The Metric Tensor and the Line Element

Strip a manifold of everything but its smooth structure and you are left with a bureaucrat's map: points carry labels, charts tell you which label sits next to which, and that is all. You cannot yet ask how far apart two points are, whether two directions are perpendicular, how much time a clock reads along a worldline, or which events can signal which. Every one of those questions is answered by a single extra object laid over the manifold — the metric tensor $g$ . At each point $p$ it is a symmetric, nondegenerate bilinear map $g_p:T_pM\times T_pM\to\mathbb{R}$ that eats two tangent vectors and returns a real number, the exact generalization of the Euclidean dot product to a space that may be curved and may carry a coordinate mesh as crooked as you like. From $g$ everything metrical descends. Its components $g_{\mu\nu}$ assemble into the line element $\mathrm{d}s^2=g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ , the infinitesimal squared interval between neighbouring events; integrating $\sqrt{|\mathrm{d}s^2|}$ along a curve returns its length, or in Lorentzian signature the proper time a clock accumulates. Nondegeneracy hands you the inverse metric $g^{\mu\nu}$ and with it the raising and lowering of indices. The signature — the fixed pattern of signs of the eigenvalues, protected by Sylvester's law of inertia — decides whether you are doing Riemannian geometry ( $+\!+\!+$ ) or the Lorentzian geometry ( $-\!+\!+\!+$ ) of relativity, where the same tensor that measures distance also carves the light cones and orders cause before effect. Through Einstein's equations this geometric object becomes dynamical: the metric is the gravitational field.

Key takeaways

The metric tensor $g$ is a symmetric, nondegenerate rank- $(0,2)$ tensor field that puts an inner product on every tangent space, turning a bare manifold into a geometry.
Its components define the line element $\mathrm{d}s^2=g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ — the invariant infinitesimal squared interval from which lengths, angles, and proper time all follow.
Nondegeneracy ( $\det g_{\mu\nu}\neq0$ ) guarantees an inverse $g^{\mu\nu}$ , letting you raise and lower indices and identify each vector with a unique one-form.
Signature is a coordinate-independent invariant: positive-definite gives Riemannian geometry, the Lorentzian $-\!+\!+\!+$ gives spacetime and its causal structure.
Position-dependent components are not curvature — flat space in polar coordinates already has $g_{\theta\theta}=r^2$ ; genuine curvature is what no coordinate change can remove.
Key idea: the metric is the one geometric object that converts a featureless smooth manifold into a space with measurable lengths, angles, and causal relations, and the line element $\mathrm{d}s^2=g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ is its coordinate-invariant fingerprint.

The core insight

The metric tensor is the single geometric object that converts a featureless smooth manifold into a space with measurable lengths, angles, and causal relations; the line element $\mathrm{d}s^2=g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ is its coordinate-invariant fingerprint.

Where this lives today

The metric is not a historical relic but the daily working object of gravitational physics. When LIGO and Virgo detect a gravitational wave, what passes through the detector is a ripple $h_{\mu\nu}$ in the spacetime metric itself: the line element becomes $\mathrm{d}s^2=(\eta_{\mu\nu}+h_{\mu\nu})\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ , and the strain read out — a fractional length change near $10^{-21}$ — directly measures a perturbation to $g_{\mu\nu}$ . The Event Horizon Telescope's black-hole images trace light along null geodesics of the Kerr metric, whose $g_{tt}$ and $g_{t\phi}$ components fix the shadow's size and asymmetry. Closer to home, every GPS satellite clock runs fast by about $38$ microseconds per day — a shift computed entirely from the $g_{tt}$ difference between orbit and ground, one that would drift positioning by kilometres a day if uncorrected. From Riemann's abstract quadratic form to detector templates, the metric is what modern gravitation actually measures.

