Angular Width on a Sphere: Intrinsic Geometry

⚡ Key Takeaways

One-line statement: The angular width subtended by a geodesic segment on a sphere, measured intrinsically, depends nonlinearly on separation through the spherical law of cosines and exhibits curvature-induced deficits that distinguish it from Euclidean angular measurement. Key result in words: On a sphere of radius $R$ , the angle subtended at one point by a geodesic arc is governed entirely by the metric $ds^2 = R^2(d\theta^2 + \sin^2\theta\, d\phi^2)$ , and the sum of angles in any geodesic triangle exceeds $\pi$ by an amount equal to the enclosed area divided by $R^2$ . Where it applies: Geodesy, astrometry, computer graphics on curved domains, general relativity (spatial sections of constant curvature), navigation, antenna beamwidth on planetary surfaces. Who/when: Spherical trigonometry traces to Menelaus (~100 CE) and Islamic astronomers; the intrinsic formulation is due to Gauss (1827) and Riemann (1854).

💡 The Core Insight

The angle between two directions at a point on the sphere is a first-order (tangent-plane) quantity defined by the metric tensor, but the relationship between that angle and the geodesic distances it spans is a global quantity controlled by curvature. The single equation that ties everything together is the spherical law of cosines, and its infinitesimal limit recovers Euclidean geometry while its finite form encodes the deviation. Curvature manifests as angular excess, making "how wide does it look" a question whose answer depends on where you stand and how far away the object is, in ways that vanish as $R \to \infty$ .

DefinitionIntrinsic Metric and Angular Measurement

Let $S^2_R$ be the round sphere of radius $R$ embedded in $\mathbb{R}^3$ , parametrized by colatitude $\theta \in [0,\pi]$ and longitude $\phi \in [0,2\pi)$ . The intrinsic (first fundamental) metric is

1

\quad ds^2 = R^2\,d\theta^2 + R^2 \sin^2\theta\, d\phi^2.

Derivation hint: Restrict the Euclidean metric $dx^2+dy^2+dz^2$ to the parametrization $x=R\sin\theta\cos\phi$ , $y=R\sin\theta\sin\phi$ , $z=R\cos\theta$ and simplify using $\cos^2+\sin^2=1$ . The angle $\alpha$ between two tangent vectors $u, v \in T_pS^2_R$ is

2

\quad \cos\alpha = \frac{g_{ij}u^i v^j}{\sqrt{g_{ij}u^iu^j}\,\sqrt{g_{ij}v^iv^j}},

where $g_{ij} = \mathrm{diag}(R^2, R^2\sin^2\theta)$ are the components of the metric in $(\theta,\phi)$ coordinates. The symbols: $u^i, v^i$ are contravariant components of the tangent vectors; $g_{ij}$ is the metric tensor; the angle $\alpha$ is the intrinsic angular width as seen at $p$ . Derivation hint: This is the Riemannian generalization of the Euclidean dot-product formula; the metric supplies the inner product on each tangent space.

TheoremSpherical Law of Cosines

Let a geodesic triangle on $S^2_R$ have vertices $A, B, C$ , with sides (geodesic arcs) of angular length $a, b, c$ (measured as the central angle, i.e., arc-length divided by $R$ ) opposite the respective vertices, and interior angles $\alpha, \beta, \gamma$ . Then

3

\quad \cos a = \cos b \cos c + \sin b \sin c \cos\alpha.

Derivation hint: Apply the Euclidean law of cosines twice—once in the tangent plane at $A$ and once to the chord triangle in $\mathbb{R}^3$ —or use the dot product of the unit position vectors $\hat{B},\hat{C}$ expressed in a frame at $A$ . The dual (angular) law of cosines holds:

4

\quad \cos\alpha = -\cos\beta\cos\gamma + \sin\beta\sin\gamma\cos a.

Derivation hint: Apply [3] to the polar (dual) triangle, whose sides are the supplements of the original triangle's angles and vice versa.

