The Del Operator (Nabla)

Maxwell's equations, the wave equation, and nearly every field relation in electromagnetism are written with a single symbol: the inverted Greek delta $\nabla$ , called del or nabla. It is a compact bookkeeping device that packages the three first-order spatial derivatives into a vector. Introduced as a formal vector operator, $\nabla$ lets the gradient, divergence, and curl — three distinct operations — be written and manipulated with the familiar algebra of dot and cross products. The power of the notation is that $\nabla$ behaves like an ordinary vector in many manipulations, so vector identities can be guessed and checked by treating $\nabla$ as if it were a vector $\mathbf{V}$ — provided one always remembers two things it does not share with an ordinary vector. First, it does not commute with the fields it differentiates: $\nabla f \ne f\nabla$ . Second, it is meaningless until it is given an operand. These two caveats are the source of nearly every sign error and every subtlety in vector-calculus identities. From $\nabla$ flow the three first-order operations and the second-order ones built by applying $\nabla$ twice — most importantly the Laplacian $\nabla^2$ , which governs electrostatic potentials, diffusion, and waves. Two of the five possible second derivatives vanish identically, and these vanishing identities underlie the existence of the scalar and vector potentials in electromagnetism. Finally, $\nabla$ is the differential half of the great integral theorems — the fundamental theorem for gradients, the divergence theorem, and Stokes's theorem each integrate one of the three $\nabla$ operations. Understanding $\nabla$ as one unified object is the gateway to the entire formalism of field theory.

Key takeaways

The del operator $\nabla$ is a vector whose components are the partial-derivative operators $\partial/\partial x,\ \partial/\partial y,\ \partial/\partial z$ .
Acting on a scalar field it produces the gradient $\nabla f$ ; dotted into a vector field it produces the divergence $\nabla\cdot\mathbf{A}$ ; crossed with a vector field it produces the curl $\nabla\times\mathbf{A}$ .
$\nabla$ is a true vector under rotations, but it is an operator, not a number: order matters and it must always be told what it acts on.
Second derivatives built from $\nabla$ give the Laplacian $\nabla^2$ and identities such as $\nabla\times(\nabla f)=0$ and $\nabla\cdot(\nabla\times\mathbf{A})=0$ .
The three integral theorems of vector calculus (gradient, divergence, Stokes) are all statements about integrating the three $\nabla$ operations.
In curvilinear coordinates $\nabla$ acquires metric factors — it is not simply $(\partial_r,\partial_\theta,\partial_\phi)$ .
Key idea: $\nabla \equiv \hat{\mathbf{x}}\,\dfrac{\partial}{\partial x}+\hat{\mathbf{y}}\,\dfrac{\partial}{\partial y}+\hat{\mathbf{z}}\,\dfrac{\partial}{\partial z}$ is a vector operator: multiply it onto a scalar ( $\nabla f$ ), dot it into a vector ( $\nabla\cdot\mathbf{A}$ ), or cross it into a vector ( $\nabla\times\mathbf{A}$ ) to get the three derivatives of vector calculus.

Where this lives today

Del is arguably the most-typed symbol in computational physics. Every finite-element electromagnetic solver — the software that designs phone antennas, MRI coils and particle-accelerator cavities — discretizes exactly the three del operations onto meshes, and the two vanishing identities of equation (5) are enforced structurally in modern "mimetic" and finite-element-exterior-calculus schemes: the discrete curl of the discrete gradient is exactly zero in the code, not just approximately, because otherwise spurious numerical charges and monopoles appear. Machine learning borrowed the symbol outright: the training signal for a neural network is written $\nabla_\theta L$ , the gradient of the loss with respect to millions of parameters, computed by automatic differentiation. And in plasma physics and astrophysics, the operator's grammar organizes whole research fields — dynamo theory is the study of how $\nabla\times(\mathbf{v}\times\mathbf{B})$ can outrun magnetic diffusion $\eta\nabla^2\mathbf{B}$ in the induction equation. Fluency in del is transferable across all of it.

The physical picture

Picture $\nabla$ as a 'measuring stick with three prongs', one along each axis. Each prong asks: how fast is the field changing in this direction? What you do with the answers — collect them into a vector (gradient), add them up (divergence), or twist them (curl) — determines the operation. One symbol, three grammatical sentences.

