Scalar and Vector Fields

A field is the physicist's way of saying that something has a value everywhere, not just at one place. Two kinds sit at the foundation of the entire subject. A scalar field $\phi(\mathbf r)$ pins one number to every point of space — the temperature in a room, the pressure in the ocean, the electrostatic potential around a charge. A vector field $\mathbf F(\mathbf r)$ pins an arrow, with magnitude and direction, to every point — the wind, the flow of a river, the electric and gravitational fields. Almost every equation in classical physics is a statement about how such fields are shaped and how they change from point to point, so getting these two objects exactly right is the price of admission to vector calculus, electromagnetism, and fluid dynamics. What separates a genuine scalar or vector from an arbitrary bag of numbers is not what it looks like in one frame but how it behaves when you rotate the axes. A scalar is invariant: turn your head and the temperature you feel does not change. A vector's three components mix among themselves under a rotation in exactly the same lockstep as the coordinates of the position vector — they transform with an orthogonal matrix. This transformation law is the operational definition working physicists actually use, because it is the guarantee that an equation describes real geometry rather than an accident of the chosen coordinate grid. Everything that follows — the gradient that turns a scalar into a vector, the divergence and curl that dissect a vector, the great integral theorems — is built by differentiating and integrating these two objects while respecting that law. This article establishes the objects and the rule; the operators that act on them are named here and developed in the articles that follow.

Key takeaways

A scalar field $\phi(\mathbf r)$ assigns one real number to every point of space; a vector field $\mathbf F(\mathbf r)$ assigns an arrow with magnitude and direction.
The defining test is the transformation law: scalars are invariant under axis rotation, $\phi'=\phi$ ; vectors transform component-by-component as $F'_i=\sum_j R_{ij}F_j$ with an orthogonal $R$ .
Scalar fields are pictured by level surfaces $\phi=\text{const}$ (isotherms, equipotentials); vector fields by arrows and by field lines everywhere tangent to $\mathbf F$ .
The two kinds are linked by the gradient: $\mathbf F=-\nabla\phi$ is a vector field pointing along steepest descent, and $\nabla\phi$ is always perpendicular to the level surfaces of $\phi$ .
Any scalar built from vectors by the dot product — magnitude, angle, energy density — is automatically frame-independent, because $R^{\mathsf T}R=I$ .
Not every position-dependent triple is a vector: it qualifies only if it obeys $F'_i=R_{ij}F_j$ ; otherwise it is just three separate scalars wearing a bracket.
Key idea: a scalar field is a number at every point that a rotation leaves untouched; a vector field is an arrow at every point whose components rotate together as one rigid object — and all of vector calculus is the art of differentiating and integrating these two.

The core insight

A scalar field is a number at every point that is unchanged by rotating the axes; a vector field is an arrow at every point whose three components rotate together as a single rigid object. Everything in vector calculus is built by differentiating and integrating these two fields.

Where this lives today

Scalar and vector fields are not a historical relic; they are the native data type of computational physics. A modern numerical weather-prediction model such as the ECMWF system stores the atmosphere as scalar fields (temperature, pressure, humidity) and vector fields (the three-component wind) sampled on a global grid of roughly $9\ \mathrm{km}$ spacing, then advances them with the field equations of fluid dynamics; every forecast you read is the integration of $\nabla\cdot$ and $\nabla\times$ of those fields. Computational fluid dynamics does the same for aircraft and engines, and finite-element and finite-difference-time-domain solvers do it for the electromagnetic fields inside antennas, chips, and MRI scanners. The tensor rung of the ladder is working hard too: diffusion-tensor MRI reconstructs a rank-2 field of water-diffusion tensors in the brain to trace nerve-fibre tracts, and continuum mechanics runs entirely on the stress and strain tensor fields. At the frontier, a fundamental scalar field — the Higgs field, whose nonzero value everywhere gives particles their mass — was confirmed at the LHC in 2012, and lattice gauge theory simulates the quark and gluon vector fields of the strong force on supercomputers. From a weather map to the Higgs boson, the objects being computed are still just numbers-per-point and arrows-per-point obeying the transformation law set out above.