The physical picture

Think of the metric as a ruler-and-protractor bolted onto every point, and possibly re-calibrated as you move. Hand it a small displacement — a tangent vector with components $\mathrm{d}x^\mu$ — and it returns the true physical length of that step through $\mathrm{d}s^2=g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ . In flat Euclidean space written in Cartesian coordinates the ruler is the identity matrix everywhere and the formula is just the infinitesimal Pythagorean theorem, $\mathrm{d}s^2=\mathrm{d}x^2+\mathrm{d}y^2+\mathrm{d}z^2$ . The metric's job is to generalize Pythagoras: it allows cross terms, and it allows the coefficients to vary from point to point, so the length of a fixed coordinate step depends on where you take it and which way it points. A crucial early lesson is that varying coefficients need not mean a curved space. Lay ordinary polar coordinates on the perfectly flat plane and the metric becomes $\mathrm{d}s^2=\mathrm{d}r^2+r^2\,\mathrm{d}\theta^2$ : the angular ruler $g_{\theta\theta}=r^2$ grows with radius simply because a fixed increment $\mathrm{d}\theta$ sweeps a longer arc far from the origin. The geometry is still flat — its curvature tensor vanishes identically — but the components advertise position dependence. Coordinate artefacts and genuine curvature look identical in the raw numbers $g_{\mu\nu}$ ; only invariants can tell them apart.

Figure

The flat plane in polar coordinates: a coordinate cell has a radial side dr and a tangential side r dθ, so ds² = dr² + r²dθ² with g_rr = 1 and g_θθ = r². At r = 2 the angular ruler reads g_θθ = 4, yet the plane is exactly flat — position-dependent components are not curvature

Nondegeneracy is the second pillar. It says no nonzero vector is orthogonal to everything, equivalently $\det(g_{\mu\nu})\neq0$ , and it is exactly what lets you invert the component matrix. The inverse $g^{\mu\nu}$ is a perfect two-way dictionary between the tangent space $T_pM$ and its dual $T_p^*M$ — every vector has a unique metric-dual one-form and back — the machinery behind raising and lowering indices. Without it the inverse metric and even the Levi-Civita connection would be undefined. In Lorentzian signature the metric also sorts directions, something a Euclidean ruler never could. A displacement is timelike, null, or spacelike as $\mathrm{d}s^2$ is negative, zero, or positive (in $-\!+\!+\!+$ ). The null directions form the light cone at each point, the timelike interior is what a massive observer can reach, and the whole causal order of spacetime is fixed before a single equation of motion is written.

Figure

The Minkowski metric ds² = −c²dt² + dx² + dy² + dz² sorts directions at every event: null rays with ds² = 0 form the 45° light cone, timelike displacements (ds² negative) lie inside it and are traversable by clocks, spacelike ones (ds² positive) lie outside. The signature is η = diag −1 +1 +1 +1

The core equations

Everything begins with the metric as a pointwise inner product — a bilinear map $g_p:T_pM\times T_pM\to\mathbb{R}$ symmetric in its two slots, $g_p(u,v)=g_p(v,u)$ , so the order of the vectors never matters and "the angle between $u$ and $v$ " is well posed. It is the single operation from which lengths ( $g_p(v,v)$ ) and angles follow. Feeding it an infinitesimal displacement and expanding in a coordinate basis produces the line element, the workhorse of the whole subject — a quadratic form in the coordinate differentials whose coefficients are the metric components.

\mathrm{d}s^2=g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu,\qquad g_{\mu\nu}\equiv g(\partial_\mu,\partial_\nu) \qquad (1)

where: $\mathrm{d}s^2$ — infinitesimal squared interval, a scalar invariant; $g_{\mu\nu}$ — metric components in the chart $\{x^\mu\}$ , literally the metric evaluated on the coordinate basis vectors $\partial_\mu$ ; $\mathrm{d}x^\mu$ — coordinate-differential basis one-forms; repeated indices summed. Two algebraic conditions promote a generic $(0,2)$ -tensor to a genuine metric. Symmetry makes the inner product order-independent, and nondegeneracy makes the component matrix invertible — both coordinate-independent statements.

g_{\mu\nu}=g_{\nu\mu},\qquad \det(g_{\mu\nu})\neq 0 \qquad (2)

where: $g_{\mu\nu}=g_{\nu\mu}$ — symmetry, so the antisymmetric part is invisible to $\mathrm{d}s^2$ ; $\det(g_{\mu\nu})\neq0$ — nondegeneracy, the passport to an inverse metric. That inverse is the matrix reciprocal of the components, a $(2,0)$ -tensor whose existence is guaranteed precisely by nondegeneracy; it is the object that raises indices.