Figure

The del machine — one operator, three products, three different questions about a field

Because $\nabla$ transforms like a genuine vector under rotation of the coordinate axes — a fact we prove below — the gradient is a true vector, the divergence is a true scalar, and the curl is a true (pseudo)vector. The notation is not just shorthand: it encodes the correct geometric transformation law, which is why equations written with $\nabla$ hold in every frame. Think of $\nabla$ as a verb missing its object. 'Differentiate' is meaningless until you say what and how. $\nabla f$ , $\nabla\cdot\mathbf{A}$ , and $\nabla\times\mathbf{A}$ are the three grammatical sentences you can form. When checking a vector identity, pretend $\nabla$ is a vector to guess the answer, then go back and verify that each $\nabla$ differentiates exactly the right factor — the product rule, not just the algebra, decides the result. The electromagnetic payoff is total: all four Maxwell equations are sentences in this grammar. Two divergences ( $\nabla\cdot\mathbf{E}=\rho/\varepsilon_0$ , $\nabla\cdot\mathbf{B}=0$ ) and two curls ( $\nabla\times\mathbf{E}=-\partial_t\mathbf{B}$ , $\nabla\times\mathbf{B}=\mu_0\mathbf{J}+\mu_0\varepsilon_0\partial_t\mathbf{E}$ ). Learn to speak $\nabla$ fluently and you can read the whole theory off a t-shirt.

The operator itself

In Cartesian coordinates, del is defined as the vector whose components are the three partial-derivative operators:

\nabla \equiv \hat{\mathbf{x}}\,\frac{\partial}{\partial x}+\hat{\mathbf{y}}\,\frac{\partial}{\partial y}+\hat{\mathbf{z}}\,\frac{\partial}{\partial z} \qquad (1)

where: $\nabla$ — del (nabla), the vector differential operator; $\hat{\mathbf{x}},\hat{\mathbf{y}},\hat{\mathbf{z}}$ — orthonormal Cartesian basis vectors; $\partial/\partial x$ — partial derivative holding $y,z$ fixed. On its own, equation (1) is an unfinished sentence — a stack of instructions with nothing yet to act on. Its whole meaning comes from the three ways of attaching an operand:

\nabla f,\qquad \nabla\cdot\mathbf{A},\qquad \nabla\times\mathbf{A} \qquad (2)

where: $f$ — scalar field; $\mathbf{A}$ — vector field with components $(A_x,A_y,A_z)$ . These mimic the three ways an ordinary vector $\mathbf{V}$ can multiply: $\mathbf{V}f$ (scaling), $\mathbf{V}\cdot\mathbf{A}$ (dot), $\mathbf{V}\times\mathbf{A}$ (cross). The mimicry is exact enough that each operation inherits its component formula from vector algebra — the gradient collects $\partial_i f$ , the divergence sums $\partial_i A_i$ , and the curl is most easily remembered as a formal determinant:

\nabla\times\mathbf{A}=\begin{vmatrix}\hat{\mathbf{x}}&\hat{\mathbf{y}}&\hat{\mathbf{z}}\\[2pt]\partial_x&\partial_y&\partial_z\\[2pt]A_x&A_y&A_z\end{vmatrix} \qquad (3)

where: $\partial_x$ — shorthand for $\partial/\partial x$ ; the determinant is expanded along the top row, each $\partial_i$ acting on the $A_k$ in its cofactor. The trick works because the honest cross product $\mathbf{V}\times\mathbf{A}$ has the identical determinant form with $V_i$ in place of $\partial_i$ ; promoting constants to derivative operators, the $x$ -component reads $(\nabla\times\mathbf{A})_x = \partial_y A_z - \partial_z A_y$ , and cyclic permutation gives the rest. This determinant is a Cartesian-only mnemonic — in spherical or cylindrical coordinates it fails, because there the basis vectors themselves have derivatives.

