Spherical geometry is the study of figures and measurements on the surface of a sphere. Its basic objects are points, arcs of great circles, angles, and regions bounded by those arcs. Unlike Euclidean geometry, it describes a space in which the straightest paths eventually return to their starting points, any two complete straightest paths intersect, and the angles of a triangle generally add to more than two right angles. These differences follow from the sphere’s positive curvature rather than from imperfections in measurement.
The subject developed through problems in astronomy, geography, and navigation, where directions and positions are naturally represented on a spherical surface. It also provides a fundamental example in differential geometry and Riemannian geometry: a curved space whose distances, curvature, symmetries, and geodesics can be calculated explicitly. Its classical triangle formulas remain useful in global positioning and astronomical calculations, while its modern formulation connects it with topology, transformation groups, numerical computation, and harmonic analysis.
The sphere as a geometric space
A sphere of radius , centered at the origin of three-dimensional Euclidean space, is the set
Here the dot product supplies the ordinary ambient notion of length. Spherical geometry, however, measures travel along the surface rather than along straight segments through the surrounding space. This distinction separates the sphere’s intrinsic geometry from its particular representation as an object embedded in three dimensions.
The intersection of the sphere with a plane through its center is a great circle. Its radius is , and its circumference is . A plane that intersects the sphere without passing through the center produces a small circle. The equator and every complete meridian of an ideal spherical Earth are great circles; parallels of latitude other than the equator are small circles. The difference is geometric, not merely terminological: great-circle arcs are locally shortest surface paths, whereas small-circle arcs generally are not.
The angle between two intersecting surface curves is measured between their tangent directions at the intersection. For great circles, it can also be obtained from the angle between their defining planes, with the appropriate choice of the interior angle. These tangent directions lie in the tangent plane, which is perpendicular to the radius through the intersection point. Thus ordinary plane geometry still governs infinitesimal measurements, even though finite figures behave differently.
A sphere is a two-dimensional manifold: every sufficiently small neighborhood can be described with two coordinates, although no single nonsingular coordinate chart covers the entire surface. Its total area is , and its intrinsic diameter—the largest distance between two points—is . Changing multiplies every length by the same factor and every area by its square, while leaving angles unchanged. Consequently, calculations often use the unit sphere and restore physical units afterward.
The sphere is homogeneous and isotropic: no point is intrinsically distinguished, and no tangent direction at a point is preferred. A pole or equator becomes special only after a coordinate system or external rotation axis has been selected. This uniformity makes the sphere particularly tractable. Its Gaussian curvature has the same value, , everywhere, so local geometric properties depend on scale but not on location.
Great circles, distance, and shortest paths
A geodesic is a curve that follows an intrinsically straight direction. On a sphere, nonconstant geodesics are great circles traversed at constant speed. One way to see this is to examine the acceleration of a great-circle motion: it points directly toward the sphere’s center and therefore has no component tangent to the surface. The moving point turns in the surrounding three-dimensional space, but it does not turn relative to the surface itself.
For points represented by vectors and of length , their central angular separation is
Their intrinsic distance is
This formula defines a metric on the sphere, including the triangle inequality. It differs from the ambient chord distance,
For small , the chord and arc lengths are nearly equal, but their difference becomes substantial over large separations.
Two distinct, nonantipodal points determine a unique great circle and a unique shorter arc of that circle. That shorter arc is the unique minimizing path. The longer arc is also geodesic in the local differential sense, but it does not minimize distance between its endpoints. Opposite, or antipodal, points are exceptional: infinitely many great circles pass through them, and each connecting semicircle has the same minimum length, .
A useful unit-speed expression for a geodesic starting at with unit tangent vector is
where . This formula shows directly that geodesics close after length . It also demonstrates why local straightness does not imply global minimality: continuing past the antipodal point yields a route longer than the alternative shorter arc.
There are no parallel great circles. Two distinct planes through the center intersect in a line, and that line meets the sphere in two antipodal points. Yet the sphere is not simply Euclidean geometry with the parallel postulate reversed. The familiar assertion that two points determine exactly one complete line also fails for antipodal points. The treatment of these exceptional pairs becomes important when spherical geometry is compared with axiomatic elliptic geometry.
Spherical triangles, excess, and area
A standard spherical triangle is bounded by three shorter great-circle arcs joining three vertices that do not lie on one great circle. Classical formulas are usually stated for the convex triangle contained in an open hemisphere. Its side lengths, expressed as central angles, are denoted , and its opposite interior angles are . Physical side lengths are ; the angular sides themselves are measured in radians.
The most distinctive triangle identity is
where is the enclosed area. The quantity
is called the spherical excess. Thus , a result traditionally known as Girard’s theorem. An ordinary convex spherical triangle has positive area, so its angle sum exceeds . Unlike a Euclidean triangle, its angles determine its area once the sphere’s radius is fixed.