g^{\mu\rho}g_{\rho\nu}=\delta^\mu{}_\nu,\qquad V^\mu=g^{\mu\nu}V_\nu,\quad V_\mu=g_{\mu\nu}V^\nu \qquad (3)

where: $g^{\mu\rho}$ — inverse-metric components; $\delta^\mu{}_\nu$ — Kronecker delta ( $1$ if $\mu=\nu$ , else $0$ ); the last two relations trade a vector for its metric-dual one-form and back. Under a change of chart the components pick up two inverse-Jacobian factors, the defining behaviour of a $(0,2)$ -tensor and exactly what conspires with the differentials to leave $\mathrm{d}s^2$ untouched.

g'_{\alpha\beta}=\frac{\partial x^\mu}{\partial x'^\alpha}\frac{\partial x^\nu}{\partial x'^\beta}\,g_{\mu\nu} \qquad (4)

where: $x'^\alpha$ — new coordinates; $\partial x^\mu/\partial x'^\alpha$ — Jacobian matrix of the coordinate change; $g'_{\alpha\beta}$ — components in the new chart. The raw numbers change chart to chart; only invariants carry geometry. Integrating the line element along a curve returns a physical length (Riemannian) or proper time / proper length (Lorentzian); the absolute value inside the root handles both signs of $\mathrm{d}s^2$ .

L=\int_{\lambda_1}^{\lambda_2}\sqrt{\left|g_{\mu\nu}\frac{\mathrm{d}x^\mu}{\mathrm{d}\lambda}\frac{\mathrm{d}x^\nu}{\mathrm{d}\lambda}\right|}\;\mathrm{d}\lambda \qquad (5)

where: $\lambda$ — parameter along the curve $x^\mu(\lambda)$ ; $\mathrm{d}x^\mu/\mathrm{d}\lambda$ — tangent (velocity) vector; $L$ — arc length (Riemannian) or $c\,\tau$ / proper length (Lorentzian). For a timelike worldline $L=\int c\,\mathrm{d}\tau$ is the clock reading. The flat spacetime of special relativity is the reference point: in inertial coordinates the metric is the constant Minkowski array, whose one negative eigenvalue makes the geometry Lorentzian rather than Euclidean.

\eta_{\mu\nu}=\mathrm{diag}(-1,+1,+1,+1) \qquad (6)

where: $\eta_{\mu\nu}$ — flat-spacetime metric in inertial coordinates with $c=1$ . The lone minus sign encodes all of causality; flip the whole array and you get the equally common $+\!-\!-\!-$ convention. Finally, demanding that the connection leave the metric alone singles out a unique way to differentiate on a curved space, expressed entirely in terms of the metric and its first derivatives.

\nabla_\rho\,g_{\mu\nu}=0\;\Longrightarrow\;\Gamma^\lambda_{\mu\nu}=\tfrac12\,g^{\lambda\rho}\big(\partial_\mu g_{\rho\nu}+\partial_\nu g_{\rho\mu}-\partial_\rho g_{\mu\nu}\big) \qquad (7)

where: $\nabla_\rho$ — covariant derivative; $\Gamma^\lambda_{\mu\nu}$ — Christoffel symbols of the Levi-Civita connection. Metric compatibility means parallel transport preserves lengths and angles; with zero torsion it fixes $\Gamma$ uniquely (the fundamental theorem of Riemannian geometry).

Derivation AWhy the line element is the same in every coordinate system

The goal is to prove that although each component $g_{\mu\nu}$ changes when you switch charts, the full contraction $g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ does not — it is a genuine scalar. This is why physics is written in terms of $\mathrm{d}s^2$ rather than the bare components: the interval between two nearby events is a fact about spacetime, not about the mesh you draped over it. We transform both the differentials and the components into a new chart and watch the Jacobians annihilate pairwise.

Step 1. Coordinate differentials transform with one Jacobian factor each, $\mathrm{d}x'^\alpha=(\partial x'^\alpha/\partial x^\mu)\,\mathrm{d}x^\mu$ . Why: the $\mathrm{d}x^\mu$ are basis one-forms in the cotangent space, and this is exactly the one-form (covariant differential) transformation rule.

Step 2. The components transform with two inverse Jacobian factors, $g'_{\alpha\beta}=(\partial x^\mu/\partial x'^\alpha)(\partial x^\nu/\partial x'^\beta)\,g_{\mu\nu}$ . Why: this is the definition of a $(0,2)$ -tensor, and it carries the opposite Jacobian to the differentials — the mismatch is what will cancel.