Applying del twice

Since $\nabla f$ is a vector field and $\nabla\cdot\mathbf{A}$ a scalar field, del can act on its own outputs. Five second derivatives exist; the most important is the divergence of a gradient, the Laplacian:

\nabla^2 \equiv \nabla\cdot\nabla=\frac{\partial^2}{\partial x^2}+\frac{\partial^2}{\partial y^2}+\frac{\partial^2}{\partial z^2} \qquad (4)

where: $\nabla^2$ — Laplacian operator (also written $\Delta$ ), the subject of its own article. In charge-free space the potential obeys $\nabla^2 V = 0$ ; with sources, $\nabla^2 V = -\rho/\varepsilon_0$ — the equations that the entire boundary-value-problem chapter solves. Two of the five second derivatives are identically zero, and these two facts quietly run electromagnetism:

\nabla\times(\nabla f)=\mathbf{0},\qquad \nabla\cdot(\nabla\times\mathbf{A})=0 \qquad (5)

where: $\mathbf{0}$ — the zero vector. Both follow from the equality of mixed partial derivatives: the $z$ -component of $\nabla\times(\nabla f)$ is $\partial_x\partial_y f-\partial_y\partial_x f=0$ by Clairaut's theorem, and in $\nabla\cdot(\nabla\times\mathbf{A})$ each mixed derivative of each component appears twice with opposite signs, cancelling in pairs. Read physically: gradients never circulate (so a curl-free $\mathbf{E}$ can be written as $-\nabla V$ ), and curls never diverge (so a divergence-free $\mathbf{B}$ can be written as $\nabla\times\mathbf{A}$ ). The scalar and vector potentials of electromagnetism exist because of equation (5).

Figure

The five second derivatives of del — two survivors, two identical zeros, one hybrid

Because del differentiates, its products obey Leibniz rules rather than plain algebra. The two used constantly in this track are the gradient and divergence product rules:

\nabla(fg)=f\,\nabla g+g\,\nabla f \qquad (6)

where: $f,g$ — scalar fields; the familiar product rule, componentwise.

\nabla\cdot(f\mathbf{A})=f(\nabla\cdot\mathbf{A})+\mathbf{A}\cdot(\nabla f) \qquad (7)

where: $f$ — scalar field; $\mathbf{A}$ — vector field. Each rule says: let del differentiate one factor at a time and add the results. Rule (7) is the workhorse of Poynting-theorem and energy-density manipulations later in the track — worth committing to memory now. Finally, in curvilinear coordinates the operator dresses itself with metric scale factors, because a step in an angular coordinate is not a step in distance:

\nabla = \hat{\mathbf{r}}\,\frac{\partial}{\partial r}+\hat{\boldsymbol{\theta}}\,\frac{1}{r}\frac{\partial}{\partial\theta}+\hat{\boldsymbol{\phi}}\,\frac{1}{r\sin\theta}\frac{\partial}{\partial\phi} \qquad (8)

where: $r,\theta,\phi$ — radius, polar angle, azimuth; $\hat{\mathbf{r}},\hat{\boldsymbol{\theta}},\hat{\boldsymbol{\phi}}$ — unit vectors that change direction from point to point; $1/r$ , $1/(r\sin\theta)$ — scale factors converting per-radian rates into per-metre rates. The position-dependent basis vectors are why divergence and curl in spherical coordinates are not obtained by naively dotting or crossing (8) into a field — the derivatives also act on the unit vectors. Use the tabulated formulas (or the derivation below) instead.

Derivation ADel is a genuine vector under rotation

We are proving that the three components of $\nabla$ transform under a rotation of axes exactly like the components of an ordinary vector — the fact that licenses the name "vector operator" and guarantees that $\nabla f$ , $\nabla\cdot\mathbf{A}$ and $\nabla\times\mathbf{A}$ are a true vector, scalar, and (pseudo)vector respectively. The strategy is the chain rule: a derivative with respect to a rotated coordinate is a weighted sum of derivatives with respect to the old ones, and the weights turn out to be precisely the rotation-matrix entries. The step to watch is inverting the coordinate relation — it is the orthogonality of $R$ that makes the inverse equal the transpose and lands the indices in the right order.

Step 1. Rotate the axes, so the new coordinates are $x_i'=\sum_j R_{ij}x_j$ with $R$ orthogonal. By the chain rule, for any smooth function,

\frac{\partial}{\partial x_i'}=\sum_j \frac{\partial x_j}{\partial x_i'}\,\frac{\partial}{\partial x_j}

Why: differentiating in a new variable routes through all the old variables it depends on — this is the multivariable chain rule with nothing assumed about $R$ yet.