A simple example uses the North Pole and two equatorial points separated by of longitude. The two meridian sides meet the equator at right angles and meet each other at a right angle at the pole. All three angles are therefore , giving excess and area , exactly one eighth of the sphere. This example also shows why a triangle can have more than one right angle without contradiction.
The area formula can be understood through spherical lunes. A lune is the region between two great semicircles meeting at antipodal vertices. If its angle is , its area is : the fraction of the sphere occupied by the lune is . Combining the lunes associated with a triangle’s three vertices yields the excess formula. The argument makes the dependence on angular size particularly transparent and requires no coordinate calculations.
For a convex geodesic polygon with sides and interior angles ,
A triangulation establishes this by adding triangle excesses and canceling internal contributions. For nonconvex or oriented polygons, angle and area conventions require additional care because the same closed boundary can distinguish complementary regions.
The area divided by is also the solid angle subtended at the center, measured in steradians. A complete sphere subtends steradians. This connects spherical area to radiation patterns, optical fields of view, and the fraction of all possible directions occupied by an object.
Spherical trigonometry and triangle identities
Spherical trigonometry relates a spherical triangle’s sides and angles. Its formulas are not obtained by treating the three arcs as straight segments in a plane. Instead, they encode the relationships among the corresponding radius vectors and the planes of the three great circles. The spherical law of cosines is
together with the two cyclic versions. It determines a third side from two sides and their included angle.
A direct derivation places the vertex with angle at the north pole of a unit sphere. The other two vertices can be written as
Their dot product is both and the right-hand side of the cosine formula. This illustrates how linear algebra can replace a lengthy spatial construction while preserving the geometric meaning.
The spherical law of sines states
As in plane trigonometry, a sine equation alone can leave an ambiguity because supplementary angles have equal sines. A complete triangle-solving procedure must therefore check the allowed ranges and use additional cosine relations when necessary. A dual cosine rule expresses sides in terms of angles:
The symmetry between side and angle formulas reflects a geometric duality formed by taking poles of the triangle’s side circles.
For a right spherical triangle with , the cosine rule reduces to
This is the spherical counterpart of the Pythagorean theorem. If all sides are small compared with the sphere’s radius, expanding the cosines gives . More generally, the spherical cosine rule approaches the plane law of cosines as the triangle’s angular size tends to zero. Euclidean geometry is therefore the local small-scale limit, not a competing approximation valid at every scale.
For calculating area from side lengths, l’Huilier’s formula is often useful. With ,
For a nondegenerate convex triangle, this yields its positive excess. It is especially valuable when directly subtracting from the sum of three computed angles would lose significant numerical precision.
Coordinates and numerical computation
Latitude and longitude are the most familiar coordinates on a sphere. With latitude , longitude , and radius , the corresponding Cartesian coordinates are
Latitude lies between and , while longitude is periodic modulo . At either pole, every longitude denotes the same point. This is a coordinate singularity, not a geometric singularity of the sphere.
For two positions, their central separation satisfies
Although mathematically exact, evaluating this through an inverse cosine can be inaccurate for very small separations: numbers near contain little relative information about the tiny angle. This is a problem of numerical stability under floating-point arithmetic, rather than a defect in spherical trigonometry.
The haversine formulation is better suited to many short-distance calculations. Define
Then
A numerical implementation must accommodate rounding that places the computed slightly outside . Near antipodal configurations, sensitivity remains important, and vector-based methods can provide an alternative.
For unit vectors and , a particularly useful expression is
where the cross product gives the sine of the separation. Using both sine and cosine information generally behaves better than using inverse cosine alone. Exact antipodal endpoints still leave the connecting great-circle direction undetermined.
Interpolation along a shorter great-circle arc is given, for , by
This spherical linear interpolation moves at constant angular speed. Its limiting behavior needs special handling when endpoints nearly coincide, while antipodal endpoints require an additional choice of path. A related interpolation is widely used with unit quaternions in three-dimensional animation; that construction takes place on a three-dimensional sphere in four-dimensional space, rather than on the ordinary two-dimensional sphere.
Curvature, parallel transport, and intrinsic geometry
In latitude–longitude coordinates, the spherical metric tensor is expressed by
It describes how coordinate increments translate into physical distance. The longitude coefficient decreases toward the poles because parallels become smaller there. The associated area element is
Integrating this expression over the full latitude and longitude ranges gives .
The constant Gaussian curvature is intrinsic: it can be determined from measurements made entirely within the surface. This is an instance of the principle established by Carl Friedrich Gauss in his Theorema Egregium. A flat sheet cannot be bent onto a spherical region without stretching or tearing, because bending without intrinsic distortion preserves Gaussian curvature, whereas the plane has curvature zero.