The physical picture

Stand in a heated room. At every point there are two independent facts: a temperature, which is one number, and an air velocity, which is an arrow. The temperature reading, taken over the whole room, is a scalar field; the pattern of drafts is a vector field. Both fill the same space continuously, yet one is a number-per-point and the other an arrow-per-point, and no amount of turning your body will convert one into the other. The cleanest mental model is a landscape: the altitude of the terrain is a scalar field, its contour lines are the level surfaces, and the slope you would feel underfoot — pointing uphill, steepest where the contours crowd — is a vector field derived from it by the gradient.

Figure

A scalar field pictured as level curves of φ = ½(x² + y²) with its gradient field ∇φ = (x, y). The arrows point straight out, perpendicular to every contour; at the marked point (3, 0) the gradient is ∇φ = (3, 0) with magnitude 3, and the green tangent shows ∇φ ⟂ level curve

The transformation law is the beating heart of the whole definition, and the second picture makes it concrete with numbers. Suppose the wind at a point has components $(F_x,F_y)=(4,3)$ in your chosen frame, so its magnitude is $\sqrt{4^2+3^2}=5$ . Now rotate the axes by $\theta=36.87^\circ$ until the new $x'$ -axis lies right along the wind. In the rotated frame the very same physical arrow has components $(F'_x,F'_y)=(5,0)$ — the numbers changed completely, yet the magnitude is still $5$ and the temperature at that point is still whatever it was. The arrow never moved; only its bookkeeping did. A quantity earns the name "vector field" precisely when its components reshuffle this way, and "scalar field" when they refuse to change at all.

Figure

Rotating the axes by θ = 36.87° turns the wind components from (F_x, F_y) = (4, 3) into (F′_x, F′_y) = (5, 0). The magnitude 5 and the temperature T = 300 K are scalar invariants; only the component bookkeeping changes

The core equations

A scalar field is nothing more than a real-valued function of position: hand it a point and it returns a single number. Nothing about it changes when you rotate or translate the coordinate frame, which is exactly why temperatures, potentials, and densities are modelled this way.

\phi:\;\mathbb{R}^3\to\mathbb{R},\qquad \mathbf r\mapsto\phi(\mathbf r)=\phi(x,y,z) \qquad (1)

where: $\phi$ — the scalar field, one real number attached to each point; $\mathbf r=(x,y,z)$ — the position vector of that point. Reading: a scalar field is a smooth "number cloud" filling a region of space, and its value at a physical point is a property of the point, not of the axes. A vector field returns an arrow at each point. In Cartesian coordinates that arrow is bookkept by three components, and each component is itself an ordinary scalar function of position — but the trio only earns the name "vector" once we impose the transformation law below.

\mathbf F:\;\mathbb{R}^3\to\mathbb{R}^3,\qquad \mathbf F(\mathbf r)=F_x\,\hat{\mathbf x}+F_y\,\hat{\mathbf y}+F_z\,\hat{\mathbf z} \qquad (2)

where: $\mathbf F$ — the vector field, an arrow at each point; $F_x,F_y,F_z$ — its Cartesian components, each a scalar function of $\mathbf r$ ; $\hat{\mathbf x},\hat{\mathbf y},\hat{\mathbf z}$ — the orthonormal basis. Reading: magnitude and direction live at every point, and the components are just how that arrow is projected onto the axes you happen to have drawn. Scalar fields are visualised through their level surfaces, the loci where the field takes a fixed value. Because $\phi$ is single-valued, two different level surfaces can never intersect — a point cannot equal two constants at once — so they nest like the layers of an onion.

\phi(\mathbf r)=\text{const}\;\Longrightarrow\;\text{level (iso) surface} \qquad (3)

where: $\phi=\text{const}$ — the set of points sharing one field value: an isotherm for temperature, an equipotential for voltage, a contour for altitude. Reading: the density of these surfaces encodes how fast the field is changing — crowded surfaces mean a steep field. Vector fields are visualised through field lines: curves that are everywhere tangent to $\mathbf F$ . Demanding tangency in each coordinate direction gives a set of coupled ratios that are the differential equations of the field lines (streamlines, for a flow).

\frac{\mathrm dx}{F_x}=\frac{\mathrm dy}{F_y}=\frac{\mathrm dz}{F_z} \qquad (4)

where: $F_x,F_y,F_z$ — the components of $\mathbf F$ evaluated along the curve. Reading: integrate these ratios and you trace the arrows into continuous streamlines, whose local spacing indicates the field strength. The two transformation laws are the operational definitions, packed into one line: a scalar reports the same value in the rotated frame, while a vector's components are remixed by the rotation matrix $R$ .

