Biot-Savart Law

On 30 October 1820 — seven months after Ørsted's compass twitched and one month after Ampère's wires pulled at each other — Jean-Baptiste Biot and Félix Savart stood before the Paris Academy with the missing quantitative piece: a formula for how much magnetic field a current makes, and where. Electrostatics had enjoyed such a formula for four decades; Coulomb's inverse square let you compute $\vec E$ of any charge distribution by adding up point contributions. Magnetostatics now got its exact twin — with one structural mutation that changes everything about how the field looks. The Biot–Savart law says each element $I\,\mathrm{d}\vec\ell$ of a steady current contributes a field $\mathrm{d}\vec B = (\mu_0/4\pi)\,I\,\mathrm{d}\vec\ell\times\hat{\boldsymbol{\mathscr{r}}}/\mathscr{r}^2$ : the same $1/\mathscr{r}^2$ falloff, the same superposition — but a cross product where Coulomb has a radial unit vector. The field of an element does not point away from it; it wraps around it, right-hand-rule fashion. That single 90° twist is why magnetic field lines circulate and close instead of streaming out to infinity, why there are no magnetic charges for them to end on, and why every magnetostatic calculation carries a piece of three-dimensional geometry that electrostatics never needed. This article builds the law in its three source dialects (line, sheet, volume — the dictionary of Electric Current and Current Density), earns its canonical first results by honest integration — the finite segment, the infinite wire's $\mu_0I/2\pi s$ , a loop's center — and marks its boundaries: steady, divergence-free currents only, closed circuits only. From this seed the whole differential structure of magnetostatics grows: the next three articles take its divergence (zero — no monopoles), its full loop-axis and solenoid geometry, and its curl (Ampère's law).

Key takeaways

The Biot–Savart law gives $\vec B$ of a steady current as an inverse-square superposition: $\mathrm{d}\vec B = \dfrac{\mu_0}{4\pi}\,\dfrac{I\,\mathrm{d}\vec\ell\times\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}$ , with $\boldsymbol{\mathscr{r}}$ running from the source element to the field point.
It is Coulomb's law with a twist — literally: the cross product rotates each contribution 90°, so fields circulate around currents instead of pointing away from them.
Three equivalent forms via the substitution rule: $I\,\mathrm{d}\vec\ell \to \vec K\,\mathrm{d}a' \to \vec J\,\mathrm{d}\tau'$ ; the volume form is fundamental.
Valid only for steady currents ( $\nabla\cdot\vec J = 0$ , the previous article's contract) — an isolated current element is unphysical, and only closed-circuit integrals are measurable.
Canonical results earned here: finite segment $B = \frac{\mu_0I}{4\pi s}(\sin\theta_2 - \sin\theta_1)$ ; infinite wire $B = \mu_0I/2\pi s$ , azimuthal; loop center $B = \mu_0I/2R$ .
The per-element $1/\mathscr{r}^2$ is not the total field's falloff: geometry sums it to $1/s$ for a wire and cancels it down to $1/r^3$ for a distant loop.
KEY IDEA: the field of a steady current is the inverse-square, cross-product superposition over its elements — each $I\,\mathrm{d}\vec\ell$ wraps field around itself by the right-hand rule, and all of magnetostatics is this one integral plus geometry.

The core insight

The magnetic field of a steady current is the inverse-square, cross-product superposition of contributions from every current element: each element $I\,\mathrm{d}\vec\ell$ produces a field perpendicular to both itself and the line of sight, circulating around the current direction.

Where this lives today

Where symmetry is generous, the shortcut of Ampère's law (two articles ahead) beats integration — but almost nothing engineered is that symmetric, and there the Biot–Savart law is not a historical relic but the production algorithm. Magnet-design codes for MRI machines, particle-accelerator steering magnets, fusion-device coils, and PCB electromagnetic-compatibility analysis all run discretized equation (1): chop every conductor into segments, sum the cross products. A modern gradient-coil design loop evaluates millions of Biot–Savart sums per optimization pass. The 2019 SI redefinition gave the law a metrological cameo too: $\mu_0$ , exactly $4\pi\times10^{-7}$ by fiat for seventy years, became a measured quantity — its continued agreement with the old value (to $\sim10^{-10}$ ) is a precision test rippling straight out of the fine-structure constant. The law's name also travels far beyond magnetism. The same kernel — a source element crossed into an inverse-square sight line — computes the velocity field induced by a vortex filament in fluid dynamics: aerodynamicists summing the influence of a wingtip vortex are "using the Biot–Savart law" verbatim, one of physics' cleanest structural analogies (current ↔ vorticity, $\vec B$ ↔ velocity). And in the magnetostatics of this track, the law is the trunk of the tree: the next article integrates it into the standard field catalog, and the two after that differentiate it — divergence to banish monopoles, curl to earn Ampère's law.

