Skip to content

Exterior Derivative

The exterior derivative is the coordinate-independent operation that differentiates differential forms:

d:Ωk(M)→Ωk+1(M).d:\Omega^k(M)\to\Omega^{k+1}(M).

It sends functions to differentials, one-forms to curl-like two-forms, two-forms to divergence-like three-forms, and so on. Its defining structural identity is

d2=0.d^2=0.

This page explains the coordinate formula, why applying dd twice gives zero, and how the familiar gradient, curl, and divergence operations fit into the same pattern.

On a smooth manifold MM, a kk-form is a smooth section of ΛkT∗M\Lambda^kT^*M. The exterior derivative raises degree by one:

Ω0(M)→ d Ω1(M)→ d Ω2(M)→ d ⋯fdfd(df)\begin{array}{cccccc} \Omega^0(M)&\xrightarrow{\ d\ }&\Omega^1(M)&\xrightarrow{\ d\ }&\Omega^2(M)&\xrightarrow{\ d\ }\cdots\\ f&&df&&d(df)& \end{array}

For a scalar function ff, the exterior derivative is the ordinary differential:

df=∂f∂xidxi.df = \frac{\partial f}{\partial x^i}dx^i.

For a 1-form

α=αidxi,\alpha=\alpha_i dx^i,

the exterior derivative is a 2-form:

dα=∂αi∂xjdxj∧dxi.d\alpha = \frac{\partial \alpha_i}{\partial x^j} dx^j\wedge dx^i.

After collecting antisymmetric terms, this is equivalently

dα=∑i<j(∂αj∂xi−∂αi∂xj)dxi∧dxj.d\alpha = \sum_{i<j} \left( \frac{\partial\alpha_j}{\partial x^i} - \frac{\partial\alpha_i}{\partial x^j} \right) dx^i\wedge dx^j.

The wedge product is what makes this derivative coordinate-independent rather than merely a list of component derivatives.

Let

ω=1k!ωi1⋯ik(x) dxi1∧⋯∧dxik\omega = \frac{1}{k!} \omega_{i_1\cdots i_k}(x)\, dx^{i_1}\wedge\cdots\wedge dx^{i_k}

be a kk-form. Then

dω=1k!∂ωi1⋯ik∂xjdxj∧dxi1∧⋯∧dxik.d\omega = \frac{1}{k!} \frac{\partial\omega_{i_1\cdots i_k}}{\partial x^j} dx^j\wedge dx^{i_1}\wedge\cdots\wedge dx^{i_k}.

In words: differentiate the coefficient functions and wedge the new differential onto the front. Antisymmetry then handles the signs and repeated-coordinate cancellations.

This formula is local, but the operation dd is geometric. On chart overlaps, the coordinate formulas transform consistently.

The exterior derivative is linear:

d(aω+bη)=a dω+b dηd(a\omega+b\eta) = a\,d\omega+b\,d\eta

for constants a,ba,b and forms of the same degree.

It also obeys a graded product rule. If ω\omega is a kk-form, then

d(ω∧η)=dω∧η+(−1)kω∧dη.d(\omega\wedge\eta) = d\omega\wedge\eta + (-1)^k\omega\wedge d\eta.

The sign is not decoration. It records the fact that dd has degree 11 and must pass through a kk-form before differentiating the second factor.

For a function ff and a form η\eta, this reduces to

d(fη)=df∧η+f dη.d(f\eta) = df\wedge\eta+f\,d\eta.

The identity

d2=0d^2=0

means that applying the exterior derivative twice always gives zero:

d(dω)=0.d(d\omega)=0.

For a function ff, compute

d(df)=d(∂f∂xidxi).d(df) = d \left( \frac{\partial f}{\partial x^i}dx^i \right).

Using the coordinate rule,

d(df)=∂2f∂xj∂xidxj∧dxi.d(df) = \frac{\partial^2 f}{\partial x^j\partial x^i} dx^j\wedge dx^i.