\phi'(\mathbf r')=\phi(\mathbf r),\qquad F'_i=\sum_{j}R_{ij}\,F_j,\qquad R^{\mathsf T}R=I \qquad (5)

where: $R_{ij}=\hat{\mathbf e}'_i\cdot\hat{\mathbf e}_j$ — the direction cosine between new axis $i$ and old axis $j$ ; $R^{\mathsf T}R=I$ — orthogonality, which keeps lengths and angles fixed. Reading: this single equation is the membership card; obey it and you are a genuine field, fail it and you are a coordinate artefact. The gradient welds the two kinds of field together. It is a vector field built from a scalar field, and it always points normal to the level surfaces — steepest ascent — so the unit normal to any isosurface is just the normalised gradient.

\hat{\mathbf n}=\frac{\nabla\phi}{|\nabla\phi|},\qquad \mathbf F=-\nabla\phi \qquad (6)

where: $\hat{\mathbf n}$ — unit normal to $\phi=\text{const}$ ; $\nabla\phi$ — the gradient, perpendicular to the level surfaces; $\mathbf F=-\nabla\phi$ — a conservative field derived from a scalar potential, the minus sign being the physics convention so that force points downhill in energy. Reading: whenever a vector field is the gradient of a potential, the potential's level surfaces and the field lines cross at right angles.

Derivation AAny scalar built from vectors is the same in every frame

The goal here is to prove the single most useful consequence of the transformation laws: that the dot product of two vector fields — and therefore every length, angle, and energy density constructed from it — is a true scalar, identical in every rotated frame. This is what licenses us to write physics as equations between invariants and trust that they mean the same thing to every observer no matter how their axes are turned. We start from the vector law $A'_i=\sum_j R_{ij}A_j$ and lean entirely on orthogonality, $R^{\mathsf T}R=I$ .

Step 1. Write the dot product in the primed frame and substitute the transformation law into each factor: $\mathbf A'\!\cdot\mathbf B'=\sum_i A'_iB'_i=\sum_i\big(\sum_j R_{ij}A_j\big)\big(\sum_k R_{ik}B_k\big)$ . Why: to test invariance we must express the primed quantity entirely in terms of unprimed components, so both factors are rewritten with the same rule.

Step 2. Reorder the finite sums and gather the two rotation matrices onto the summed index $i$ : $\mathbf A'\!\cdot\mathbf B'=\sum_{j,k}\big(\sum_i R_{ij}R_{ik}\big)A_jB_k$ . Why: the components $A_j,B_k$ do not depend on $i$ , so the whole $R$ -dependence collapses into the inner sum, which is a matrix product waiting to be simplified.

Step 3. Apply orthogonality: $\sum_i R_{ij}R_{ik}=(R^{\mathsf T}R)_{jk}=\delta_{jk}$ . Why: this is the entire content of "rotations preserve the metric" — the columns of $R$ are orthonormal, which is exactly the statement $R^{\mathsf T}R=I$ .

Step 4. Collapse the Kronecker delta: $\mathbf A'\!\cdot\mathbf B'=\sum_{j,k}\delta_{jk}A_jB_k=\sum_j A_jB_j=\mathbf A\!\cdot\mathbf B$ . Why: the delta forces $k=j$ , leaving the original dot product — so $\mathbf A\!\cdot\mathbf B$ , and in particular $|\mathbf A|^2$ , is a genuine scalar field.

Go deeper · Derivation B · 1 minThe gradient stands perpendicular to the level surfaces

Derivation BThe gradient stands perpendicular to the level surfaces

The aim is to show that $\nabla\phi$ is everywhere normal to the surface $\phi=\text{const}$ , which is what makes the gradient the natural bridge from a scalar field to a vector field and underlies the claim that electric field lines cross equipotentials at right angles. The argument is pure chain rule: we take an arbitrary curve lying inside a level surface, differentiate the scalar along it, and discover that the gradient must be orthogonal to every tangent direction the surface admits.

Step 1. Let $\mathbf r(t)$ be any smooth curve confined to the surface, so $\phi(\mathbf r(t))=c$ for all $t$ . Why: by choosing the curve to live in the surface, we hold $\phi$ constant and can exploit that constancy in the next step.