Origins · 1 minJean-Baptiste Biot and how this was discovered

Jean-Baptiste Biot

1774–1862 · Paris

Biot was already a Collège de France professor and a scientific celebrity — he had ridden a balloon to 4,000 metres with Gay-Lussac to measure magnetism at altitude, and proven the extraterrestrial origin of meteorites from the L'Aigle fall — when Ørsted's discovery reached Paris in the autumn of 1820. With the young physician-turned-physicist Félix Savart, he attacked the new phenomenon quantitatively: oscillating a magnetized needle near a current-carrying wire and timing its swings to extract how the force falls with distance. Their announcement of 30 October 1820 — barely a month after Ampère's first wire-force memoir — gave magnetostatics its field law, refined by Laplace into the differential element form used today. Biot's restless breadth also gave optics the "Biot" of optical rotation (saccharimetry — sugar-content measurement by polarized light — descends from him) and geodesy some of its best early arcs. The crater Biot on the Moon and the law in every electromagnetics text keep both halves of the partnership's name current.

History

1820

— April: Ørsted's compass twitches beside a current. September: Ampère quantifies wire-wire forces. 30 October: Biot and Savart announce the distance law of the current's magnetic action.

1820

— Laplace helps cast the measurements into the differential element form — the law as integrated today.

1826

— Ampère's Mémoire embeds the law in a complete mechanics of currents.

1873

— Maxwell's Treatise absorbs Biot–Savart as the static limit of the full field equations.

1892

— Lorentz's electron theory grounds the law microscopically: fields of individual moving charges, averaged into currents.

1920

s — Prandtl's aerodynamics reuses the identical kernel for vortex filaments — the Biot–Savart law of wings.

The physical picture

Picture each tiny segment of current as a little arrow, and its field as rings wrapped around that arrow: thumb along $I\,\mathrm{d}\vec\ell$ , fingers curl along $\mathrm{d}\vec B$ . The cross product $\mathrm{d}\vec\ell\times\hat{\boldsymbol{\mathscr{r}}}$ encodes the whole geometry — the contribution is strongest broadside to the element ( $\sin\theta = 1$ ) and dies to nothing along its line ( $\sin\theta = 0$ ): a current element is magnetically loud sideways and silent along its own axis.

Figure

One current element: the field wraps around it, strongest broadside, silent along the axis

Set the two great inverse-square laws side by side and the family resemblance — and the mutation — are exact. Coulomb: $\mathrm{d}\vec E = \frac{1}{4\pi\varepsilon_0}\frac{\mathrm{d}q\,\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}$ , pointing along the sight line. Biot–Savart: same $1/\mathscr{r}^2$ , same additivity, but rotated 90° by the cross product. Everything distinctive about magnetism downstream — field lines that close on themselves, zero divergence, no monopoles, circulation as the natural integral — is that one rotation, propagated through the mathematics. The next two articles will make this precise by differentiating the law; here it is visible already in the raw geometry. One number-sense warning before the calculations: the $1/\mathscr{r}^2$ belongs to each element, not to the total. Geometry decides what survives the sum. An infinite wire recruits ever more contributing length as you back away, and the total decays only as $1/s$ ; a small loop's opposite sides nearly cancel at long range, and the total dies as $1/r^3$ (the dipole field, previewed in the next article's catalog and owned in full by the magnetic-matter chapter). Same law, opposite corrections — source geometry is half of every magnetostatics problem. And a scope stamp, inherited directly from the Continuity Equation: the law is a creature of steady currents, $\nabla\cdot\vec J = 0$ . A lone current element flunks that condition — charge would pile up at its two ends — so $\mathrm{d}\vec B$ of a single element is calculational scaffolding, not an observable; only complete circuits (or divergence-free distributions) produce the fields experiments measure. And when currents genuinely change in time, Biot–Savart does not bend — it breaks, and the retarded fields of the radiation chapter (Jefimenko's equations) take over.