TheoremGirard's Theorem — Angular Excess

For any geodesic triangle on $S^2_R$ with interior angles $\alpha,\beta,\gamma$ and enclosed area $\mathcal{A}$ ,

5

\quad \alpha + \beta + \gamma - \pi = \frac{\mathcal{A}}{R^2} = E,

where $E$ is the spherical excess. The angle sum strictly exceeds $\pi$ , and $E$ equals the integral of the Gaussian curvature over the triangle. Derivation hint: Decompose the sphere into lunes generated by the great circles carrying each side; a lune of dihedral angle $\theta$ has area $2R^2\theta$ ; sum the three lune-pairs and subtract the triangle counted multiple times.

ProofSpherical Law of Cosines from the Embedding

Step 1. Place the three vertices as unit vectors (scaled by $R$ ) on the sphere. Without loss of generality let $A$ point along $\hat{z}$ . The position vectors are $\hat{A},\hat{B},\hat{C}$ on the unit sphere, with $\hat{A}\cdot\hat{B} = \cos c$ , $\hat{A}\cdot\hat{C} = \cos b$ , $\hat{B}\cdot\hat{C} = \cos a$ (the central angles equal the side lengths).

Step 2. Express $\hat{B}$ and $\hat{C}$ in spherical coordinates relative to $A$ at the pole:

\hat{B} = (\sin c\cos\phi_B, \sin c\sin\phi_B, \cos c),\quad \hat{C} = (\sin b\cos\phi_C, \sin b\sin\phi_C, \cos b).

Step 3. The interior angle $\alpha$ at $A$ is the difference in longitudes, $\alpha = |\phi_B - \phi_C|$ , because the geodesics through the pole are meridians whose tangent directions differ precisely by the longitude difference.

Step 4. Compute the dot product directly:

\hat{B}\cdot\hat{C} = \sin b\sin c\cos(\phi_B-\phi_C) + \cos b\cos c = \cos b\cos c + \sin b\sin c\cos\alpha.

Step 5. Since $\hat{B}\cdot\hat{C} = \cos a$ , this yields equation [3]. $\blacksquare$

∎

ProofGirard's Theorem via Lunes

Step 1. Each pair of distinct great circles intersects in two antipodal points and divides the sphere into four lunes. A lune subtending dihedral angle $\psi$ has area $\dfrac{\psi}{2\pi}\cdot 4\pi R^2 = 2\psi R^2$ .

Step 2. Extend each side of triangle $ABC$ to its full great circle. These three great circles create the triangle $T$ and its antipodal copy $T'$ . Consider the three lunes (one per vertex) on the hemisphere containing $T$ , with areas $2\alpha R^2$ , $2\beta R^2$ , $2\gamma R^2$ .

Step 3. Together with their antipodal partners these six lunes cover the whole sphere, but $T$ and $T'$ are each covered three times instead of once (once by each lune sharing a vertex). Hence

2(2\alpha R^2 + 2\beta R^2 + 2\gamma R^2) = 4\pi R^2 + 4\mathcal{A},

where the surplus $4\mathcal{A}$ accounts for the double over-counting of $T$ and $T'$ (each over-counted twice).

Step 4. Divide by $4R^2$ :

\alpha + \beta + \gamma = \pi + \frac{\mathcal{A}}{R^2}.

This is equation [5], and applying the Gauss–Bonnet theorem $\int_T K\,dA = E$ with $K=1/R^2$ confirms it. $\blacksquare$

∎

Figure

Geodesic triangle on the sphere with subtended angle α and excess area 𝒜

CF

Carl Friedrich Gauss

1777–1855

Gauss founded the intrinsic theory of surfaces with his 1827 Disquisitiones generales circa superficies curvas, proving the Theorema Egregium: Gaussian curvature is determined entirely by the first fundamental form (the metric), independent of how the surface sits in space. For the sphere this means angular width can be computed by an inhabitant who never leaves the surface—curvature is detectable from internal angle measurements alone. Gauss also conducted the Hannover geodetic survey (1820s), measuring a large triangle between the peaks Brocken, Hoher Hagen, and Inselsberg, partly to test whether terrestrial space deviated from Euclidean geometry. His insight that the angular excess equals the integrated curvature is the operational heart of this topic.