Step 2. Invert the coordinate relation to find the weights. Orthogonality gives $R^{-1}=R^{\mathsf T}$ , so $x_j=\sum_i R_{ij}\,x_i'$ and

\frac{\partial x_j}{\partial x_i'}=R_{ij}

Why: solving $x'=Rx$ for $x$ requires $R^{-1}$ ; for a rotation the inverse is the transpose, which is what places $R_{ij}$ (not $R_{ji}$ ) in the derivative.

Step 3. Substitute back:

\frac{\partial}{\partial x_i'}=\sum_j R_{ij}\,\frac{\partial}{\partial x_j}

This is exactly the transformation law $V_i'=\sum_j R_{ij}V_j$ obeyed by the components of an ordinary vector. The derivative triple rotates like an arrow, so everything built from it inherits honest tensorial character: gradient (vector), divergence (scalar), curl (pseudovector). Verified: matches the transformation argument of Griffiths §1.2.2; combined with the invariance proofs in the scalar-and-vector-fields article it confirms $\nabla\cdot\mathbf{A}$ is rotation-invariant.

Go deeper · Derivation B · 1 minThe Laplacian in spherical coordinates

Derivation BThe Laplacian in spherical coordinates

We are computing $\nabla^2 f=\nabla\cdot(\nabla f)$ in spherical coordinates, metric factors included — the formula that later collapses Laplace's equation for a point charge to a one-line ODE. The strategy: take the spherical gradient components (from the gradient article), feed them into the general curvilinear divergence formula, and simplify. The danger is exactly the warning above: you cannot just dot (8) into (8), because the unit vectors $\hat{\boldsymbol\theta}, \hat{\boldsymbol\phi}$ have position-dependent directions; the curvilinear divergence formula with the Jacobian factor $r^2\sin\theta$ handles their derivatives for us.

Step 1. Gradient components in spherical coordinates:

(\nabla f)_r=\partial_r f,\qquad (\nabla f)_\theta=\tfrac{1}{r}\partial_\theta f,\qquad (\nabla f)_\phi=\tfrac{1}{r\sin\theta}\partial_\phi f

Why: each physical component is the coordinate derivative divided by its scale factor $h_r=1$ , $h_\theta=r$ , $h_\phi=r\sin\theta$ , as forced by $\mathrm{d}T=\nabla T\cdot\mathrm{d}\mathbf{l}$ .

Step 2. Divergence of a general field in orthogonal curvilinear coordinates:

\nabla\cdot\mathbf{A}=\frac{1}{r^2\sin\theta}\Big[\partial_r(r^2\sin\theta\,A_r)+\partial_\theta(r\sin\theta\,A_\theta)+\partial_\phi(r\,A_\phi)\Big]

Why: the prefactor is the inverse Jacobian $1/(h_rh_\theta h_\phi)$ and each term carries the product of the other two scale factors — this bookkeeping is what correctly differentiates the moving basis vectors.

Step 3. Insert the gradient components and simplify term by term:

\nabla^2 f=\frac{1}{r^2}\frac{\partial}{\partial r}\!\left(r^2\frac{\partial f}{\partial r}\right)+\frac{1}{r^2\sin\theta}\frac{\partial}{\partial\theta}\!\left(\sin\theta\frac{\partial f}{\partial\theta}\right)+\frac{1}{r^2\sin^2\theta}\frac{\partial^2 f}{\partial\phi^2}

The $\sin\theta$ cancels from the radial term, one factor of $r$ from the polar term, leaving the standard spherical Laplacian. Verified: agrees with the formula inside the cover of Griffiths and in Jackson's appendix; for $f=1/r$ the radial term gives $\partial_r(r^2\cdot(-1/r^2))=0$ away from the origin, consistent with the Coulomb potential solving Laplace's equation.

Go deeper · Derivation C · 1 minThe curl of a curl via the BAC-CAB template

Derivation CThe curl of a curl via the BAC-CAB template

We are deriving $\nabla\times(\nabla\times\mathbf{A})=\nabla(\nabla\cdot\mathbf{A})-\nabla^2\mathbf{A}$ — the identity that, applied to Maxwell's equations, produces the electromagnetic wave equation and with it light itself. The strategy: use the algebraic triple-product rule BAC-CAB as a template to guess the form, then confirm with index notation that each derivative really acts on $\mathbf{A}$ (both dels differentiate the field, so no product-rule subtleties arise, unlike identities with two different fields). The step to watch is the epsilon-contraction identity — it is where the guessed structure is actually proved.