The law, in three dialects

For a thin wire carrying steady current $I$ , the field at position $\vec r$ is the closed-loop sum of element contributions:

\mathrm{d}\vec B = \frac{\mu_0}{4\pi}\,\frac{I\,\mathrm{d}\vec\ell'\times\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}, \qquad \vec B(\vec r) = \frac{\mu_0 I}{4\pi}\oint_{\mathcal{C}}\frac{\mathrm{d}\vec\ell'\times\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2} \qquad (1)

where: $\mathrm{d}\vec\ell'$ — line element at source point $\vec r'$ , along the current; $\boldsymbol{\mathscr{r}} = \vec r - \vec r'$ — the separation vector, from source to field point (the sign trap of this subject — reverse it and every field circulates backwards); $\mu_0$ — the permeability of free space, $4\pi\times10^{-7}\ \mathrm{T\,m/A}$ to ten significant figures (exact by definition before 2019, measured — and unchanged — since). The $4\pi$ up front is the rational-units bookkeeping that keeps factors of $\pi$ out of the field equations later. The sheet and volume dialects follow from the substitution rule of Electric Current and Current Density — $I\,\mathrm{d}\vec\ell' \to \vec K\,\mathrm{d}a' \to \vec J\,\mathrm{d}\tau'$ :

\vec B(\vec r) = \frac{\mu_0}{4\pi}\int\frac{\vec K(\vec r')\times\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}\,\mathrm{d}a', \qquad \vec B(\vec r) = \frac{\mu_0}{4\pi}\int\frac{\vec J(\vec r')\times\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}\,\mathrm{d}\tau' \qquad (2)

where: the volume form is the fundamental one — lines and sheets are its squeezed limits — and it is the form whose divergence and curl the next articles will take to extract the field equations of magnetostatics. Everything is linear in the sources: superposition holds exactly, fields of separate circuits simply add, and the numerical "Biot–Savart solvers" of magnet-design software do nothing but chop conductors into segments and sum equation (1). Now the calculations that made the law famous. The first is the workhorse of the subject — and doing the finite segment before taking the infinite limit buys a formula that handles polygons, half-lines, and real bench wires for free.

Figure

Segment geometry: endpoint angles from the perpendicular foot decide everything

Derivation AThe straight segment, then the infinite wire

The goal is the field a perpendicular distance $s$ from a straight run of current — first for a finite segment seen between angles $\theta_1$ and $\theta_2$ , then the infinite-wire limit $B = \mu_0I/2\pi s$ . The strategy is direct integration of (1) in the cleanest coordinates: source points along $z$ , field point on the perpendicular axis. Watch two things: how the cross product locks the field's direction to pure azimuth before any integral is done, and how the finite-segment answer packages into sines of the endpoint angles — the form every polygon problem reuses.

Step 1. Geometry. Wire along $\hat{\vec z}$ , current $+z$ ; field point at distance $s$ in the plane $z = 0$ . A source element at height $z'$ has

\mathrm{d}\vec\ell' = \mathrm{d}z'\,\hat{\vec z}, \qquad \boldsymbol{\mathscr{r}} = s\,\hat{\vec s} - z'\hat{\vec z}, \qquad \mathscr{r} = \sqrt{s^2 + z'^2}.

Step 2. Cross product first:

\mathrm{d}\vec\ell'\times\hat{\boldsymbol{\mathscr{r}}} = \frac{\mathrm{d}z'}{\mathscr{r}}\,\hat{\vec z}\times(s\hat{\vec s} - z'\hat{\vec z}) = \frac{s\,\mathrm{d}z'}{\mathscr{r}}\,\hat{\boldsymbol\phi}.

Why this settles the direction for good: every element of the wire — near or far, above or below — contributes along the same $\hat{\boldsymbol\phi}$ . The field of a straight current is purely azimuthal, circling the wire per the right-hand rule; no integration required for that conclusion, only for the magnitude.