Step 3. Insert both into the primed line element: $g'_{\alpha\beta}\,\mathrm{d}x'^\alpha\,\mathrm{d}x'^\beta=\dfrac{\partial x^\mu}{\partial x'^\alpha}\dfrac{\partial x^\nu}{\partial x'^\beta}\,g_{\mu\nu}\,\dfrac{\partial x'^\alpha}{\partial x^\rho}\dfrac{\partial x'^\beta}{\partial x^\sigma}\,\mathrm{d}x^\rho\,\mathrm{d}x^\sigma$ . Why: everything is now written in the unprimed chart, so the surviving structure becomes visible once the Jacobians meet.

Step 4. Collapse each Jacobian against its inverse by the chain rule, $\dfrac{\partial x^\mu}{\partial x'^\alpha}\dfrac{\partial x'^\alpha}{\partial x^\rho}=\delta^\mu{}_\rho$ and likewise for $\nu,\sigma$ . Why: a coordinate transformation and its inverse are literally inverse matrices, so their product is the identity — this is where the two opposite transformation laws pay off.

Step 5. The deltas collapse the sums, leaving $g'_{\alpha\beta}\,\mathrm{d}x'^\alpha\,\mathrm{d}x'^\beta=g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ . Why: the line element is identical in both charts, so it is a scalar — the invariant physical interval that the coordinate-dependent components merely encode.

Go deeper · Derivation B · 1 minFlat space wearing curved-looking components: the polar plane

Derivation BFlat space wearing curved-looking components: the polar plane

The goal is to show explicitly that position-dependent metric components are not the same thing as curvature, by taking the unambiguously flat Euclidean plane and rewriting its metric in polar coordinates. The result, $\mathrm{d}s^2=\mathrm{d}r^2+r^2\,\mathrm{d}\theta^2$ , has a component $g_{\theta\theta}=r^2$ that varies with position, yet the space is the same flat plane it always was — a distinction that trips up nearly every newcomer and is worth nailing down by direct calculation.

Step 1. Start Cartesian: $\mathrm{d}s^2=\mathrm{d}x^2+\mathrm{d}y^2$ with $g_{\mu\nu}=\delta_{\mu\nu}$ , the constant identity. Why: this is manifestly the flat plane with the simplest possible metric, our fixed reference point.

Step 2. Substitute $x=r\cos\theta,\;y=r\sin\theta$ and differentiate: $\mathrm{d}x=\cos\theta\,\mathrm{d}r-r\sin\theta\,\mathrm{d}\theta$ and $\mathrm{d}y=\sin\theta\,\mathrm{d}r+r\cos\theta\,\mathrm{d}\theta$ . Why: the line element is built from the differentials, so we need how $\mathrm{d}x,\mathrm{d}y$ depend on $\mathrm{d}r,\mathrm{d}\theta$ via the chain rule.

Step 3. Square and add. The $\mathrm{d}r^2$ coefficient is $\cos^2\theta+\sin^2\theta=1$ ; the $\mathrm{d}\theta^2$ coefficient is $r^2(\sin^2\theta+\cos^2\theta)=r^2$ ; the cross term is $(-r\sin\theta\cos\theta+r\sin\theta\cos\theta)\,\mathrm{d}r\,\mathrm{d}\theta=0$ . Why: the Pythagorean identity cleans up the diagonal terms and the cross term cancels exactly, leaving a diagonal metric.

Step 4. Read off $\mathrm{d}s^2=\mathrm{d}r^2+r^2\,\mathrm{d}\theta^2$ , so $g_{rr}=1,\;g_{\theta\theta}=r^2,\;g_{r\theta}=0$ . Why: the angular ruler grows with radius because a fixed $\mathrm{d}\theta$ subtends a longer arc farther out — pure coordinate geometry.

Step 5. Compute the curvature of this metric and it vanishes identically; every component of the Riemann tensor is zero. Why: curvature is coordinate-independent, so the flat plane stays flat no matter how baroque the chart — position dependence of $g_{\mu\nu}$ alone is never evidence of curvature.

Go deeper · Derivation C · 1 minWhy a metric may be taken symmetric with no loss

Derivation CWhy a metric may be taken symmetric with no loss

The goal is to justify the standing assumption $g_{\mu\nu}=g_{\nu\mu}$ . A general rank-two tensor need not be symmetric, so why insist a metric is? Because the line element contracts $g_{\mu\nu}$ against the manifestly symmetric object $\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ , and any antisymmetric part contributes exactly nothing there — it is invisible to every length and angle, so keeping it would be idle baggage.