Step 1. Algebraic template with two equal vectors:

\mathbf{V}\times(\mathbf{V}\times\mathbf{A})=\mathbf{V}(\mathbf{V}\cdot\mathbf{A})-(\mathbf{V}\cdot\mathbf{V})\mathbf{A}

Why: this is BAC-CAB, $\mathbf{a}\times(\mathbf{b}\times\mathbf{c})=\mathbf{b}(\mathbf{a}\cdot\mathbf{c})-\mathbf{c}(\mathbf{a}\cdot\mathbf{b})$ , with $\mathbf{a}=\mathbf{b}=\mathbf{V}$ — it suggests, but does not yet prove, the operator identity.

Step 2. Prove it in index form using $\varepsilon_{ijk}\varepsilon_{klm}=\delta_{il}\delta_{jm}-\delta_{im}\delta_{jl}$ :

[\nabla\times(\nabla\times\mathbf{A})]_i=\varepsilon_{ijk}\,\partial_j\,\varepsilon_{klm}\partial_l A_m = \partial_i(\partial_m A_m)-\partial_j\partial_j A_i

Why: both derivatives act on $\mathbf{A}$ and commute with each other, so the two delta-terms give exactly the divergence-gradient piece and the componentwise Laplacian piece.

Step 3. Reassemble in vector notation:

\nabla\times(\nabla\times\mathbf{A})=\nabla(\nabla\cdot\mathbf{A})-\nabla^2\mathbf{A}

with $\nabla^2\mathbf{A}$ meaning the Laplacian applied to each Cartesian component — a definition that is safe only in Cartesian coordinates; in curvilinear systems this identity is taken as the definition of the vector Laplacian. Verified: identity printed inside the covers of Griffiths and Jackson; applied to $\nabla\times(\nabla\times\mathbf{E})$ with Maxwell's vacuum equations it yields $\nabla^2\mathbf{E}=\mu_0\varepsilon_0\,\partial_t^2\mathbf{E}$ , waves at $c=1/\sqrt{\mu_0\varepsilon_0}$ .

Limiting cases

The single symbol $\nabla$ specializes into different operations depending on how it meets its operand:

Case	Result	Meaning
On a scalar field	$\nabla f = \hat{\mathbf{x}}\,\partial_x f+\hat{\mathbf{y}}\,\partial_y f+\hat{\mathbf{z}}\,\partial_z f$	Gradient: a vector pointing toward steepest increase of $f$ .
Dotted into a vector	$\nabla\cdot\mathbf{A}=\partial_x A_x+\partial_y A_y+\partial_z A_z$	Divergence: a scalar measuring net outflow per unit volume.
Crossed into a vector	$\nabla\times\mathbf{A}=\sum_{i,j,k}\varepsilon_{ijk}\,\hat{\mathbf{e}}_i\,\partial_j A_k$	Curl: a vector measuring local circulation.
Applied twice (scalar)	$\nabla\cdot(\nabla f)=\nabla^2 f$	Laplacian: governs potentials, diffusion, and waves.
Dotted from the left	$\mathbf{A}\cdot\nabla = A_x\partial_x + A_y\partial_y + A_z\partial_z$	Still an operator: the directional derivative along $\mathbf{A}$ , central to the convective derivative of fluid dynamics.

Worked examples

Example 1 — all three operations on concrete fields (confidence builder). Let $f = x^2yz$ and $\mathbf{A} = xz\,\hat{\mathbf{x}} - y^2\,\hat{\mathbf{y}} + 2xy\,\hat{\mathbf{z}}$ . Evaluate $\nabla f$ , $\nabla\cdot\mathbf{A}$ , and $\nabla\times\mathbf{A}$ at the point $P = (1, 2, 3)$ .

Step 1. Gradient of $f$ : the partials are $\partial_x f = 2xyz$ , $\partial_y f = x^2z$ , $\partial_z f = x^2y$ . At $P$ :

\nabla f\big|_P = (2)(1)(2)(3)\,\hat{\mathbf{x}} + (1)(3)\,\hat{\mathbf{y}} + (1)(2)\,\hat{\mathbf{z}} = 12\,\hat{\mathbf{x}} + 3\,\hat{\mathbf{y}} + 2\,\hat{\mathbf{z}}

Step 2. Divergence of $\mathbf{A}$ : $\partial_x(xz) = z$ , $\partial_y(-y^2) = -2y$ , $\partial_z(2xy) = 0$ , so $\nabla\cdot\mathbf{A} = z - 2y$ . At $P$ :

\nabla\cdot\mathbf{A}\big|_P = 3 - 4 = -1

a mild local sink.