Step 3. Assemble the magnitude integral over a segment running from $z_1'$ to $z_2'$ :

B = \frac{\mu_0 I}{4\pi}\int_{z_1'}^{z_2'}\frac{s\,\mathrm{d}z'}{(s^2 + z'^2)^{3/2}}.

Step 4. Evaluate with the standard antiderivative (or substitute $z' = s\tan\theta$ , $\theta$ the angle of the sight line off the perpendicular):

\int\frac{s\,\mathrm{d}z'}{(s^2+z'^2)^{3/2}} = \frac{z'}{s\sqrt{s^2+z'^2}} = \frac{\sin\theta}{s},

so, between endpoint angles $\theta_1$ and $\theta_2$ (measured from the perpendicular foot to the segment's ends),

B = \frac{\mu_0 I}{4\pi s}\left(\sin\theta_2 - \sin\theta_1\right) \qquad (3)

Why keep this intermediate form: it is the reusable tool. Any circuit made of straight runs — square loops, hairpins, PCB traces — is a sum of (3)'s with the right angles, as worked example 2 shows. It also exposes the near-field/far-field anatomy of a real finite wire: close in ( $s$ small compared to the segment) the sines saturate at $\pm1$ and the wire acts infinite; far out it acts like a point element again.

Step 5. Take the infinite limit $\theta_1 \to -90^\circ$ , $\theta_2 \to +90^\circ$ :

\vec B = \frac{\mu_0 I}{2\pi s}\,\hat{\boldsymbol\phi} \qquad (4)

Why the $1/s$ (not $1/s^2$ ): backing away from the wire, each element weakens as $1/\mathscr{r}^2$ , but the stretch of wire contributing effectively — the part within roughly $\pm45°$ of the perpendicular — grows like $s$ . One power cancels. This is the formula the parallel-wire force borrowed last article, now paid for honestly. *Verified: the antiderivative differentiates back to the integrand ✓; (4) matches Griffiths Ex. 5.5 / Eq. 5.39 and the Ampère's-law one-liner ( $B\cdot2\pi s = \mu_0I$ ) that Curl of B will legitimize; numbers — 15 A at 2.0 cm gives $B = 1.5\times10^{-4}$ T, three times Earth's field, the reason a compass jumps near a jumper cable.*

Go deeper · Derivation B · 1 minThe center of a circular loop — the cross product pays off in one line

Derivation BThe center of a circular loop — the cross product pays off in one line

The goal is the field at the exact center of a circular loop of radius $R$ carrying current $I$ — the simplest closed-circuit Biot–Savart result and the calibration point for every coil. The strategy is to notice that the loop's geometry makes every factor in (1) constant around the ring, so the integral collapses to a circumference. (The field off center along the axis, with its $\theta$ -resolution and dipole limit, is the opening business of the next article, Magnetic Field of Steady Currents — this derivation deliberately stops at the sweet spot where no components need resolving.)

Step 1. Geometry at the center. For every element of the ring, $\mathrm{d}\vec\ell'$ is tangent and $\boldsymbol{\mathscr{r}}$ points from the rim to the center: the two are exactly perpendicular, and $\mathscr{r} = R$ for all elements. So each contribution has magnitude

\mathrm{d}B = \frac{\mu_0 I}{4\pi}\,\frac{\mathrm{d}\ell'}{R^2},

and — by the right-hand rule applied around the ring — every single one points the same way: along the loop's axis. Why no components cancel here: at the center, the perpendicularity is perfect and universal; this is the one point where the Biot–Savart integrand is constant in both magnitude and direction.

Step 2. Sum the circumference:

B = \frac{\mu_0 I}{4\pi R^2}\oint\mathrm{d}\ell' = \frac{\mu_0 I}{4\pi R^2}\,(2\pi R) = \frac{\mu_0 I}{2R} \qquad (5)

Step 3. Stack $N$ turns — superposition at its laziest: $B = \mu_0NI/2R$ . A 10-turn, 5.0-cm-radius coil at 2.0 A puts $2.5\times10^{-4}$ T at its center — five Earth-fields from bench-top hardware, and the standard classroom "tangent galvanometer" geometry. Verified: (5) is the $z = 0$ value of the full axial formula $\mu_0IR^2/2(R^2+z^2)^{3/2}$ derived next article — the two agree at the center ✓; numeric example recomputed; direction check: current counterclockwise seen from above → field up, right-hand rule through the loop.

