix $T$ tells you instantly whether an orientation (e.g., of a triangle mesh) is preserved or flipped, and by what volumetric factor — a direct computational shadow of the coordinate-free scalar constructed here.

Ask the tutor: Why does the alternating property alone force $\dim \Lambda^n(V^*) = 1$ when $\dim V = n$, and where exactly does the proof of that fact use the assumption that we're looking at exactly $n$-linear forms rather than $k$-linear forms for $k \ne n$?

Think of an alternating $n$-form $\omega$ on an $n$-dimensional space $V$ as a signed volume functional: $\omega(v_1,\dots,v_n)$ measures the signed volume of the parallelepiped spanned by $v_1,\dots,v_n$, relative to some fixed unit of volume. Multilinearity captures the idea that stretching one edge of the parallelepiped by a scalar $c$ scales the volume by $c$, and additivity in each slot encodes how volumes combine when you shear one edge by adding another vector to it. The alternating condition — vanishing when two arguments coincide — encodes the geometric fact that a degenerate parallelepiped (two edges equal, hence lying in a common line) has zero volume. The one-dimensionality of $\Lambda^n(V^*)$ says something remarkable: up to an overall scalar, there is only one notion of signed volume on an $n$-dimensional space. Any two alternating top-forms are proportional. This matches intuition — once you fix a unit cube's volume to be $1$, the volume of every other parallelepiped is determined. Since a linear operator $T$ sends the "unit" parallelepiped spanned by any $v_1,\dots,v_n$ to the parallelepiped spanned by $Tv_1,\dots,Tv_n$, and since all volume functionals are proportional, $T$ must scale every volume functional by the same factor. That universal scaling factor is $\det T$.

Because the entire framework rests on $\Lambda^n(V^*)$ being one-dimensional, the theorem below both re-derives the key structural facts (that $T^*$ is a well-defined scalar operator) and proves that this scalar, computed via any basis, is exactly the classical matrix determinant. The technique is direct construction plus uniqueness: build the candidate scalar, then show it must equal the standard determinant by testing it against the standard basis.

Example 1: a shear-and-scale operator on $\mathbb{R}^3$, computed via the top exterior power directly. Let $V = \mathbb{R}^3$ with standard basis $e_1,e_2,e_3$, and let $T$ be defined by

So in the standard basis,

We compute $\det T$ two ways to illustrate the equivalence proven above. Via the classical formula. Expanding along the first row:

Via the alternating form directly. Let $\omega_0$ be the unique alternating $3$-form with $\omega_0(e_1,e_2,e_3) = 1$ (this is the signed-volume form, and $\omega_0(u,v,w) = \det[u\,|\,v\,|\,w]$ when $u,v,w$ are written in coordinates). By Step 3 of the theorem, $\det T = \omega_0(Te_1,Te_2,Te_3)$. Expand by multilinearity:

Rather than expanding all $27$ terms, we use the alternating property to discard repeated basis vectors term by term (equivalently, this reconstructs the permutation sum), which is exactly why the formula in Step 3 collapses to the six permutation terms of $S_3$. Carrying out that six-term sum with $A$'s entries: the permutations of $\{1,2,3\}$ contribute

Summing: $24 - 2 + 0 + 0 - 1 + 0 = 21$. This matches $\det A = 21$ found above, confirming that $\omega_0(Te_1,Te_2,Te_3)$ — the definition-theoretic computation of $\det T$ — agrees exactly with the classical cofactor expansion, as guaranteed by the theorem.

Example 2: basis-change invariance made explicit. Let $T : \mathbb{R}^2 \to \mathbb{R}^2$ be given in the standard basis $\beta = (e_1,e_2)$ by

Now switch to the (non-orthogonal, asymmetric) basis $\beta' = (f_1, f_2)$ with $f_1 = e_1 + e_2$, $f_2 = e_1 - 2e_2$. The change-of-basis matrix from $\beta'$-coordinates to $\beta$-coordinates is

The matrix of $T$ in the new basis is $A' = P^{-1}AP$. Compute:

Then $\det A' = 5 \cdot 2 - 0 \cdot 0 = 10$, matching $\det A = 10$ exactly. This is not a coincidence: the theorem shows both computations equal $\omega_0(Tv_1, Tv_2)/\omega_0(v_1,v_2)$-type invariants — more precisely, the scalar $c$ by which $T^*$ acts on the one-dimensional space $\Lambda^2(V^*)$ — which by Step 2 of the proof does not depend on which basis was used to build $A$ or $A'$. This equality is the content foreshadowed by Basis-Independence of the Determinant: the exterior-power definition of $\det T$ makes such basis-invariance automatic rather than something to verify by direct matrix computation each time.