The second derivative coefficient is symmetric in ii and jj for smooth ff, while dxj∧dxidx^j\wedge dx^i is antisymmetric. The contraction of a symmetric coefficient with an antisymmetric basis vanishes, so

d(df)=0.d(df)=0.

The same symmetry-versus-antisymmetry cancellation proves d2=0d^2=0 for all forms.

In three-dimensional Euclidean vector calculus, the exterior derivative packages the sequence

gradient→curl→divergence\text{gradient} \quad\to\quad \text{curl} \quad\to\quad \text{divergence}

into one operation:

Ω0→ d Ω1→ d Ω2→ d Ω3.\Omega^0 \xrightarrow{\ d\ } \Omega^1 \xrightarrow{\ d\ } \Omega^2 \xrightarrow{\ d\ } \Omega^3.

The translation into vector fields uses the Euclidean metric and orientation. The exterior derivative itself does not require them.

Gradient as Exterior Derivative of a Function

Section titled “Gradient as Exterior Derivative of a Function”

For a scalar function f(x,y,z)f(x,y,z),

df=∂f∂xdx+∂f∂ydy+∂f∂zdz.df = \frac{\partial f}{\partial x}dx + \frac{\partial f}{\partial y}dy + \frac{\partial f}{\partial z}dz.

This is a 1-form. In Euclidean space, the metric identifies it with the gradient vector

∇f=(∂f∂x,∂f∂y,∂f∂z).\nabla f = \left( \frac{\partial f}{\partial x}, \frac{\partial f}{\partial y}, \frac{\partial f}{\partial z} \right).

Without a metric, dfdf still exists, but the gradient vector does not have a canonical meaning.

The identity d2=0d^2=0 becomes the familiar statement that the curl of a gradient vanishes:

d(df)=0↔∇×∇f=0.d(df)=0 \quad \leftrightarrow \quad \nabla\times\nabla f=\mathbf0.

The vector statement is a metric-dependent translation of the form statement.

Let

A=Ax dx+Ay dy+Az dz.A = A_x\,dx+A_y\,dy+A_z\,dz.

Then

dA=(∂Ay∂x−∂Ax∂y)dx∧dy+(∂Az∂y−∂Ay∂z)dy∧dz+(∂Ax∂z−∂Az∂x)dz∧dx.\begin{aligned} dA &= \left( \frac{\partial A_y}{\partial x} - \frac{\partial A_x}{\partial y} \right) dx\wedge dy\\ &\quad+ \left( \frac{\partial A_z}{\partial y} - \frac{\partial A_y}{\partial z} \right) dy\wedge dz\\ &\quad+ \left( \frac{\partial A_x}{\partial z} - \frac{\partial A_z}{\partial x} \right) dz\wedge dx. \end{aligned}

With the usual Euclidean identification of 2-forms and pseudovectors, this corresponds to

∇×A.\nabla\times\mathbf A.

For electromagnetic notation, AA may be read as a vector-potential 1-form and dAdA as a magnetic-flux 2-form.

The identity d2=0d^2=0 then gives

d(dA)=0,d(dA)=0,

which corresponds to the vector-calculus identity

∇⋅(∇×A)=0.\nabla\cdot(\nabla\times\mathbf A)=0.

Divergence as Exterior Derivative of a Two-Form

Section titled “Divergence as Exterior Derivative of a Two-Form”

In oriented Euclidean three-space, a vector field B=(Bx,By,Bz)\mathbf B=(B_x,B_y,B_z) can be represented by the flux 2-form

B=Bx dy∧dz+By dz∧dx+Bz dx∧dy.\mathcal B = B_x\,dy\wedge dz + B_y\,dz\wedge dx + B_z\,dx\wedge dy.

Then

dB=(∂Bx∂x+∂By∂y+∂Bz∂z)dx∧dy∧dz.d\mathcal B = \left( \frac{\partial B_x}{\partial x} + \frac{\partial B_y}{\partial y} + \frac{\partial B_z}{\partial z} \right) dx\wedge dy\wedge dz.