Step 1. Split any $(0,2)$ -tensor uniquely into symmetric and antisymmetric parts, $T_{\mu\nu}=T_{(\mu\nu)}+T_{[\mu\nu]}$ with $T_{(\mu\nu)}=\tfrac12(T_{\mu\nu}+T_{\nu\mu})$ and $T_{[\mu\nu]}=\tfrac12(T_{\mu\nu}-T_{\nu\mu})$ . Why: this decomposition is basis-independent and exhausts the tensor, so we can treat the two pieces separately.

Step 2. Contract the antisymmetric part with the differentials: $T_{[\mu\nu]}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu=0$ . Why: $\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ is symmetric under $\mu\leftrightarrow\nu$ , and the contraction of a symmetric object with an antisymmetric one always vanishes.

Step 3. Therefore $\mathrm{d}s^2=T_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu=T_{(\mu\nu)}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ . Why: only the symmetric part reaches the line element, so all metrical content lives there and the antisymmetric part is geometrically inert.

Step 4. Impose nondegeneracy on that symmetric part, $\det(g_{\mu\nu})\neq0$ . Why: this promotes a mere symmetric bilinear form into a genuine metric, equipping it with an inverse and the power to raise and lower indices.

Limiting cases

Case	Result	Meaning
Flat Euclidean 3-space, Cartesian	$\mathrm{d}s^2=\mathrm{d}x^2+\mathrm{d}y^2+\mathrm{d}z^2$	$g_{\mu\nu}=\delta_{\mu\nu}$ , the identity; positive-definite, signature $(+\!+\!+)$ ; ordinary geometry
Flat space, spherical coordinates	$\mathrm{d}s^2=\mathrm{d}r^2+r^2\mathrm{d}\theta^2+r^2\sin^2\theta\,\mathrm{d}\phi^2$	same flat geometry, but $g_{\mu\nu}$ now depends on position — coordinate artefact, not curvature
Minkowski spacetime, inertial	$\mathrm{d}s^2=-c^2\mathrm{d}t^2+\mathrm{d}x^2+\mathrm{d}y^2+\mathrm{d}z^2$	$g_{\mu\nu}=\eta_{\mu\nu}=\mathrm{diag}(-1,1,1,1)$ ; Lorentzian $(-\!+\!+\!+)$ ; defines light cones
2-sphere of radius $a$	$\mathrm{d}s^2=a^2\mathrm{d}\theta^2+a^2\sin^2\theta\,\mathrm{d}\phi^2$	genuinely curved Riemannian metric; no chart makes $g_{\mu\nu}$ constant everywhere
Weak static field	$-g_{tt}\approx1+2\Phi/c^2$	a single component carries Newtonian gravity and clock rates; $\Phi$ the Newtonian potential

Common trapPosition-dependent metric components mean the space is curved

They do not. The components $g_{\mu\nu}$ are chart-dependent bookkeeping; curvature is a coordinate-independent invariant built from their second derivatives. Polar coordinates on the flat plane already give $g_{\theta\theta}=r^2$ and spherical coordinates an even busier array — yet the Riemann tensor of both is exactly zero. Genuine curvature is certified only when no coordinate change can reduce the metric to constants everywhere (equivalently, the Riemann tensor is nonzero), as on the 2-sphere. A related trap: $\mathrm{d}s^2$ is called a "squared" interval but in Lorentzian signature it is not a true square and is freely negative — a timelike displacement has $\mathrm{d}s^2<0$ in the $-\!+\!+\!+$ convention, a feature (it marks the direction as clock-traversable), not an error to fix.

Worked examples

1 · Proper time and differential aging (numeric)

In Minkowski spacetime a stay-at-home twin remains at rest while a traveller flies out and back at constant speed $v=0.6c$ , returning after a coordinate time $\Delta t=20$ years read by the stay-at-home clock. Use the metric alone to find how much each twin ages.

Step 1. The elapsed time on any clock is the proper time $\tau$ read from the metric along its worldline, $c^2\,\mathrm{d}\tau^2=-\mathrm{d}s^2=c^2\mathrm{d}t^2-\mathrm{d}x^2$ , so $\mathrm{d}\tau=\mathrm{d}t\sqrt{1-v^2/c^2}$ for motion at speed $v$ . This is equation (5) specialized to a timelike straight segment.