Curvature also appears in the behavior of neighboring geodesics. Two meridians leaving the equator northward initially point in parallel local directions, yet converge at the North Pole. More generally, an infinitesimal transverse separation between nearby geodesics obeys
The oscillatory solutions describe focusing, in contrast to the linear separation behavior of Euclidean geodesics. The antipodal point is the first conjugate point along a geodesic from a given starting point.
Parallel transport gives another way to detect curvature. A tangent vector is transported along a curve while remaining intrinsically parallel according to the sphere’s Levi-Civita connection. If it travels around a simple closed loop, it generally returns rotated relative to its original direction. For a positively oriented loop enclosing a disk-shaped region of area , the rotation is modulo , with its sign determined by orientation conventions.
These facts are organized by the Gauss–Bonnet theorem. For a disk-shaped region with piecewise smooth boundary,
where is boundary geodesic curvature and are exterior turning angles. Great-circle sides have , so the triangle excess formula follows immediately. For the entire sphere, with no boundary, the theorem gives
linking local curvature to the Euler characteristic .
Symmetry, topology, and elliptic geometry
Every orthogonal transformation of three-dimensional Euclidean space preserves the sphere and its intrinsic distances. Conversely, every distance-preserving transformation of the round sphere comes from such a transformation. Its full isometry group is therefore ; the orientation-preserving subgroup is , the group of rotations. These groups can be represented by orthogonal matrices, and their continuous structure makes them examples of a Lie group.
This symmetry explains why many spherical problems can be simplified by rotating the configuration. One point may be placed at a pole and another on a chosen reference meridian without changing distances or angles. A spherical triangle, for example, can be described up to rotation by three suitable side or angle parameters. Reflection may also preserve all metric data while reversing orientation, which matters when signed areas or oriented boundaries are involved.
Topologically, the sphere is a compact, connected surface without boundary. It is simply connected: every closed loop can be continuously contracted to a point. Neither periodic longitude nor the closure of great circles implies a hole in the surface. Longitude’s discontinuity along a chosen meridian arises from coordinates, while a closed great circle can contract by moving through smaller latitude circles toward a pole.
Ordinary spherical geometry should be distinguished from elliptic geometry. Identifying each point with its antipode produces the real projective plane. A point in this quotient is an antipodal pair, and a line is the image of a great circle. Two distinct projective points determine exactly one line, and two distinct lines meet in exactly one projective point. The two antipodal intersection points of great circles have become a single point.
The natural distance between projective points represented by unit vectors is
Its maximum is , and each projective line has length . Locally, elliptic and spherical geometries have the same constant positive curvature, but globally they differ: the projective plane is not orientable and is not simply connected.
This distinction matters in discussions of non-Euclidean geometry. The sphere clearly violates Euclidean parallel behavior, but antipodal points also violate the usual uniqueness-of-line axiom. The projective quotient supplies a cleaner incidence structure for elliptic geometry. It illustrates how local metric properties and global topology contribute independently to the character of a geometric space.
Projections and the geometry of maps
A map projection represents part of a sphere on a plane. Because spherical and Euclidean curvature differ, no projection can preserve all local distances. Different projections preserve different structures: angles, areas, selected distances, or the straight appearance of geodesics. Understanding these distinctions is essential when interpreting routes and shapes on maps.
Stereographic projection sends a sphere, minus one point, to a plane by projecting from the omitted point. For the unit sphere, projection from the north pole onto the equatorial plane gives
It is conformal, meaning that it preserves angles between intersecting curves. Circles on the sphere become circles or straight lines in the plane; a line is obtained when the original circle passes through the projection point. Great circles do not, in general, become straight lines.
With this normalization, the spherical metric becomes
The same position-dependent factor multiplies both coordinate directions, explaining conformality. The scale factor is not constant, so conformality does not imply preservation of distances or areas. Stereographic projection also identifies the sphere with the complex plane together with one point at infinity, the Riemann sphere used in complex analysis.
A gnomonic projection instead projects from the sphere’s center onto a tangent plane. Each great circle lies in a plane through that center, so its projected image lies on a straight line. This makes the projection useful for visualizing great-circle routes. Its limitations are substantial: only an open hemisphere can be represented on one such plane, and distortion grows without bound near its boundary.
The Mercator projection uses coordinates
It preserves angles, but its scale factor is , and its local area factor is . The poles lie infinitely far away in the map plane. Its straight lines represent constant-bearing routes, or loxodromes, rather than great-circle routes in general.
Equal-area projections make a different tradeoff. The Lambert cylindrical equal-area form
preserves area because its planar area element equals . It does not generally preserve angles or shapes. These examples show why a route that bends on a flat map can nevertheless be the shortest route on the sphere, and why the apparent sizes of regions depend on the projection.