Step 3. Curl of $\mathbf{A}$ from the determinant: the components are $(\partial_y(2xy) - \partial_z(-y^2),\ \partial_z(xz) - \partial_x(2xy),\ \partial_x(-y^2) - \partial_y(xz)) = (2x,\ x - 2y,\ 0)$ . At $P$ :

\nabla\times\mathbf{A}\big|_P = 2\,\hat{\mathbf{x}} - 3\,\hat{\mathbf{y}} + 0\,\hat{\mathbf{z}}

Step 4. Free consistency check of the vanishing identity: $\nabla\cdot(\nabla\times\mathbf{A}) = \partial_x(2x) + \partial_y(x-2y) + \partial_z(0) = 2 - 2 + 0 = \mathbf{0}$ — identically zero at every point, as equation (5) demands.

Example 2 — verifying curl-of-curl on a wave-like field (exam level). Let $\mathbf{A} = \sin(kx)\,\hat{\mathbf{z}}$ with $k = 2\ \mathrm{m}^{-1}$ — a frozen snapshot of a wave. Verify $\nabla\times(\nabla\times\mathbf{A}) = \nabla(\nabla\cdot\mathbf{A}) - \nabla^2\mathbf{A}$ explicitly, and evaluate at $x = \pi/4\ \mathrm{m}$ .

Step 1. First curl: for $\mathbf{A} = (0, 0, A_z(x))$ the determinant gives $\nabla\times\mathbf{A} = (\partial_y A_z)\,\hat{\mathbf{x}} - (\partial_x A_z)\,\hat{\mathbf{y}} = -k\cos(kx)\,\hat{\mathbf{y}}$ .

Step 2. Second curl: for $(0, A_y(x), 0)$ the only surviving component is $\hat{\mathbf{z}}\,\partial_x A_y$ :

\nabla\times(\nabla\times\mathbf{A}) = \partial_x\big(-k\cos kx\big)\,\hat{\mathbf{z}} = k^2\sin(kx)\,\hat{\mathbf{z}}

Step 3. Right-hand side: $\nabla\cdot\mathbf{A} = \partial_z\sin(kx) = 0$ , so the first term dies; and $\nabla^2\mathbf{A} = \partial_x^2\sin(kx)\,\hat{\mathbf{z}} = -k^2\sin(kx)\,\hat{\mathbf{z}}$ . Hence

\nabla(\nabla\cdot\mathbf{A}) - \nabla^2\mathbf{A} = 0 + k^2\sin(kx)\,\hat{\mathbf{z}}

— the two sides agree exactly.

Step 4. Numbers: at $x = \pi/4$ with $k = 2$ : $kx = \pi/2$ , $\sin(kx) = 1$ , so both sides equal $k^2 = 4\ \mathrm{m}^{-2}$ (times $\hat{\mathbf{z}}$ , in the field's units). The curl-of-curl of this transverse field is $4\,\hat{\mathbf{z}}$ — and the same algebra, run on $\mathbf{E}$ instead of $\mathbf{A}$ , is precisely how Maxwell's equations turn into the wave equation.

Applications

Maxwell's equations in compact form. Every Maxwell equation is one of the three $\nabla$ operations: $\nabla\cdot\mathbf{E}=\rho/\varepsilon_0$ and $\nabla\cdot\mathbf{B}=0$ use the divergence, while $\nabla\times\mathbf{E}=-\partial_t\mathbf{B}$ and $\nabla\times\mathbf{B}=\mu_0\mathbf{J}+\mu_0\varepsilon_0\partial_t\mathbf{E}$ use the curl. The single symbol $\nabla$ thus expresses all of electromagnetism's local field laws.

Existence of potentials. Because $\nabla\times(\nabla f)=0$ , the curl-free electrostatic field can be written $\mathbf{E}=-\nabla V$ . Because $\nabla\cdot(\nabla\times\mathbf{A})=0$ , the divergence-free magnetic field can be written $\mathbf{B}=\nabla\times\mathbf{A}$ . The two vanishing identities of del are precisely what guarantee the scalar and vector potentials exist.

