The cross-ratio is a quantity associated with four ordered points on a line that remains unchanged under projective transformations. Unlike distances or ordinary ratios of lengths, it is preserved by perspective projection, making it a fundamental invariant of projective geometry. It also applies to four points on the extended complex plane, where it is preserved by Möbius transformations. Its value depends on the order of the points, so the convention used to define it must always be specified.
Definition
For four distinct finite coordinates , one common convention is
The coordinates may be real numbers, complex numbers, or, more generally, elements of a field.
On a real affine line, this expression can be viewed as a ratio of two ratios of directed distances:
Directed distances, rather than unsigned lengths, are essential because the sign records information about the arrangement of the points.
For example,
Different authors use formulas that differ by a permutation of the arguments. These conventions are equivalent in their geometric content, but a numerical value or formula cannot be transferred between them without checking the ordering.
Points at infinity
The projective line adds a point at infinity to an affine line. When one argument is infinity, the formula is interpreted by cancellation or a limit. With the convention above,
and
For four pairwise distinct points, the cross-ratio is finite and is neither nor . Coincident points can produce , , or infinity, but some coincidence patterns leave the expression undefined.
Homogeneous-coordinate formulation
A coordinate-independent definition uses homogeneous coordinates. Represent points on the projective line by nonzero vectors in a two-dimensional vector space, with each vector defined only up to multiplication by a nonzero scalar.
Writing for the determinant of the matrix with columns , define
Rescaling any representative vector multiplies the numerator and denominator by the same factor, so the value depends only on the projective points. Taking , and similarly for the other finite points, recovers the affine formula.
This formulation works over arbitrary fields and incorporates the point at infinity without separate limiting arguments.
Projective invariance
A projective transformation of a projective line has the coordinate form
Over the complex numbers, such a map is called a Möbius transformation.
The central invariance property is
For finite expressions, this follows from
Substitution into the cross-ratio causes all additional factors to cancel. The homogeneous-coordinate proof is more uniform: applying an invertible matrix multiplies every determinant by the determinant of that matrix, which again cancels.
Geometrically, a perspective projection from one line to another preserves the cross-ratio wherever the projection defines a projectivity. Individual distances and ordinary ratios of distances generally change under such a projection.
Why four points are needed
For any two ordered triples of distinct points on a projective line, there is a unique projective transformation carrying the first triple to the second. Consequently, three points have no nontrivial numerical invariant under all projective transformations. A fourth point introduces one.
For fixed distinct , the transformation
sends
and satisfies
The cross-ratio is therefore the coordinate of the remaining point after three reference points have been normalized.
It is also a complete invariant of ordered quadruples: two ordered quadruples of distinct points have the same cross-ratio if and only if a projective transformation maps one quadruple to the other in the specified order.
Reordering the points
Although four distinct points admit orderings, their cross-ratios take at most six values. If
these values are
For example,
while exchanging the two pairs leaves the value unchanged:
The action of the symmetric group on the four arguments thus reduces to six transformations of the cross-ratio. The four permutations that preserve the expression identically form the Klein four-group.
Over the real or complex numbers, the six values are usually distinct, but special configurations have fewer:
- If , only three distinct values occur.
- If , only two distinct values occur over the complex numbers.
The corresponding configurations are called harmonic and equianharmonic, respectively. In fields of small positive characteristic, these distinctions can change.
Harmonic division and real ordering
Four ordered points form a harmonic range when
In terms of directed ratios, this means that and divide the pair in opposite ratios.
For instance,
More generally, if , the harmonic condition becomes
so
Thus an affine midpoint is a special case of a projective harmonic relation involving the point at infinity.
Harmonic division is central to straightedge constructions and to the geometry of conic sections. Harmonic configurations can be constructed using only line intersections, without measuring lengths.
For four distinct points on the real projective line, a negative cross-ratio means that the pair separates the pair in the cyclic ordering of the projective line. The sign is therefore geometric information, not an incidental consequence of coordinates.
Complex analysis and circles
In complex analysis, the natural setting is the Riemann sphere
Four distinct points on the sphere lie on a common generalized circle if and only if their cross-ratio is real. A generalized circle means either an ordinary circle or a straight line together with the point at infinity.
One way to establish this criterion is to apply a Möbius transformation sending three of the points to . Generalized circles are preserved by Möbius transformations, and the generalized circle through is the extended real line. The fourth point lies on that circle exactly when its normalized coordinate—the cross-ratio—is real.
This criterion connects a geometric incidence condition with an algebraic condition on complex coordinates. It is useful in studying circle configurations, conformal maps, and transformations of the sphere.
Applications
Perspective geometry and imaging
In ideal perspective imaging, four collinear scene points retain their cross-ratio in the image, provided the projection is nondegenerate. This allows projectively invariant comparisons between line configurations even though apparent lengths and spacing change.
The restriction to collinear points is important: four arbitrary points in a plane do not acquire a scalar cross-ratio simply by inserting their two-dimensional coordinates into the one-dimensional formula.
Hyperbolic and Hilbert geometry
Logarithms of cross-ratios define distances in several geometric models. For example, in the Klein model of hyperbolic geometry with curvature normalized to , let the line through interior points meet the boundary at , in the order
Their hyperbolic distance is
The same construction on suitable convex domains gives the Hilbert metric, often with the same factor of . Projective invariance of the cross-ratio explains why projective transformations carrying one such domain onto another preserve these distances.
Configuration spaces
In algebraic geometry, the cross-ratio parametrizes ordered configurations of four distinct points on a projective line, up to projective transformation. Over the complex numbers, the parameter space is
The three excluded values correspond to limiting degenerations in which marked points collide.
Limitations and computational considerations
A cross-ratio is an invariant of an ordered configuration, not generally of an unordered set of four points. For an unordered configuration, the collection of up to six permutation-related values must be considered.
Its invariance also concerns projective transformations, not arbitrary mappings of a line or plane. A nonlinear change of coordinates generally does not preserve it.
Numerically, direct evaluation can lose accuracy when coordinates are nearly equal or have widely different magnitudes. Homogeneous coordinates avoid the need to represent infinity by a large finite number, but they do not remove sensitivity associated with nearly coincident points. Such cases require attention to numerical stability and to the precision of the underlying point measurements.