Limiting cases

Case	Result	Meaning
finite straight segment	$B = \frac{\mu_0I}{4\pi s}(\sin\theta_2 - \sin\theta_1)$	the polygon tool; owned here
infinite straight wire	$B = \mu_0I/2\pi s$ , azimuthal	the workhorse; $1/s$ from recruited length
loop center	$B = \mu_0I/2R$ ( $\times N$ turns)	calibration point; full axis → next article
far from any closed loop	$B \sim \mu_0 m/2\pi r^3$ ( $m = IA$ )	near-cancellation → dipole falloff; the catalog article and E5 own it
sheet / volume sources	replace $I\,\mathrm{d}\vec\ell' \to \vec K\,\mathrm{d}a' \to \vec J\,\mathrm{d}\tau'$	one law, three dialects
unsteady currents	Biot–Savart FAILS	$\nabla\cdot\vec J \neq 0$ → retarded (Jefimenko) fields, radiation chapter
take $\nabla\cdot$ of the law	$\nabla\cdot\vec B = 0$	next-but-one article's theorem — the law's built-in no-monopole clause
take $\nabla\times$ of the law	$\nabla\times\vec B = \mu_0\vec J$	Ampère's law, earned two articles ahead

Common trapThe separation vector runs from source to field point — and lone elements are fictions

Two habits prevent the classic Biot–Savart disasters. First, direction discipline: $\boldsymbol{\mathscr{r}} = \vec r - \vec r'$ points from the current element to where you want the field. Reverse it and the cross product — hence every field in the problem — circulates the wrong way. Fix the convention before the first integral and re-check with the right-hand rule on the final answer (thumb along $I$ , fingers give $\vec B$ 's wrap). Second, physicality discipline: never quote the field "of a current element" as an observable. A lone element violates steady continuity (charge would pile at its ends), and only closed-circuit integrals are measurable; the differential law is scaffolding. Related: adding non-loop fragments of different circuits, or applying the law to a capacitor-charging wire (unsteady!), produces well-formed nonsense — the law's small print is $\nabla\cdot\vec J = 0$ , inherited from the previous article.

Worked examples

1 · A compass next to a jumper cable (confidence builder)

A car's starter cable carries $I = 15\ \mathrm{A}$ (treat it as long and straight). (a) Find $B$ at $s = 2.0\ \mathrm{cm}$ . (b) At what distance does the cable's field drop to Earth's $50\ \mathrm{\mu T}$ ? (c) A 10-turn circular coil of radius 5.0 cm carries 2.0 A — compare its center field.

Step 1. Equation (4): $B = \dfrac{\mu_0I}{2\pi s} = \dfrac{(4\pi\times10^{-7})(15)}{2\pi(0.020)} = \boxed{1.5\times10^{-4}\ \mathrm{T}}$ — three Earth-fields; a compass beside it swings hard toward the wire's circulation.

Step 2. Setting $\mu_0I/2\pi s = 50\ \mathrm{\mu T}$ : $s = \dfrac{(4\pi\times10^{-7})(15)}{2\pi(5.0\times10^{-5})} = 6.0\ \mathrm{cm}$ — beyond a hand-span, Earth wins. The $1/s$ falloff is gentle (halve the distance, double the field), which is why current-carrying cabling near magnetometers is a real instrumentation nuisance.

Step 3. The coil, from (5) with $N = 10$ : $B = \dfrac{\mu_0NI}{2R} = \dfrac{(4\pi\times10^{-7})(10)(2.0)}{2(0.050)} = \boxed{2.5\times10^{-4}\ \mathrm{T}}$ — a coil concentrates field better than a wire at the same current: geometry, not brute amperage, is the first tool of magnet design.

2 · The square loop — the segment formula earns its keep (exam level)

A square loop of side $a = 8.0\ \mathrm{cm}$ carries $I = 3.0\ \mathrm{A}$ . Find the field at its center, and compare with a circular loop inscribed in the square (radius $a/2$ ) carrying the same current.