🧠 Intuition

Imagine standing at the North Pole and looking down two meridians that strike the equator $90°$ of longitude apart. At the pole the angle between your two sight-lines is exactly $90°$ . But both meridians are perpendicular to the equator where they meet it. So this triangle has three right angles summing to $270°$ —a full $90°$ of excess, precisely $\mathcal{A}/R^2$ where $\mathcal{A}$ is one-eighth of the sphere's surface. Now shrink the triangle: as the meridians' equatorial intersections move closer, the excess shrinks toward zero and the geometry becomes Euclidean. The key intuition is that angular width "knows" about the area it encloses. On a flat plane, a tiny triangle and a huge triangle with the same shape have identical angle sums ( $180°$ ). On a sphere, you cannot scale a triangle up without changing its angles. There is no similarity, only congruence—the absolute length scale $R$ is baked in. Angular width is therefore non-conformally coupled to distance: a fixed object subtends an angle that depends on the curved path your gaze travels.

ExampleRight Spherical Triangle: Subtended Angle from Two Arcs

Given: On Earth (mean radius $R = 6371\,\text{km}$ ), two geodesic legs from a station $A$ run to $B$ (arc $c = 2000\,\text{km}$ ) and to $C$ (arc $b = 1500\,\text{km}$ ), and the direct geodesic $B$ to $C$ measures $a = 2500\,\text{km}$ . Find the angle $\alpha$ subtended at $A$ .

Step 1. Convert arc lengths to central angles via (angle) = (arc)/ $R$ :

a = 2500/6371 = 0.39240\,\text{rad},\quad b = 1500/6371 = 0.23544\,\text{rad},\quad c = 2000/6371 = 0.31392\,\text{rad}.

Step 2. Apply the spherical law of cosines [3], solved for $\cos\alpha$ :

\cos\alpha = \frac{\cos a - \cos b\cos c}{\sin b\sin c}.

Step 3. Evaluate trig values: $\cos a = 0.92388$ , $\cos b = 0.97244$ , $\cos c = 0.95108$ , $\sin b = 0.23327$ , $\sin c = 0.30878$ .

\cos\alpha = \frac{0.92388 - (0.97244)(0.95108)}{(0.23327)(0.30878)} = \frac{0.92388 - 0.92492}{0.072032} = \frac{-0.00104}{0.072032} = -0.01444.

Step 4. Invert: $\alpha = \arccos(-0.01444) = 1.5853\,\text{rad}$ .

\boxed{\alpha \approx 1.5853\ \text{rad} \approx 90.83^\circ}

(Compare to the Euclidean planar value via the flat law of cosines: $\cos\alpha_{\text{flat}} = (b^2+c^2-a^2)/(2bc)$ with arc lengths $\approx -0.01042$ , giving $\alpha_{\text{flat}}\approx 90.60°$ . The $0.23°$ difference is the curvature signature.)

ExampleAngular Excess of a Large Triangle

Given: A geodesic triangle on $S^2_R$ ( $R = 6371\,\text{km}$ ) has interior angles $\alpha = 62.0°$ , $\beta = 71.5°$ , $\gamma = 49.3°$ . Find the spherical excess and the enclosed area.

Step 1. Sum the angles: $\alpha+\beta+\gamma = 62.0+71.5+49.3 = 182.8°$ .

Step 2. Spherical excess $E = 182.8° - 180° = 2.8° = 0.048869\,\text{rad}$ .

Step 3. Apply Girard [5]: $\mathcal{A} = E\,R^2 = 0.048869 \times (6371)^2\,\text{km}^2$ .

\mathcal{A} = 0.048869 \times 4.0590\times10^{7} = 1.9836\times10^{6}\,\text{km}^2.

\boxed{E = 2.8^\circ = 0.04887\ \text{rad},\qquad \mathcal{A} \approx 1.984\times10^{6}\ \text{km}^2}

(For scale, that is roughly the area of Mexico; only a triangle this enormous yields a few degrees of measurable excess, which is why everyday surveying treats the Earth as locally flat.)

⚠️ Common mistake

The angle "subtended" intrinsically is not the visual angle through 3D space. A common error is to compute the angle between the chords in the embedding $\mathbb{R}^3$ rather than between the geodesic tangent directions on the surface. These differ. Intrinsic angular width is defined purely by equation [2] using the surface metric; the tangent vectors lie in the tangent plane, not along straight Euclidean secants. Also note that arc length and central angle are interchangeable only after dividing by $R$ —mixing kilometers and radians inside the law of cosines is a frequent and silent source of error.

