A vector field that at each point gives the direction and rate of greatest increase of a differentiable scalar function; in Euclidean coordinates it is the tuple of partial derivatives ∇f = (∂_1 f, ..., ∂_n f).

Principle

It is the unique vector field corresponding to the differential 1-form df via the inner-product (musical) isomorphism: for any direction v, the directional derivative D_v f equals the inner product ⟨∇f, v⟩.

Demonstration

For f(x,y) = x^2 + y^2 on R^2, ∇f = (2x, 2y); along the radial unit vector r̂ the directional derivative equals ⟨∇f, r̂⟩ = 2r, the rate of fastest increase.

Misapplication

Treating the gradient as invariant without specifying a metric or using it on manifolds without performing the musical isomorphism (converting df to a vector) leads to coordinate-dependent errors.

Consequence

The gradient gives the steepest ascent direction and magnitude, determines level set normals, and underlies conservative vector fields: a vector field is a gradient field iff its curl (in simply connected domains) vanishes.

Reversal

The curl or rotational component of a vector field measures failure to be a gradient; taking divergence instead emphasizes flux rather than directional increase—these are orthogonal differential notions.

Boundary

Defined for differentiable scalar functions on manifolds endowed with an inner product (Riemannian metric); on a general differentiable manifold one must use the metric to identify df with ∇f.

Semantic Tension

Distinct from the differential (df), which is a 1-form; the gradient is the metric-dependent vector representation of df—confusion arises when metric is omitted or an ambient Euclidean structure is assumed silently.

Synthesis

The gradient is the metric-dependent vector field representing the differential of a scalar: it points along steepest increase, its magnitude is the maximal directional derivative, and it links 1-forms to vector calculus constructions.