Step 2. Differentiate with the multivariable chain rule: $\dfrac{\mathrm d}{\mathrm dt}\phi(\mathbf r(t))=\dfrac{\partial\phi}{\partial x}\dfrac{\mathrm dx}{\mathrm dt}+\dfrac{\partial\phi}{\partial y}\dfrac{\mathrm dy}{\mathrm dt}+\dfrac{\partial\phi}{\partial z}\dfrac{\mathrm dz}{\mathrm dt}=0$ . Why: the derivative of the constant $c$ is zero, and the chain rule turns the left side into a sum over the three coordinate rates.

Step 3. Recognise the sum as a dot product: $\nabla\phi\cdot\dfrac{\mathrm d\mathbf r}{\mathrm dt}=0$ , where $\mathrm d\mathbf r/\mathrm dt$ is tangent to the surface. Why: the triple of partial derivatives is by definition $\nabla\phi$ , and the triple of coordinate rates is the tangent vector, so the sum is exactly $\nabla\phi\cdot\mathbf t$ .

Step 4. Since the curve — and hence its tangent — was arbitrary within the surface, $\nabla\phi$ is orthogonal to every tangent, i.e. normal to the surface, giving $\hat{\mathbf n}=\nabla\phi/|\nabla\phi|$ . Why: a vector perpendicular to all tangents of a surface is by definition its normal, so the normalised gradient is the unit normal.

Go deeper · Derivation C · 1 minThe gradient of a scalar really is a vector field

Derivation CThe gradient of a scalar really is a vector field

It is not obvious that stacking the three partial derivatives $\partial\phi/\partial x_i$ into a bracket produces something that obeys the vector transformation law — plenty of index-carrying triples do not. The goal is to prove that $\nabla\phi$ passes the membership test of equation (5), so that calling it a "vector field" is earned rather than assumed. The engine is the chain rule applied to the coordinates themselves, plus the fact that for a rotation the inverse is the transpose.

Step 1. In the rotated frame the components of the gradient are $(\nabla\phi)'_i=\dfrac{\partial\phi}{\partial x'_i}$ ; expand by the chain rule over the old coordinates: $\dfrac{\partial\phi}{\partial x'_i}=\sum_j\dfrac{\partial x_j}{\partial x'_i}\dfrac{\partial\phi}{\partial x_j}$ . Why: $\phi$ is the same scalar in both frames, so its primed derivative is reached by differentiating through the coordinate change $x_j(x')$ .

Step 2. For a rotation $\mathbf x'=R\mathbf x$ the inverse is $\mathbf x=R^{\mathsf T}\mathbf x'$ , so $x_j=\sum_k R_{kj}x'_k$ and therefore $\partial x_j/\partial x'_i=R_{ij}$ . Why: orthogonality $R^{-1}=R^{\mathsf T}$ makes the Jacobian of the coordinate change just the rotation matrix again, with indices in the pattern needed below.

Step 3. Substitute to obtain $(\nabla\phi)'_i=\sum_j R_{ij}\,(\nabla\phi)_j$ , which is precisely the vector transformation law. Why: matching the pattern $F'_i=\sum_j R_{ij}F_j$ confirms that the gradient transforms as a rank-one tensor — a genuine vector field derived from a scalar field.

Limiting cases

The scalar-versus-vector split is the bottom rung of a ladder classified entirely by how many factors of $R$ appear in the transformation law. Reading the table downward shows scalars and vectors as the rank-0 and rank-1 members of the tensor family.

Case	Result	Meaning
Rank-0 tensor (scalar)	$\phi'=\phi$	invariant under rotation — zero indices, one component; temperature, potential, density
Rank-1 tensor (vector)	$F'_i=R_{ij}F_j$	one factor of $R$ — one index, three components; velocity, force, electric field
Rank-2 tensor	$T'_{ij}=R_{ik}R_{jl}T_{kl}$	two factors of $R$ — two indices, nine components; stress, strain, moment of inertia
Pseudoscalar / pseudovector	$\phi'=\det(R)\,\phi$	an extra sign under reflections, $\det R=\pm1$ ; the curl of a true vector is really a pseudovector
Time-dependent field	$\phi(\mathbf r,t),\ \mathbf F(\mathbf r,t)$	the same objects with an extra parameter $t$ ; the spatial transformation law is unchanged at each instant

Common trapA position-dependent triple of numbers is automatically a vector field

It is not. A field's defining feature is behaviour under a change of frame, not the mere fact of having three components that depend on $\mathbf r$ . Write down $(x^2,\,0,\,0)$ in one frame: this is a perfectly good triple of position-dependent numbers, but rotate the axes and ask whether the new components equal $R_{ij}$ times the old ones. In general they do not, because the recipe "square the first coordinate, put zeros elsewhere" is tied to the particular axes you drew — it is three unrelated scalars wearing a bracket, not a vector. The same caution runs the other way: do not confuse a vector field (an arrow that varies from point to point) with a constant vector, nor a scalar field with a lone number. Membership is decided by equation (5), and only by equation (5).