Thus dBd\mathcal B corresponds to the divergence of B\mathbf B times the oriented volume form:

dB↔(∇⋅B) dV.d\mathcal B \quad \leftrightarrow \quad (\nabla\cdot\mathbf B)\,dV.

Again, the exterior derivative gives the invariant form statement. The vector-divergence interpretation uses Euclidean structure.

In a local gauge, the Berry connection is a 1-form on parameter space:

An=Ai(n)(R) dRi.A_n = A_i^{(n)}(R)\,dR^i.

Its exterior derivative is the Berry curvature 2-form:

Fn=dAn.F_n=dA_n.

In coordinates,

Fn=12Fij(n) dRi∧dRj,F_n = \frac12 F_{ij}^{(n)}\,dR^i\wedge dR^j,

where

Fij(n)=∂Aj(n)∂Ri−∂Ai(n)∂Rj.F_{ij}^{(n)} = \frac{\partial A_j^{(n)}}{\partial R^i} - \frac{\partial A_i^{(n)}}{\partial R^j}.

Under a gauge change of the local eigenvector,

An↦An−dχ.A_n\mapsto A_n-d\chi.

The curvature is unchanged:

Fn↦d(An−dχ)=dAn−d2χ=Fn.F_n \mapsto d(A_n-d\chi) = dA_n-d^2\chi = F_n.

The last equality is exactly d2=0d^2=0. The mathematical Berry-specific connection and curvature formulas are developed in Berry Connection as a Mathematical Object. The physical adiabatic phase and gauge convention are treated in Berry Phase. The broader connection language is the subject of Connections and Curvature.

A form ω\omega is closed if

dω=0.d\omega=0.

It is exact if there is a form η\eta such that

ω=dη.\omega=d\eta.

Every exact form is closed because

dω=d(dη)=0.d\omega=d(d\eta)=0.

The converse is locally true under suitable hypotheses but not globally true on every manifold. Global failures of “closed implies exact” are one way topology enters physics. For example, a locally flat connection can still have nontrivial holonomy around a noncontractible loop.

This is a warning, not the full topological theory. Homotopy and Winding owns the first loop-based examples, and Chern Numbers owns the first curvature-integral invariant.

The exterior derivative is the operator that appears in the general Stokes theorem:

∫∂Σω=∫Σdω.\int_{\partial\Sigma}\omega = \int_\Sigma d\omega.

This single formula contains the fundamental theorem of calculus, Green’s theorem, the Kelvin-Stokes curl theorem, and the divergence theorem as special cases. The integration and orientation details belong to Integration on Manifolds.

For this page, the key idea is simpler: dd turns the integrand on a boundary into the integrand on the region it bounds.

  • Forgetting that dd raises form degree by one.
  • Treating dAdA as a vector curl before choosing a metric and orientation.
  • Dropping the sign in the graded product rule.
  • Thinking d2=0d^2=0 is a special property of scalar functions only.
  • Assuming that every closed form is globally exact.
  • Confusing the exterior derivative with an arbitrary componentwise derivative.
  • Using vector-calculus identities in curvilinear coordinates without checking the underlying form and metric.
  • Calling a Berry curvature gauge invariant without checking that the gauge transformation is the usual A↦A−dχA\mapsto A-d\chi form.
  • B. Schutz, Geometrical Methods of Mathematical Physics, Cambridge University Press, 1980.
  • T. Frankel, The Geometry of Physics, 3rd ed., Cambridge University Press, 2011.
  • J. M. Lee, Introduction to Smooth Manifolds, 2nd ed., Springer, 2013.
  • L. W. Tu, An Introduction to Manifolds, 2nd ed., Springer, 2011.
  • M. Nakahara, Geometry, Topology and Physics, 2nd ed., CRC Press, 2003.
  • M. Spivak, Calculus on Manifolds, Addison-Wesley, 1965.
  • A. Shapere and F. Wilczek, eds., Geometric Phases in Physics, World Scientific, 1989.
  1. For f(x,y)=x2yf(x,y)=x^2y, compute dfdf and verify that d(df)=0d(df)=0.
Solution

First,

df=2xy dx+x2 dy.df = 2xy\,dx+x^2\,dy.