The wave equation. Taking the curl of Faraday's law and using the curl-of-curl identity $\nabla\times(\nabla\times\mathbf{E})=\nabla(\nabla\cdot\mathbf{E})-\nabla^2\mathbf{E}$ yields, in vacuum, $\nabla^2\mathbf{E}=\mu_0\varepsilon_0\,\partial_t^2\mathbf{E}$ — electromagnetic waves traveling at speed $c=1/\sqrt{\mu_0\varepsilon_0}$ .

Coordinate-adapted problems. For spherically symmetric charge distributions, writing $\nabla$ in spherical coordinates collapses Laplace's equation $\nabla^2 V=0$ to an ordinary differential equation in $r$ , giving the $1/r$ Coulomb potential. The metric factors in $\nabla$ are essential here.

Origins · 1 minWilliam Rowan Hamilton and how this was discovered

William Rowan Hamilton

1805–1865 · Trinity College Dublin and Dunsink Observatory

Hamilton read Hebrew, Persian and Sanskrit as a child, was appointed Royal Astronomer of Ireland while still an undergraduate, and spent his last two decades obsessed with a single algebraic idea. On 16 October 1843, walking along Dublin's Royal Canal, he saw in a flash how to multiply triples of numbers by using quadruples instead — and carved the quaternion relations $i^2=j^2=k^2=ijk=-1$ into the stone of Broome Bridge. Inside quaternion algebra lurked a differential operator he wrote with a rotated delta: applied to a function of position it packaged all three space derivatives at once. Hamilton himself barely glimpsed what it would become; it was Tait who championed the operator, whimsically nicknamed nabla after an Assyrian harp of the same shape, and Maxwell who deployed it on electromagnetic fields. When Gibbs and Heaviside later broke quaternions apart into the modern dot and cross products, the operator survived the demolition intact — the sharpest tool salvaged from Hamilton's canal-side epiphany.

History

1843

— Hamilton discovers quaternions at Broome Bridge, Dublin, embedding what will become the del operator in their algebra

1846

— Hamilton introduces the rotated-delta operator explicitly in his quaternion papers

1867

— Tait's Elementary Treatise on Quaternions develops the operator's calculus; the nickname nabla, after the harp, sticks

1873

— Maxwell's Treatise names the three operations physically — slope, convergence, curl — and uses them on the electromagnetic field

1881

— Gibbs (and Heaviside independently) strip away quaternions, leaving del with the dot and cross products as modern vector analysis

1892

— Heaviside recasts Maxwell's twenty component equations into the four compact del-equations used ever since

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §1.2 · Jackson, Classical Electrodynamics, 3rd ed., front-cover identities and §1 · Purcell & Morin, Electricity and Magnetism, 3rd ed., Ch. 2 · Zangwill, Modern Electrodynamics, §1.2

Provenance. Definitions and identities cross-checked against Griffiths Introduction to Electrodynamics 4th ed. §1.2 and inside-cover identity tables, Jackson 3rd ed., and Purcell 3rd ed. A verifier should: (1) confirm the Cartesian definition and the three first-order operations; (2) re-expand the curl determinant; (3) reproduce the spherical $\nabla$ and Laplacian against the standard cover formulas; (4) re-derive the curl-of-curl identity via $\varepsilon\varepsilon$ contraction; (5) verify the two vanishing second-derivative identities by equality of mixed partials. No empirical numbers are quoted except $c=1/\sqrt{\mu_0\varepsilon_0}$ , a standard relation.

Builds on: Scalar and Vector Fields · Partial derivatives · Dot and cross products Leads to: The Gradient Operator · The Divergence of a Vector Field · The Curl of a Vector Field · Second Derivatives and the Laplacian · The Divergence (Gauss) Theorem

Is del actually a vector, or is calling it one just a convenient abuse of language?

Both, in a precise sense. Its three components transform under rotations exactly like a vector's (the chain-rule derivation above), which is the physicist's definition — so gradient, divergence and curl have honest tensorial character. But it is a vector whose "components" are instructions rather than numbers, so vector algebra only partially applies: no commuting past its operand, no magnitude, no direction until it acts. The working rule: trust del's vector nature for transformation questions, distrust it for algebraic manipulations until every derivative has an assigned operand.

