Scheme implementation overview

Various parts of schemes were implemented by Volker Braun, David Joyner, David Kohel, Andrey Novoseltsev, and William Stein.


  • David Kohel (2006-01-03): initial version
  • William Stein (2006-01-05)
  • William Stein (2006-01-20)
  • Andrey Novoseltsev (2010-09-24): update due to addition of toric varieties.
  • Scheme: A scheme whose datatype might not be defined in terms of algebraic equations: e.g. the Jacobian of a curve may be represented by means of a Scheme.

  • AlgebraicScheme: A scheme defined by means of polynomial equations, which may be reducible or defined over a ring other than a field. In particular, the defining ideal need not be a radical ideal, and an algebraic scheme may be defined over Spec(R).

  • AmbientSpaces: Most effective models of algebraic scheme will be defined not by generic gluings, but by embeddings in some fixed ambient space.

  • AffineSpace: Affine spaces and their affine subschemes form the most important universal objects from which algebraic schemes are built. The affine spaces form universal objects in the sense that a morphism is uniquely determined by the images of its coordinate functions and any such images determine a well-defined morphism.

    By default affine spaces will embed in some ordinary projective space, unless it is created as an affine patch of another object.

  • ProjectiveSpace: Projective spaces are the most natural ambient spaces for most projective objects. They are locally universal objects.

  • ProjectiveSpace_ordinary (not implemented) The ordinary projective spaces have the standard weights \([1,..,1]\) on their coefficients.

  • ProjectiveSpace_weighted (not implemented): A special subtype for non-standard weights.

  • ToricVariety: Toric varieties are (partial) compactifications of algebraic tori \(\left(\CC^*\right)^n\) compatible with torus action. Affine and projective spaces are examples of toric varieties, but it is not envisioned that these special cases should inherit from ToricVariety.

  • AlgebraicScheme_subscheme_affine: An algebraic scheme defined by means of an embedding in a fixed ambient affine space.

  • AlgebraicScheme_subscheme_projective: An algebraic scheme defined by means of an embedding in a fixed ambient projective space.

  • QuasiAffineScheme (not yet implemented): An open subset \(U = X \setminus Z\) of a closed subset \(X\) of affine space; note that this is mathematically a quasi-projective scheme, but its ambient space is an affine space and its points are represented by affine rather than projective points.


    AlgebraicScheme_quasi is implemented, as a base class for this.

  • QuasiProjectiveScheme (not yet implemented): An open subset of a closed subset of projective space; this datatype stores the defining polynomial, polynomials, or ideal defining the projective closure \(X\) plus the closed subscheme \(Z\) of \(X\) whose complement \(U = X \setminus Z\) is the quasi-projective scheme.


    The quasi-affine and quasi-projective datatype lets one create schemes like the multiplicative group scheme \(\mathbb{G}_m = \mathbb{A}^1\setminus \{(0)\}\) and the non-affine scheme \(\mathbb{A}^2\setminus \{(0,0)\}\). The latter is not affine and is not of the form \(\mathrm{Spec}(R)\).


  • PointSets and points over a ring: For algebraic schemes \(X/S\) and \(T/S\) over \(S\), one can form the point set \(X(T)\) of morphisms from \(T\to X\) over \(S\).

    sage: PP.<X,Y,Z> = ProjectiveSpace(2, QQ)
    sage: PP
    Projective Space of dimension 2 over Rational Field

    The first line is an abuse of language – returning the generators of the coordinate ring by gens().

    A projective space object in the category of schemes is a locally free object – the images of the generator functions locally determine a point. Over a field, one can choose one of the standard affine patches by the condition that a coordinate function \(X_i \ne 0\)

    sage: PP(QQ)
    Set of rational points of Projective Space
    of dimension 2 over Rational Field
    sage: PP(QQ)([-2,3,5])
    (-2/5 : 3/5 : 1)

    Over a ring, this is not true, e.g. even over an integral domain which is not a PID, there may be no single affine patch which covers a point.

    sage: R.<x> = ZZ[]
    sage: S.<t> = R.quo(x^2+5)
    sage: P.<X,Y,Z> = ProjectiveSpace(2, S)
    sage: P(S)
    Set of rational points of Projective Space of dimension 2 over
    Univariate Quotient Polynomial Ring in t over Integer Ring with
    modulus x^2 + 5

    In order to represent the projective point \((2:1+t) = (1-t:3)\) we note that the first representative is not well-defined at the prime \(pp = (2,1+t)\) and the second element is not well-defined at the prime \(qq = (1-t,3)\), but that \(pp + qq = (1)\), so globally the pair of coordinate representatives is well-defined.

    sage: P( [2, 1+t] )
    (2 : t + 1 : 1)

    In fact, we need a test R.ideal([2,1+t]) == R.ideal([1]) in order to make this meaningful.