Step 1. One side, seen from the center: perpendicular distance $s = a/2$ , and the side's ends sit at $\pm45°$ from the perpendicular foot ( $\tan\theta = (a/2)/(a/2) = 1$ ). The segment formula (3):

B_{\rm side} = \frac{\mu_0I}{4\pi(a/2)}\left[\sin45^\circ - \sin(-45^\circ)\right] = \frac{\mu_0I}{2\pi a}\,\sqrt2\, \cdot \, 2 \cdot \frac{1}{2} = \frac{\sqrt2\,\mu_0 I}{2\pi a}.

Step 2. Four sides, all contributions along the axis by the right-hand rule (each side's field at the center points the same way for a consistent current sense):

B = 4B_{\rm side} = \frac{2\sqrt2\,\mu_0 I}{\pi a} = \frac{2\sqrt2\,(4\pi\times10^{-7})(3.0)}{\pi(0.080)} = \boxed{4.2\times10^{-5}\ \mathrm{T}}.

Step 3. The inscribed circle, radius $R = a/2$ , from (5): $B_{\rm circ} = \dfrac{\mu_0I}{2(a/2)} = \dfrac{\mu_0I}{a} = 4.7\times10^{-5}\ \mathrm{T}$ .

Step 4. Ratio: $B_{\rm square}/B_{\rm circ} = \dfrac{2\sqrt2/\pi}{1} \approx \boxed{0.90}$ — the square's corners sit farther from the center than the circle's rim, and those long-diagonal stretches contribute weakly ( $1/\mathscr{r}^2$ per element). Rule of thumb it teaches: for the same enclosed scale and current, rounder windings concentrate more central field — one reason coils are round, beyond ease of manufacture.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §5.2 · Jackson, Classical Electrodynamics, 3rd ed., §5.1–5.3 · Purcell & Morin, Electricity and Magnetism, 3rd ed., Ch. 6 · Zangwill, Modern Electrodynamics, Ch. 10.

Provenance. Equations from Griffiths §5.2–5.4, Jackson §5.1–5.3, Purcell Ch. 6, Zangwill Ch. 10. A verifier should recompute: (1) the infinite-wire integral $\int_{-\infty}^\infty (s^2+z'^2)^{-3/2}\mathrm{d}z'=2/s^2$ giving $B=\mu_0 I/2\pi s$ ; (2) the loop integral giving $B_z=\mu_0 I R^2/2(R^2+z^2)^{3/2}$ with center value $\mu_0 I/2R$ ; (3) the divergence and curl derivations via standard vector identities and $\nabla(1/\mathfrak{r})=-\hat{\boldsymbol{\mathfrak{r}}}/\mathfrak{r}^2$ . Constant $\mu_0=4\pi\times10^{-7}\,\mathrm{T\,m/A}$ (exact pre-2019; $\approx1.25663706\times10^{-6}$ post-redefinition). Parallel-wire force $2\times10^{-7}\,\mathrm{N/m}$ at 1 A, 1 m is the historical ampere definition. (In this edition the full loop-axis integral is owned by Magnetic Field of Steady Currents, the divergence proof by Divergence of B, the curl by Curl of B, and the vector-potential form by Magnetic Vector Potential — all forward-referenced; this article owns the law, the finite segment, the infinite wire, and the loop center.) Worked numbers recomputed: 15 A at 2 cm → $1.5\times10^{-4}$ T, Earth-crossover 6.0 cm; 10-turn 5-cm coil at 2 A → $2.5\times10^{-4}$ T; square loop 3 A, 8 cm → $4.2\times10^{-5}$ T vs inscribed circle $4.7\times10^{-5}$ T, ratio $2\sqrt2/\pi = 0.90$ .

Builds on: Electric Current and Current Density · Continuity Equation · Lorentz Force Law

Leads to: Divergence of B and Absence of Magnetic Monopoles · Curl of B · Ampère's Law · Magnetic Vector Potential · Magnetic Field of Steady Current Configurations

Is the Biot-Savart law a fundamental law of nature or a derived result?

Within magnetostatics it is fundamental — the experimental input, playing exactly Coulomb's role. Within full electrodynamics it is derived: it is the steady-current limit of the retarded solutions of Maxwell's equations (Jefimenko's fields), which is also why it fails the moment currents change appreciably in time. Best mental filing: Coulomb's law and Biot–Savart's law are the two static pillars, each exact in its regime, each the low-frequency shadow of the one relativistic field theory — a unification this track cashes in the relativistic tier, where $\vec B$ of a moving charge emerges from boosting $\vec E$ .