Step 2. The stay-at-home twin has $v=0$ , so $\mathrm{d}\tau=\mathrm{d}t$ and the ageing is simply $\tau_{\text{home}}=\Delta t=\mathbf{20\ \text{years}}$ . Coordinate time and this twin's proper time coincide because the twin sits at fixed position.

Step 3. The traveller moves at $v=0.6c$ throughout (the turnaround is brief), so $\sqrt{1-v^2/c^2}=\sqrt{1-0.36}=\sqrt{0.64}=0.8$ . Integrating, $\tau_{\text{trav}}=0.8\times20=\mathbf{16\ \text{years}}$ .

Step 4. The traveller returns $\mathbf{4}$ years younger. There is no paradox: the two worldlines are different curves between the same pair of events, and the metric assigns each its own length. The straight inertial worldline of the stay-at-home twin has the longest proper time — a maximum, not a minimum. Home twin ages 20 yr, traveller ages 16 yr, difference 4 yr.

2 · Gravitational redshift from one metric component (numeric)

A photon climbs the $h=22.5$ m tower of the 1959 Pound-Rebka experiment. Using the weak-field metric component $-g_{tt}\approx1+2\Phi/c^2$ , predict the fractional frequency shift.

Step 1. Static clocks tick at rates set by $g_{tt}$ : the proper time per unit coordinate time is $\mathrm{d}\tau=\sqrt{-g_{tt}}\,\mathrm{d}t$ . A photon's frequency measured by local static observers therefore shifts between two heights by $\dfrac{\nu_{\text{top}}}{\nu_{\text{bot}}}=\sqrt{\dfrac{-g_{tt}(\text{bot})}{-g_{tt}(\text{top})}}$ .

Step 2. In the weak field $-g_{tt}=1+2\Phi/c^2$ with $\Phi$ the Newtonian potential; to first order the fractional shift is $\dfrac{\Delta\nu}{\nu}\approx\dfrac{\Delta\Phi}{c^2}=\dfrac{g\,h}{c^2}$ , where $g=9.81\ \mathrm{m/s^2}$ is the local gravitational acceleration.

Step 3. Insert the numbers: $\dfrac{\Delta\nu}{\nu}=\dfrac{9.81\times22.5}{(2.998\times10^8)^2}=\dfrac{220.7}{8.988\times10^{16}}=\mathbf{2.46\times10^{-15}}$ .

Step 4. A photon climbing out of the well is redshifted by this fraction — barely two parts in $10^{15}$ , yet Pound and Rebka measured it to about 10% via the Mössbauer effect, and modern optical clocks confirm the same $g_{tt}$ physics at the $10^{-18}$ level. One metric component reproduces gravitational time dilation. Δν/ν ≈ 2.46×10⁻¹⁵ redshift.

3 · Area from the metric determinant (numeric)

Model the Earth as a sphere of radius $a=6371$ km. From the induced 2-sphere metric, compute the total surface area and the area of the polar cap above latitude $60^\circ$ N.

Step 1. The sphere's line element is $\mathrm{d}s^2=a^2\,\mathrm{d}\theta^2+a^2\sin^2\theta\,\mathrm{d}\phi^2$ , so the component matrix is $\mathrm{diag}(a^2,\,a^2\sin^2\theta)$ and its determinant is $g=a^4\sin^2\theta$ . The invariant area element is $\mathrm{d}A=\sqrt{g}\,\mathrm{d}\theta\,\mathrm{d}\phi=a^2\sin\theta\,\mathrm{d}\theta\,\mathrm{d}\phi$ .

Step 2. Total area integrates $\theta$ from $0$ to $\pi$ and $\phi$ from $0$ to $2\pi$ : $A=a^2\!\int_0^{2\pi}\!\mathrm{d}\phi\int_0^{\pi}\!\sin\theta\,\mathrm{d}\theta=a^2(2\pi)(2)=4\pi a^2$ . Numerically $4\pi(6371)^2=\mathbf{5.10\times10^{8}\ \text{km}^2}$ , the textbook surface area of the Earth.