Worked examples

1 · Gradient, unit normal, and perpendicularity on a sphere (fully worked)

The scalar field is $\phi(x,y,z)=x^2+y^2+z^2$ . At the point $P=(1,2,2)$ find the level surface through $P$ , the gradient, the unit normal, and verify that the gradient is perpendicular to the surface.

Step 1. Evaluate the field: $\phi(P)=1^2+2^2+2^2=9$ , so $P$ lies on the level surface $x^2+y^2+z^2=9$ , a sphere of radius $3$ centred at the origin. Every point at distance $3$ from the origin shares this value.

Step 2. Gradient: $\nabla\phi=(2x,2y,2z)$ , which at $P$ is $\nabla\phi(P)=(2,4,4)$ . Its length is $|\nabla\phi|=\sqrt{2^2+4^2+4^2}=\sqrt{4+16+16}=\sqrt{36}=6$ .

Step 3. Unit normal: $\hat{\mathbf n}=\nabla\phi/|\nabla\phi|=(2,4,4)/6=\big(\tfrac13,\tfrac23,\tfrac23\big)$ . Compare with the radial direction $\mathbf r/|\mathbf r|=(1,2,2)/3=\big(\tfrac13,\tfrac23,\tfrac23\big)$ — they are identical, exactly as expected, since the normal to a sphere points radially outward.

Step 4. Perpendicularity check: a tangent to the sphere at $P$ is any vector orthogonal to $\mathbf r=(1,2,2)$ , for instance $\mathbf t=(2,-1,0)$ (since $\mathbf r\cdot\mathbf t=2-2+0=0$ ). Then $\nabla\phi\cdot\mathbf t=(2)(2)+(4)(-1)+(4)(0)=4-4+0=0$ . The gradient is orthogonal to the tangent, confirming $\nabla\phi\perp$ level surface. Key results: surface $r=3$ , $\nabla\phi=(2,4,4)$ , $|\nabla\phi|=6$ , $\hat{\mathbf n}=(1,2,2)/3$ .

2 · A scalar and a vector under a rotation of axes (fully worked)

At a point the temperature is $T=300\ \mathrm K$ and the wind velocity has components $(F_x,F_y)=(4,3)\ \mathrm{m\,s^{-1}}$ . Rotate the axes counter-clockwise by $\theta=36.87^\circ$ (so $\cos\theta=0.8,\ \sin\theta=0.6$ ). Find the new components and the invariants.

Step 1. The passive rotation of components is $F'_x=F_x\cos\theta+F_y\sin\theta$ and $F'_y=-F_x\sin\theta+F_y\cos\theta$ . Substitute: $F'_x=4(0.8)+3(0.6)=3.2+1.8=\mathbf{5.0}$ and $F'_y=-4(0.6)+3(0.8)=-2.4+2.4=\mathbf{0}$ .

Step 2. The vector magnitude is a scalar invariant: $|\mathbf F|=\sqrt{4^2+3^2}=5$ before, and $\sqrt{5^2+0^2}=5$ after — unchanged, as Derivation 1 guarantees.

Step 3. The temperature is a rank-0 scalar, so $T'=T=\mathbf{300\ K}$ : rotating the axes cannot change a reading that never referred to the axes. Key results: components $(4,3)\to(5,0)$ , magnitude $5$ invariant, $T=300\ \mathrm K$ invariant. This is the numerical content of the second figure: the arrow held still while its bracket of numbers reshuffled.

3 · Field lines of a rotational vector field

Find the field lines of $\mathbf F=(-y,\,x,\,0)$ and compute its local rotation.

Step 1. Apply the field-line equation (4) in the plane: $\dfrac{\mathrm dx}{-y}=\dfrac{\mathrm dy}{x}$ , which cross-multiplies to $x\,\mathrm dx+y\,\mathrm dy=0$ .