What is A·∇ good for if it is not the divergence?

It is the directional derivative operator scaled by the magnitude of A: $(\mathbf{A}\cdot\nabla)f = A_x\partial_x f + A_y\partial_y f + A_z\partial_z f$ measures how fast $f$ changes as you move along $\mathbf{A}$ . Its most famous job is in fluid dynamics: the acceleration of a fluid parcel is $\partial_t\mathbf{v} + (\mathbf{v}\cdot\nabla)\mathbf{v}$ , where the second term — the convective derivative — captures the change a parcel feels by being carried into a region where the field differs. It also appears in the E&M multipole expansion and in the force on a dipole, $(\mathbf{p}\cdot\nabla)\mathbf{E}$ .

Where does the name nabla come from, and should I say del or nabla?

"Nabla" is the Greek word for a Phoenician-Assyrian harp with roughly the symbol's triangular shape — the scholar William Robertson Smith suggested it to Tait, and it stuck as laboratory slang before becoming official. "Del" is the more common spoken form in American usage. They are perfectly synonymous; Griffiths says del, much of the European and mathematical literature says nabla, and LaTeX splits the difference by naming the command \nabla. Use either — just never confuse the operator with the finite-difference Delta.

Can I use the determinant mnemonic for the curl in cylindrical or spherical coordinates?

No — this is a classic trap. The determinant works in Cartesian coordinates because the basis vectors are constant, so the derivatives pass straight through to the components. In curvilinear systems $\hat{\mathbf{r}}, \hat{\boldsymbol\theta}, \hat{\boldsymbol\phi}$ rotate as you move, and the derivatives must act on them too. The correct curvilinear curl has scale factors woven through it (there is a modified determinant with $h$ -factors in each row, if you insist on a mnemonic). Safest practice at this stage: look up the tabulated spherical/cylindrical formulas, and reserve the bare determinant for Cartesian work.

What does the Laplacian of a vector field even mean?

In Cartesian coordinates, $\nabla^2\mathbf{A}$ means "apply the scalar Laplacian to each component": $(\nabla^2 A_x, \nabla^2 A_y, \nabla^2 A_z)$ . That definition silently fails in curvilinear coordinates, where the basis vectors have their own derivatives — the spherical " $\nabla^2 A_r$ " is not the $r$ -component of the vector Laplacian. The coordinate-safe definition inverts the curl-of-curl identity: $\nabla^2\mathbf{A} \equiv \nabla(\nabla\cdot\mathbf{A}) - \nabla\times(\nabla\times\mathbf{A})$ . When you meet the vector wave equation later in the track, that is the object being written.

How many product rules does del have, and do I need to memorize all of them?

There are six standard ones — two for gradients ( $\nabla(fg)$ and $\nabla(\mathbf{A}\cdot\mathbf{B})$ ), two for divergences ( $\nabla\cdot(f\mathbf{A})$ and $\nabla\cdot(\mathbf{A}\times\mathbf{B})$ ), two for curls ( $\nabla\times(f\mathbf{A})$ and $\nabla\times(\mathbf{A}\times\mathbf{B})$ ). Memorize the two simple scalar-times-field rules (equations 6 and 7) plus $\nabla\cdot(\mathbf{A}\times\mathbf{B}) = \mathbf{B}\cdot(\nabla\times\mathbf{A}) - \mathbf{A}\cdot(\nabla\times\mathbf{B})$ , which powers Poynting's theorem. The rest you can reconstruct on demand: write the expression in index notation, apply the ordinary product rule to each derivative, and reassemble — the derivations in this article model the technique.

If del is coordinate-dependent, how can the gradient, divergence and curl be geometric objects?

Because each operation has a coordinate-free definition that the component recipes merely implement: the gradient via $\mathrm{d}T = \nabla T\cdot\mathrm{d}\mathbf{l}$ , the divergence as flux per unit volume in the shrinking-volume limit, the curl as circulation per unit area. Those limits mention no axes at all. Equation (1) and its spherical cousin (8) are two different unpackings of the same geometry — which is why physics written in del-notation is automatically valid in every coordinate system, and why the divergence and curl articles each open with the geometric definition rather than the formula.