Definition

Let VV be a finite-dimensional over a field kk. An invertible A:VVA:V\to V induces a projective transformation

P(A):P(V)P(V),[v][Av].\mathbb P(A):\mathbb P(V)\longrightarrow\mathbb P(V), \qquad [v]\longmapsto[Av].

More generally, a linear isomorphism VWV\to W induces a projective isomorphism P(V)P(W)\mathbb P(V)\to\mathbb P(W).

Scalar ambiguity

Two invertible linear maps A,BA,B induce the same transformation exactly when B=λAB=\lambda A for some λk×\lambda\in k^\times, provided V0V\ne0. Thus the group of projective transformations is

GL(V)/(k×I)=PGL(V),\operatorname{GL}(V)/(k^\times I) =\operatorname{PGL}(V),

the .

Incidence preservation

Projective transformations send projective subspaces to projective subspaces and preserve incidence. A bijection with this line-preserving property is called a collineation. Projective transformations are collineations, but over a field with nontrivial automorphisms there can also be collineations induced by non-linear . The accounts for all of them in projective dimension at least two.

References
  1. Emil Artin, Geometric Algebra, Interscience, 1957. Relevant: Chapter II, §§3–4.
  2. Joe Harris, Algebraic Geometry: A First Course, Springer, 1992. Publisher record. Relevant: Lecture 1, projective linear transformations.