Step 3. Latitude $60^\circ$ N is polar angle $\theta=30^\circ=\pi/6$ , so the cap runs $\theta\in[0,\pi/6]$ : $A_{\text{cap}}=a^2(2\pi)\!\int_0^{\pi/6}\!\sin\theta\,\mathrm{d}\theta=2\pi a^2\big(1-\cos\tfrac{\pi}{6}\big)=2\pi a^2(1-0.8660)$ .

Step 4. Evaluate: $A_{\text{cap}}=2\pi(6371)^2(0.1340)=\mathbf{3.42\times10^{7}\ \text{km}^2}$ , about $6.7\%$ of the globe. Every step used only $\sqrt{g}$ — the determinant that reappears as the invariant volume element on any curved background. Total 5.10×10⁸ km², cap above 60°N 3.42×10⁷ km².

Physics of the portrait

The metric tensor is not a relativist's invention: it descends from a nineteenth-century revolution in pure geometry, the moment mathematicians realized a space could carry its own intrinsic notion of distance with no ambient room to sit in. The person who made that leap in full generality gave his name to the whole subject.

Origins · 1 minBernhard Riemann and how this was discovered

Bernhard Riemann

1826–1866 · founder of the intrinsic metric and Riemannian geometry

Georg Friedrich Bernhard Riemann was a shy, sickly, deeply original mathematician who in a single 1854 lecture laid the foundation of everything in this article. Asked by Gauss for a habilitation talk, he chose the boldest of three proposed topics, "Über die Hypothesen, welche der Geometrie zu Grunde liegen" (On the hypotheses that lie at the foundations of geometry), and used it to invent the manifold of arbitrary dimension carrying a positive-definite quadratic form $\mathrm{d}s^2=\sum g_{ij}\,\mathrm{d}x^i\,\mathrm{d}x^j$ — the metric — as its intrinsic measure of length. Crucially he separated the metric from curvature, showing distance is extra structure freely specified on the manifold, and introduced the curvature now bearing his name to say when a space is genuinely bent rather than merely oddly coordinated. Gauss, who had glimpsed the intrinsic geometry of surfaces in his 1827 Theorema Egregium, reportedly left astonished. Riemann had no physics in mind, yet speculated that physical space might be non-Euclidean and tied to the forces within it — an idea sixty years ahead of Einstein. He died of tuberculosis at thirty-nine; his geometry waited for Einstein, who with Marcel Grossmann's help from 1912 recognized that Riemann's metric, given Lorentzian signature and made dynamical, is exactly the gravitational field — a programme completed by the generally covariant field equations of November 1915.

History

1827

— Gauss's Disquisitiones generales circa superficies curvas defines the first fundamental form, the intrinsic metric of a surface

1854

— Riemann's habilitation lecture extends the metric to n-dimensional manifolds, founding Riemannian geometry

1869

— Christoffel introduces the symbols, built from first derivatives of the metric, that fix the connection

1908

— Minkowski recasts special relativity with the indefinite spacetime metric η, uniting space and time

1915

— Einstein's field equations make the Lorentzian metric the dynamical gravitational field

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §2.2–2.5 and §13.1 · Wald, General Relativity, §2.3 and §3.1 · Carroll, Spacetime and Geometry, §2.4–2.5 · Hartle, Gravity, ch. 7 · Schutz, A First Course in General Relativity, ch. 2 and 6. Key results: the pairwise cancellation of Jacobians makes $\mathrm{d}s^2$ a scalar; the polar-coordinate line element $\mathrm{d}s^2=\mathrm{d}r^2+r^2\mathrm{d}\theta^2$ has a vanishing Riemann tensor; metric compatibility $\nabla_\rho g_{\mu\nu}=0$ fixes the Christoffel symbols uniquely; the Minkowski signature is $\eta_{\mu\nu}=\mathrm{diag}(-1,1,1,1)$ with the timelike/null/spacelike classification following from the sign of $\mathrm{d}s^2$ . Worked numbers: twin proper time $0.8\times20=16$ yr; Pound-Rebka $gh/c^2=9.81\times22.5/(2.998\times10^8)^2=2.46\times10^{-15}$ ; sphere area $4\pi a^2=5.10\times10^8$ km² and cap $2\pi a^2(1-\cos30^\circ)=3.42\times10^7$ km² at $a=6371$ km.

Prerequisites. Tangent Vectors and the Tangent Space · Cotangent Space and One-Forms · General Tensors and the Tensor Product · The Tensor Transformation Law. Leads to. Metric Signature and Lorentzian Geometry · Raising and Lowering Indices · The Invariant Volume Element · Tensor Densities and Weights.