Step 2. Integrate: $\tfrac12 x^2+\tfrac12 y^2=\text{const}$ , i.e. $x^2+y^2=R^2$ . The field lines are concentric circles about the origin — the field circulates, as a rigid rotation would.

Step 3. Measure the circulation locally with the curl (developed in a later article): $(\nabla\times\mathbf F)_z=\dfrac{\partial F_y}{\partial x}-\dfrac{\partial F_x}{\partial y}=\dfrac{\partial x}{\partial x}-\dfrac{\partial(-y)}{\partial y}=1-(-1)=\mathbf{2}$ . The vorticity is $2\hat{\mathbf z}$ , uniform everywhere — this field is the velocity field of a fluid rotating rigidly at angular speed $1$ , whose vorticity is twice the angular velocity.

Physics of the portrait

The idea that space itself carries physical values at every point — rather than particles simply tugging one another across empty distance — is one of the great conceptual inventions of physics, and it was born in the mind of a bookbinder's apprentice who could barely do algebra.

Origins · 2 minMichael Faraday and how this was discovered

Michael Faraday

1791–1867 · the field concept and lines of force

Michael Faraday came to science through a bookshop, not a university, and never mastered the mathematics that would later be built on his ideas. Yet it was Faraday who, in the 1830s through the 1850s, insisted that the space around a magnet or a charge is not empty but is filled with real "lines of force" — curves you could imagine threading through the region, dense where the influence is strong, tangent to the direction a test object would move. This was the first genuinely field-theoretic picture of nature: influence residing everywhere in space, not merely in the bodies. James Clerk Maxwell took Faraday's lines of force literally and gave them equations, writing in the preface to his 1873 Treatise that he was simply translating Faraday's ideas into mathematics; the modern vector fields $\mathbf E$ and $\mathbf B$ are the direct descendants of Faraday's lines. The mathematical machinery to manipulate such fields arrived through William Rowan Hamilton, whose 1843 quaternions supplied the words scalar and vector, and then through Josiah Willard Gibbs and Oliver Heaviside in the 1880s and 1890s, who stripped the quaternion algebra down to the streamlined vector analysis — scalar fields, vector fields, and the operators grad, div, and curl — that every physicist now uses. The final layer, the transformation-law definition that decides what truly is a scalar or a vector, came with the tensor calculus of Ricci and Levi-Civita around 1900 and Voigt's coining of "tensor". So the humble contour map and arrow diagram of today rest on a lineage that runs from Faraday's intuition to Maxwell's equations to Gibbs's notation to the tensor transformation law.

History

1837

— Faraday pictures electric and magnetic action as lines of force filling space

1843

— Hamilton discovers quaternions, from which the terms scalar and vector arise

1873

— Maxwell's Treatise recasts electromagnetism as fields of force

1881

— Gibbs distributes Elements of Vector Analysis, separating the scalar and vector products

1893

— Heaviside's Electromagnetic Theory establishes the grad, div, and curl notation

1901

— Gibbs and Wilson publish Vector Analysis, standardising the modern language of fields

Going further

Texts. Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., Ch. 3 (vector analysis) · Riley, Hobson & Bence, Mathematical Methods for Physics and Engineering, 3rd ed., Ch. 10 · Stone & Goldbart, Mathematics for Physics, Ch. 11 · Hassani, Mathematical Physics, Ch. 1 and Part on tensors · Nakahara, Geometry, Topology and Physics, §2.5 (tensor character) · Byron & Fuller, Mathematics of Classical and Quantum Physics, Ch. 1. Key results collected: dot-product invariance follows from $R^{\mathsf T}R=I$ ; $\nabla\phi$ is normal to level surfaces by the chain-rule argument; $\nabla\phi$ transforms as $F'_i=R_{ij}F_j$ . Worked numbers: sphere $r=3$ with $\nabla\phi=(2,4,4)$ , $|\nabla\phi|=6$ , $\hat{\mathbf n}=(1,2,2)/3$ ; rotation $(4,3)\to(5,0)$ with magnitude $5$ invariant; field lines $x^2+y^2=R^2$ of $\mathbf F=(-y,x,0)$ with $(\nabla\times\mathbf F)_z=2$ . Physical relations such as Fourier's law $\mathbf q=-k\nabla T$ , $\mathbf E=-\nabla V$ , and Poisson's equation $\nabla^2\Phi=4\pi G\rho$ are standard applications.