Now apply dd:

d(df)=d(2xy)∧dx+d(x2)∧dy=(2y dx+2x dy)∧dx+2x dx∧dy=2x dy∧dx+2x dx∧dy=0.\begin{aligned} d(df) &= d(2xy)\wedge dx+d(x^2)\wedge dy\\ &= (2y\,dx+2x\,dy)\wedge dx + 2x\,dx\wedge dy\\ &= 2x\,dy\wedge dx+2x\,dx\wedge dy\\ &= 0. \end{aligned}
  1. Let α=P(x,y) dx+Q(x,y) dy\alpha=P(x,y)\,dx+Q(x,y)\,dy. Compute dαd\alpha.
Solution

Use the product rule and d(dx)=d(dy)=0d(dx)=d(dy)=0:

dα=dP∧dx+dQ∧dy.d\alpha = dP\wedge dx+dQ\wedge dy.

Since

dP=Px dx+Py dy,dQ=Qx dx+Qy dy,dP=P_x\,dx+P_y\,dy, \qquad dQ=Q_x\,dx+Q_y\,dy,

one gets

dα=Py dy∧dx+Qx dx∧dy=(Qx−Py) dx∧dy.d\alpha = P_y\,dy\wedge dx+Q_x\,dx\wedge dy = (Q_x-P_y)\,dx\wedge dy.
  1. Apply the previous result to α=x dy−y dx\alpha=x\,dy-y\,dx.
Solution

Here P=−yP=-y and Q=xQ=x. Therefore

Qx=1,Py=−1.Q_x=1, \qquad P_y=-1.

Thus

dα=(Qx−Py) dx∧dy=2 dx∧dy.d\alpha = (Q_x-P_y)\,dx\wedge dy = 2\,dx\wedge dy.
  1. Let
B=Bx dy∧dz+By dz∧dx+Bz dx∧dy.\mathcal B = B_x\,dy\wedge dz + B_y\,dz\wedge dx + B_z\,dx\wedge dy.

Compute dBd\mathcal B.

Solution

Only the derivative in the missing coordinate survives in each term, because repeated differentials wedge to zero:

dB=∂Bx∂xdx∧dy∧dz+∂By∂ydy∧dz∧dx+∂Bz∂zdz∧dx∧dy.\begin{aligned} d\mathcal B &= \frac{\partial B_x}{\partial x} dx\wedge dy\wedge dz\\ &\quad+ \frac{\partial B_y}{\partial y} dy\wedge dz\wedge dx\\ &\quad+ \frac{\partial B_z}{\partial z} dz\wedge dx\wedge dy. \end{aligned}

The last two triple wedges are cyclic permutations of dx∧dy∧dzdx\wedge dy\wedge dz, so

dB=(∂Bx∂x+∂By∂y+∂Bz∂z)dx∧dy∧dz.d\mathcal B = \left( \frac{\partial B_x}{\partial x} + \frac{\partial B_y}{\partial y} + \frac{\partial B_z}{\partial z} \right) dx\wedge dy\wedge dz.
  1. If A↦A−dχA\mapsto A-d\chi, show that F=dAF=dA is unchanged.
Solution

The transformed curvature is

F′=d(A−dχ)=dA−d2χ.F' = d(A-d\chi) = dA-d^2\chi.

Since d2=0d^2=0,

F′=dA=F.F'=dA=F.
  1. Why does dω=0d\omega=0 not always imply ω=dη\omega=d\eta globally?
Solution

The equation dω=0d\omega=0 is local differential information. The equation ω=dη\omega=d\eta asks for a globally defined potential η\eta. On manifolds with nontrivial topology, local potentials may fail to patch together globally. Thus every exact form is closed, but a closed form need not be globally exact.