Ask the tutor: Give me any line element and I will read off its metric, tell you whether the space is actually curved, and walk you through the proper-time or area calculation step by step.

Is the metric tensor the same thing as a matrix?

Not quite. The components $g_{\mu\nu}$ in a chosen chart do assemble into a symmetric matrix you compute with — invert it, take its determinant, contract it with vectors. But the tensor is the coordinate-independent object those numbers represent, and the matrix changes chart to chart via two Jacobian factors (equation 4), a transformation law a plain matrix has no reason to obey. Two very different matrices — $\mathrm{diag}(1,1)$ in Cartesian coordinates and $\mathrm{diag}(1,r^2)$ in polar — represent the same metric on the same flat plane. The tensor is physical; the matrix is one of its infinitely many coordinate portraits.

Why must the metric be symmetric?

Because its only job that involves order — the line element — cannot see any antisymmetric part. When you contract $g_{\mu\nu}$ with $\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ , and the differentials are symmetric under swapping $\mu$ and $\nu$ , any antisymmetric piece $g_{[\mu\nu]}$ contributes exactly zero (the third derivation shows this in one line). So an antisymmetric part would be pure decoration: it would never affect a length, an angle, or a proper time. Since geometry is defined by those measurable quantities, we discard the inert antisymmetric part from the start and take $g_{\mu\nu}=g_{\nu\mu}$ with no loss of generality. Antisymmetric rank-two tensors matter elsewhere — the electromagnetic field $F_{\mu\nu}$ is one — but they are not metrics.

What is the difference between the metric and curvature?

The metric is more basic: it directly gives distances, angles, and times, and you can write it down freely on any manifold. Curvature is derived — built from the second derivatives of the metric through the Christoffel symbols and the Riemann tensor — and measures whether the geometry can be flattened. This is why position-dependent components are not automatically curvature: polar coordinates make $g_{\theta\theta}=r^2$ on the flat plane, whose curvature is zero. Genuine curvature, as on a sphere, is the obstruction to finding any chart in which the metric becomes constant — an honest invariant that no change of coordinates can fake or remove.

Why can the squared interval be negative?

Because in relativity the metric is Lorentzian, not Euclidean — its signature carries one minus sign, $\eta_{\mu\nu}=\mathrm{diag}(-1,+1,+1,+1)$ . The name " $\mathrm{d}s^2$ " is traditional but misleading: it is a quadratic form, not the square of a real number, so it can take either sign. The sign is precisely the useful information. A displacement with $\mathrm{d}s^2<0$ is timelike, meaning a massive clock can travel along it, and $\sqrt{-\mathrm{d}s^2}/c$ is the proper time it reads. A displacement with $\mathrm{d}s^2=0$ is null, the path of light, and $\mathrm{d}s^2>0$ is spacelike, connecting events that cannot influence each other. The single minus sign is where all of causality is stored.

How does one metric component encode gravity?

In a static weak field the time-time component carries the Newtonian potential, $-g_{tt}\approx1+2\Phi/c^2$ . Because clocks tick at the rate $\mathrm{d}\tau=\sqrt{-g_{tt}}\,\mathrm{d}t$ , a clock lower in the well runs slower — gravitational time dilation. Worked example 2 turns this into the Pound-Rebka redshift $gh/c^2\approx2.46\times10^{-15}$ , and the same $g_{tt}$ difference between orbit and ground dominates the 38 microsecond-per-day correction that keeps GPS accurate. Newtonian gravity is just the leading deviation of a single metric component from its flat value $-1$ ; the other components govern light bending, frame dragging, and strong fields.

Why does the length formula use the absolute value of the squared interval?

Because a single formula, equation (5), serves both signatures and every kind of curve. In Riemannian geometry $g_{\mu\nu}\dot x^\mu\dot x^\nu$ is always positive and the bars do nothing. In Lorentzian geometry the sign tracks the curve's causal character: negative for a timelike worldline, positive for a spacelike one, zero for a null one. Taking $|\cdot|$ lets $\int\sqrt{|\mathrm{d}s^2|}$ return a real proper time along a timelike path and a real proper length along a spacelike path with one expression. For null curves the integrand vanishes and one must use an affine parameter instead, since light rays have zero proper time — "length" degenerates exactly on the light cone.