Why does the law involve a cross product at all — where does the 90° twist physically come from?

Deep answer: relativity. The magnetic field is what the Coulomb field of moving charges looks like after a boost, and the boost geometry inserts the velocity direction into the field — the only rotationally consistent way to combine "current direction" and "sight line" into a vector perpendicular to the force it must produce (itself a cross product, $q\vec v\times\vec B$ ) is the cross product structure. Phenomenologically, it is simply what Biot and Savart measured: a needle near a wire feels a torque around the wire, not toward it. The two cross products — one in making the field, one in feeling it — are relativity's double signature on magnetism.

The formula has no time in it. What if the current is switched on suddenly?

Then Biot–Savart is silent — and wrong if you force it. The law assumes the current is steady and has been: its field adjusts instantaneously everywhere, which real physics forbids. Switch on a current and the news travels outward at $c$ ; the honest fields during the transient come from the retarded Jefimenko equations, which reduce to Biot–Savart only when the source is (and has long been) steady. The practical criterion is quasistatic validity: if currents change on timescale $T$ , Biot–Savart is a good approximation within distances $r \ll cT$ . At 50 Hz that is a thousand kilometres — power engineers may use magnetostatics freely; antenna engineers, changing currents at GHz, may not.

How do I use the law for a current that isn't a thin wire — a thick busbar or a spinning charged sphere?

Switch dialects via the substitution rule: $I\,\mathrm{d}\vec\ell' \to \vec J\,\mathrm{d}\tau'$ for volume distributions (the busbar — integrate over the cross-section too), $\to \vec K\,\mathrm{d}a'$ for surface currents (the spinning charged shell, whose $\vec K = \sigma\vec v$ was built in Electric Current and Current Density). The integrals get harder — often much — but nothing conceptual changes: superpose inverse-square, cross-product contributions from wherever current actually flows. The spinning-shell calculation, a classic, is usually routed through the vector potential instead, precisely because $\vec A$ 's integrand ( $\vec J/\mathscr{r}$ , no cross product) is friendlier — a preview of why Magnetic Vector Potential exists.

Which should I reach for first on a new problem — Biot-Savart or Ampère's law?

Symmetry decides. Ampère's law (two articles ahead) turns the field into a one-line quotient — but only when symmetry pins the field's direction and constancy on a whole loop: infinite wires, solenoids, toroids, planes, and essentially nothing else. Everything else — finite segments, single loops off-axis, Helmholtz pairs, real bent conductors — is Biot–Savart territory, by hand where geometry is kind and numerically where it is not. Good practice: classify the source's symmetry first; if you find yourself inventing an "Ampèrian loop" on which $B$ is not actually constant, you have left Ampère country, and honest integration is the only road.

Does each dB from one element separately satisfy the magnetostatic field equations?

Its divergence, yes — the wrap-around geometry of a single element's field is already divergence-free (the proof next article never uses closedness of the circuit). Its curl, no: a lone element's field has curl in the wrong places, because the element by itself is not a legal steady current (its charge piles up). Only after summing around a closed circuit — whereupon the unphysical end effects cancel — does the total field satisfy Ampère's law with the true current as source. This is the differential-level restatement of "the element law is scaffolding; the loop integral is physics," and it is why the curl derivation in Curl of B will need $\nabla\cdot\vec J = 0$ as an explicit ingredient while the divergence derivation does not.

Is it Biot and Savart's law, or Ampère's? Who actually found what in 1820?

A genuinely shared trophy, split by method. Ørsted (April) found the phenomenon. Ampère (September onward) built the theory of forces between currents and the grand hypothesis that all magnetism is currents. Biot and Savart (October) did the precision measurement — oscillating a needle near a long wire to extract the $1/s$ force law — and, with Laplace's analytical help, the element form that bears their names. Maxwell later judged Ampère's force law "one of the most brilliant achievements in science"; the field-element formula, though, was Biot and Savart's experimental territory, and the modern division — their name on the field law, Ampère's on the circulation law two articles hence — roughly honors history as well as convention.