Prerequisites. Vectors and vector algebra · Functions of several variables · Cartesian coordinate systems. Leads to. The Gradient Operator · Directional Derivative · Divergence of a Vector Field · Curl of a Vector Field · The Del Operator and Vector Differential Identities · Line Integrals.

Ask the tutor: Ask me to rotate a specific vector field into a new frame, sketch the level surfaces and gradient of any $\phi(x,y,z)$ you choose, or show why a given triple of numbers does or does not qualify as a genuine vector field.

What is the difference between a field and an ordinary function of one variable?

Nothing, in principle — a field is a function, just one whose domain is the points of space (and sometimes time) rather than a single number line. A scalar field $\phi(x,y,z)$ is a function of three variables returning one number; a vector field is really three such functions bundled together. What earns the special name "field" is the physical interpretation and, crucially, the transformation law: we care how the output behaves when we relabel the input axes. An ordinary graph $y=f(x)$ carries no such requirement, which is why we do not call it a scalar field even though it, too, assigns a number to each point of a line.

Why do physicists insist on the transformation law instead of just listing components?

Because the laws of physics are supposed to describe nature, not our choice of graph paper. If you define a quantity by its components in one frame and those components do not transform correctly, then a colleague using rotated axes will compute a physically different answer for the same situation — the "law" would depend on who is looking. Insisting that scalars satisfy $\phi'=\phi$ and vectors satisfy $F'_i=R_{ij}F_j$ guarantees that any equation built from them means the same thing in every frame. It is the mathematical expression of the idea that space has no preferred directions.

Is the magnitude of a vector field a scalar field?

Yes, and this is a genuinely useful fact. The magnitude $|\mathbf F|=\sqrt{F_x^2+F_y^2+F_z^2}$ is built from the dot product $\mathbf F\cdot\mathbf F$ , and Derivation 1 shows any dot product of vectors is invariant under rotation. So $|\mathbf F|$ assigns one rotation-independent number to each point — the definition of a scalar field. The same reasoning makes the kinetic-energy density $\tfrac12\rho|\mathbf v|^2$ , the field energy density $\tfrac12\varepsilon_0|\mathbf E|^2$ , and the angle between two vector fields all legitimate scalar fields. Whenever you contract indices down to none, you land back on a scalar.

How exactly does the gradient turn a scalar field into a vector field?

The gradient $\nabla\phi=(\partial\phi/\partial x,\ \partial\phi/\partial y,\ \partial\phi/\partial z)$ takes the scalar $\phi$ and reads off how fast it changes along each axis, stacking those rates into an arrow. Two facts make that arrow a bona-fide vector field. Geometrically it points along steepest ascent and is perpendicular to the level surfaces (Derivation 2). Formally its components obey the vector transformation law (Derivation 3), so it is not merely a triple but a true rank-one object. This is why potentials are so powerful: solve one scalar equation for $\phi$ , then differentiate once to get the entire vector field $\mathbf F=-\nabla\phi$ for free.

Can a field depend on time as well as position?

Absolutely — most real fields do. The velocity of a turbulent fluid is $\mathbf v(\mathbf r,t)$ , and an electromagnetic wave is a time-varying pair $\mathbf E(\mathbf r,t),\ \mathbf B(\mathbf r,t)$ . Adding time does not change the scalar-versus-vector classification at all: at each fixed instant the object still transforms under spatial rotations exactly as before, so $T(\mathbf r,t)$ is a scalar field for every $t$ and $\mathbf v(\mathbf r,t)$ a vector field for every $t$ . Time simply comes along as an extra parameter. Relativity later promotes space and time to a single four-dimensional stage, where scalars and vectors are re-examined under Lorentz transformations, but that is a separate story.

What comes after scalars and vectors — are tensors just harder versions?

Tensors are the natural continuation of the same idea, indexed by how many factors of the rotation matrix appear in their transformation law. A scalar needs zero factors (rank 0), a vector needs one (rank 1), and a rank-2 tensor like the stress or inertia tensor needs two, transforming as $T'_{ij}=R_{ik}R_{jl}T_{kl}$ with nine components. They are not "harder" so much as richer: a rank-2 tensor encodes a linear relationship between two vectors — for instance, how a stress in one direction produces strain in another. Scalars and vectors are simply the first two rungs, and the transformation law you have learned here is the ladder that reaches all of them.