Vector Calculus
Gradients, Divergence, Curl, and the theorems that bind them together
1. Building Intuition
Regular calculus asks: how fast is a single number changing? Vector calculus asks the same question about quantities that live in 2D or 3D space — temperature, wind speed, electric field, fluid velocity. These quantities have both magnitude and direction at every point, forming a vector field.
The three fundamental operations of vector calculus each answer a different question:
- Gradient (∇f) — Given a scalar field (e.g. temperature), in which direction is it increasing fastest? The gradient is a vector pointing uphill, perpendicular to contour lines.
- Divergence (∇·F) — Is the vector field spreading out or converging at this point? Positive divergence means a source (fluid gushing out); negative means a sink (fluid draining in).
- Curl (∇×F) — Is the field rotating locally? Curl measures the tendency of the field to spin a tiny paddle wheel placed at that point. Zero curl means the field is irrotational.
All three are built from the same differential operator del (∇) — the vector of partial derivatives. This unifying notation is why Maxwell's equations for electromagnetism can be written in just four lines.
A vector field assigns a direction and magnitude to every point in space. In fluid mechanics this could be velocity; in electrostatics, the electric force on a test charge.
2. The Mathematics
The del operator (nabla) is the key notation. It packages the partial derivatives into a vector-like symbol that can act on scalars (giving gradient) or on vectors (giving divergence or curl):
Gradient
Applied to a scalar function f(x,y,z), del gives the gradient — a vector field pointing in the direction of maximum increase of f. The magnitude |∇f| is the rate of increase in that direction. Contour lines / level surfaces are always perpendicular to the gradient.
Elliptic contours of f(x,y) = x² + 0.5y². The yellow arrow (gradient) always points perpendicular to the contour through that point — the direction of steepest uphill climb.
Divergence
The dot product of ∇ with a vector field F measures how much the field spreads out from a point. If you imagine fluid flowing with velocity field F, then ∇·F > 0 means fluid is being created (source), ∇·F < 0 means it's being destroyed (sink), and ∇·F = 0 means the fluid is incompressible.
Curl
The cross product of ∇ with F measures the field's local rotation. If you place a tiny paddle wheel in a flowing fluid, curl tells you how fast and in which axis it spins. Conservative fields (like gravity) have zero curl everywhere.
Left: F = (x, y) — arrows point outward, ∇·F = 2 (source). Right: F = (-y, x) — arrows rotate, ∇×F = 2k (pure rotation, zero divergence).
The Great Theorems
Three theorems generalize the Fundamental Theorem of Calculus to higher dimensions. Each says: a global quantity (integral over a region) equals a boundary quantity (integral on the edge).
The line integral of a gradient only depends on endpoints — work done by a conservative force is path-independent.
The flux of curl through a surface equals the circulation around its boundary edge. Ampere's law (electromagnetism) is an application of Stokes' theorem.
The total divergence inside a volume equals the net flux through its closed surface. Gauss's law (electrostatics) is an application of the Divergence Theorem.
3. Applications
Maxwell's Equations
All of electromagnetism compressed into four equations using ∇· and ∇×. Gauss's law (∇·E = ρ/ε₀), no magnetic monopoles (∇·B = 0), Faraday's law (∇×E = -∂B/∂t), Ampere-Maxwell law (∇×B = μ₀J + μ₀ε₀∂E/∂t).
Fluid Dynamics
The Navier-Stokes equations use divergence (incompressibility: ∇·v = 0) and the gradient of pressure as the driving force. Curl of velocity gives vorticity — the spinning of fluid parcels in turbulence.
Machine Learning & Optimization
Gradient descent uses ∇L (gradient of the loss function) to find the steepest downhill direction in parameter space. Every backpropagation step computes a high-dimensional gradient.
Heat & Mass Transfer
Fourier's heat law: heat flux = -k∇T. Divergence of heat flux = heat generated. Combined with the time derivative gives the heat equation ∂T/∂t = α∇²T.
4. Worked Examples
Example 1: Gradient of a Scalar Field
Let f(x, y, z) = x²y + yz³. Find ∇f and evaluate at the point (1, 2, 1).
Compute each partial derivative:
At (1, 2, 1): ∇f = (4, 2, 6). The steepest ascent direction is (4, 2, 6), normalized: (4, 2, 6) / √(16+4+36) ≈ (0.53, 0.27, 0.80). The rate of change in that direction is |∇f| = √56 ≈ 7.48.
Example 2: Divergence and Curl of F = (xy, yz, zx)
The divergence varies by position — regions where x+y+z > 0 are sources, < 0 are sinks.
The non-zero curl confirms this field rotates locally — it is not conservative (no potential function exists).
Example 3: Divergence Theorem — Flux Through a Sphere
Calculate the outward flux of F = (x, y, z) through the unit sphere x²+y²+z² = 1.
Direct surface integration is tedious. Instead, apply the Divergence Theorem:
Compute the divergence: ∇·F = ∂x/∂x + ∂y/∂y + ∂z/∂z = 3. Since the divergence is constant, the volume integral is just 3 × (volume of unit sphere):
The Divergence Theorem turned a hard surface integral into a trivial multiplication. This is the same reasoning behind Gauss's law: the electric flux through any closed surface equals the total charge enclosed divided by ε₀.