Counterexample

"Similar triangles scale, so angular width is scale-invariant." False on a sphere. Consider two equilateral geodesic triangles with side arcs $a_1$ and $a_2 = 2a_1$ . On the plane both would have $60°$ angles. On the sphere, by the angular law of cosines [4] with $\alpha=\beta=\gamma$ , the angle satisfies $\cos\alpha = \cos a/(1+\cos a)$ . For $a_1 = 0.1\,\text{rad}$ : $\cos\alpha_1 = 0.99500/1.99500 = 0.49875$ , so $\alpha_1 = 60.083°$ . For $a_2 = 0.2\,\text{rad}$ : $\cos\alpha_2 = 0.98007/1.98007 = 0.49497$ , so $\alpha_2 = 60.334°$ . The larger triangle has visibly larger angles. Scaling a figure changes its angles—there is no notion of similarity on a curved surface, only congruence.

connection

The angular-width problem sits at the crossroads of several frameworks. It is the constant-positive-curvature instance of the general Gauss-Bonnet Theorem, which equates total curvature to angular excess for any 2-manifold with boundary. The metric [1] is a special case of a Riemannian Metric on a 2-manifold, and equation [2] is the abstract inner-product definition of angle in that setting. Spherical trigonometry is dual to Hyperbolic Trigonometry: replacing $\cos,\sin$ of arc lengths with $\cosh,\sinh$ gives the law of cosines on the hyperbolic plane, where triangles instead exhibit angular deficit. The infinitesimal limit recovers Euclidean Trigonometry, and the projective structure connects to Stereographic Projection, which is conformal and therefore preserves angles while distorting distances—a complementary tool for angle-preserving mapping.

🔬 Where this appears

⏳ History

- ~100 CE — Menelaus of Alexandria writes Sphaerica, the first systematic treatise on spherical triangles, introducing the configuration that bears his name.

- ~940–998 CE — Abu al-Wafa al-Buzjani and contemporaries develop spherical trigonometry for Islamic astronomy and qibla determination, proving the spherical law of sines.

- ~1463 — Regiomontanus's De triangulis omnimodis systematizes plane and spherical trigonometry for the European Renaissance.

- 1614 — John Napier publishes logarithms and the mnemonic Napier's rules for right spherical triangles, vastly easing computation.

- 1781 — Adrien-Marie Legendre proves the theorem that a small spherical triangle can be solved as a plane triangle with each angle reduced by one-third of the excess.

- 1827 — Gauss's Disquisitiones generales circa superficies curvas establishes the Theorema Egregium and the intrinsic notion of curvature; angular excess becomes a curvature integral.

- 1848 — Pierre Ossian Bonnet generalizes Gauss's local result to the Gauss–Bonnet theorem relating total curvature, geodesic curvature, and angular defects.

- 1854 — Bernhard Riemann's habilitation lecture Über die Hypothesen generalizes intrinsic geometry to arbitrary dimensions and metrics, framing the sphere as a constant-curvature manifold.

- 1900s–present — The formalism enters cosmology (FLRW models), satellite geodesy (WGS-84 corrections), and computational geometry (spherical Voronoi diagrams, GIS).

open-problem

Several subtle issues persist when the idealized round sphere is replaced by a realistic body. The Earth is an oblate spheroid, not a sphere, so the geodesic is no longer a great-circle arc; the law of cosines [3] must be replaced by elliptic-integral solutions (Vincenty's or Karney's algorithms), and "angular width" lacks the clean curvature-equals-area interpretation because Gaussian curvature varies over the surface. On a general 2-manifold of variable curvature, the angular excess is $\int_T K\,dA$ , but inverting measured angular widths to recover the curvature field is an inverse problem entangled with geodesic completeness and conjugate points. A second subtlety: near antipodal configurations the spherical law of cosines becomes numerically ill-conditioned (small $\sin b\sin c$ ), motivating the haversine reformulation—an open practical concern in high-precision geodesy. Finally, in the discrete setting (triangulated surfaces, regge calculus), curvature concentrates at vertices as angle deficits, and the correct discrete analogue of "intrinsic angular width" and its convergence to the smooth value remains an active topic in discrete differential geometry.

Angular Width on a Sphere: Intrinsic Geometry

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also