Historical development
Spherical geometry emerged from attempts to describe the heavens and the Earth. Greek astronomers represented stellar directions on a celestial sphere and studied circles associated with the horizon, equator, and ecliptic. These were practical geometric models: observations supplied angles, while the spherical configuration related those measurements to time, latitude, and celestial position. The subject developed alongside, but separately from, the plane geometry codified in Euclid’s Elements.
Theodosius of Bithynia, active around the second or first century BCE, composed the Sphaerics, a systematic treatment of geometric properties of spheres and their circles. Around the end of the first century CE, Menelaus of Alexandria developed spherical triangle geometry in another work called the Sphaerics. His spherical theorem related products of ratios of sines along arcs, providing a powerful tool before the modern collection of triangle-solving identities became standard.
In the second century CE, Ptolemy used spherical methods extensively in the Almagest. Computation relied on tables of chords rather than the modern sine notation, but the underlying problems—converting celestial coordinates, determining rising and setting, and relating observations at different latitudes—were fundamentally spherical. Mathematical astronomy preserved and extended these methods across later linguistic and intellectual traditions.
Between the ninth and thirteenth centuries, scholars working in Arabic and Persian developed spherical trigonometry into a more independent computational discipline. Important figures included al-Battani, Abu al-Wafa, al-Biruni, and Nasir al-Din al-Tusi. Problems included astronomical measurement, determining geographical coordinates, and finding the direction of Mecca. Al-Tusi’s thirteenth-century Treatise on the Complete Quadrilateral organized plane and spherical trigonometric results in a treatment not confined to a single astronomical model.
In Renaissance Europe, Regiomontanus’s De triangulis omnimodis, composed in the 1460s and printed in 1533, systematized plane and spherical triangle methods. The expansion of oceanic navigation gave such techniques increasing practical importance. Logarithms, introduced in the early seventeenth century, greatly reduced the labor of multiplying trigonometric quantities. Albert Girard stated the spherical triangle excess–area relation in 1629, while Thomas Harriot had obtained the result earlier in work that remained unpublished during his lifetime.
During the nineteenth century, the emphasis widened from triangle computation to intrinsic geometry. Gauss’s 1827 work on curved surfaces established the intrinsic character of Gaussian curvature. Bernhard Riemann’s 1854 habilitation lecture placed metrics and curvature within a general theory of manifolds. In that framework, spherical geometry became the basic model of constant positive curvature, alongside flat Euclidean geometry and constant-negative-curvature geometry.
Applications in Earth measurement, astronomy, and analysis
For global navigation, a spherical Earth model provides a direct relationship between angular separation and travel distance. Taking a representative mean Earth radius of about kilometers, one degree of great-circle arc is approximately kilometers, and a quarter-circle route is about kilometers. Great-circle calculations explain why long-distance routes often appear to bend toward a pole on a Mercator map. Actual travel paths can differ because of winds, currents, airspace, terrain, or operational constraints.
The spherical model is not identical to the physical Earth. Earth is more accurately approximated by an oblate ellipsoid; the WGS 84 reference ellipsoid has equatorial semiaxis meters and inverse flattening . Precision surveying and positioning therefore use ellipsoidal geodesics and reference systems rather than relying solely on spherical distances. Spherical geometry remains valuable when a simpler global model is adequate and as a conceptual foundation for more refined calculations.
In astronomy, a direction in space is represented by a point on the celestial sphere, regardless of the object’s distance. Equatorial coordinates use right ascension and declination, while horizontal coordinates use azimuth and altitude. Transformations between them are rotations of a spherical coordinate system and can be expressed through spherical triangles or rotation matrices. Angular separation, rather than surface distance, is the primary quantity, so a unit sphere is usually sufficient.
Spherical geometry also governs directional fields and spatial analysis. Spherical harmonics are eigenfunctions of the spherical Laplace operator. On a sphere of radius , the degree- harmonics satisfy
They provide the spherical counterpart of Fourier modes and are used to represent gravitational and magnetic fields, temperature distributions, and astronomical sky maps. The geometry determines the natural modes and their characteristic scales.
In computer graphics and global numerical modeling, a sphere is often divided into triangular or polygonal cells. Subdivisions based on an icosahedron distribute cells more evenly than a simple latitude–longitude grid, whose cells narrow toward the poles. Such meshes are used for terrain representation, atmospheric calculations, and numerical approximations to equations on curved surfaces.
The curvature imposes constraints on these meshes. For a triangulation of the sphere, every edge belongs to two triangles, so . Combining this with yields
where is the number of edges meeting at vertex . Thus a spherical triangular mesh cannot have six edges meeting at every vertex. An icosahedron realizes the required total through twelve degree-five vertices, a discrete expression of the sphere’s nonzero total curvature.