Hilbert's Nullstellensatz

Template:Short description In mathematics, Hilbert's Nullstellensatz (German for "theorem of zeros", or more literally, "zero-locus-theorem") is a theorem that establishes a fundamental relationship between geometry and algebra. This relationship is the basis of algebraic geometry. It relates algebraic sets to ideals in polynomial rings over algebraically closed fields. This relationship was discovered by David Hilbert, who proved the Nullstellensatz in his second major paper on invariant theory in 1893 (following his seminal 1890 paper in which he proved Hilbert's basis theorem).

FormulationEdit

Let <math>k</math> be a field (such as the rational numbers) and <math>K</math> be an algebraically closed field extension of <math>k</math> (such as the complex numbers). Consider the polynomial ring <math>k[X_1, \ldots, X_n]</math> and let <math>I</math> be an ideal in this ring. The algebraic set <math>\mathrm V(I)</math> defined by this ideal consists of all <math>n</math>-tuples <math>\mathbf x = (x_1, \dots, x_n)</math> in <math>K^n</math> such that <math>f(\mathbf x) = 0</math> for all <math>f</math> in Template:Awrap Hilbert's Nullstellensatz states that if p is some polynomial in <math>k[X_1, \ldots, X_n]</math> that vanishes on the algebraic set <math>\mathrm V(I)</math>, i.e. <math>p(\mathbf x) = 0</math> for all <math>\mathbf x</math> in <math>\mathrm V(I)</math>, then there exists a natural number <math>r</math> such that <math>p^r</math> is in <math>I</math>.<ref>Template:Harvnb.</ref>

An immediate corollary is the weak Nullstellensatz: The ideal <math>I \subseteq k[X_1, \ldots, X_n]</math> contains 1 if and only if the polynomials in <math>I</math> do not have any common zeros in Kⁿ. Specializing to the case <math>k=K=\mathbb{C}, n=1</math>, one immediately recovers a restatement of the fundamental theorem of algebra: a polynomial P in <math>\mathbb{C}[X]</math> has a root in <math>\mathbb{C}</math> if and only if deg P ≠ 0. For this reason, the (weak) Nullstellensatz has been referred to as a generalization of the fundamental theorem of algebra for multivariable polynomials.<ref>Template:Cite book</ref> The weak Nullstellensatz may also be formulated as follows: if I is a proper ideal in <math>k[X_1, \ldots, X_n],</math> then V(I) cannot be empty, i.e. there exists a common zero for all the polynomials in the ideal in every algebraically closed extension of k. This is the reason for the name of the theorem, the full version of which can be proved easily from the 'weak' form using the Rabinowitsch trick. The assumption of considering common zeros in an algebraically closed field is essential here; for example, the elements of the proper ideal (X² + 1) in <math>\R[X]</math> do not have a common zero in <math>\R.</math>

With the notation common in algebraic geometry, the Nullstellensatz can also be formulated as

for every ideal J. Here, <math>\sqrt{J}</math> denotes the radical of J and I(U) is the ideal of all polynomials that vanish on the set U.

In this way, taking <math>k = K</math> we obtain an order-reversing bijective correspondence between the algebraic sets in Kⁿ and the radical ideals of <math>K[X_1, \ldots, X_n].</math> In fact, more generally, one has a Galois connection between subsets of the space and subsets of the algebra, where "Zariski closure" and "radical of the ideal generated" are the closure operators.

As a particular example, consider a point <math>P = (a_1, \dots, a_n) \in K^n</math>. Then <math>I(P) = (X_1 - a_1, \ldots, X_n - a_n)</math>. More generally,

<math>\sqrt{I} = \bigcap_{(a_1, \dots, a_n) \in V(I)} (X_1 - a_1, \dots, X_n - a_n).</math>

Conversely, every maximal ideal of the polynomial ring <math>K[X_1,\ldots,X_n]</math> (note that <math>K</math> is algebraically closed) is of the form <math>(X_1 - a_1, \ldots, X_n - a_n)</math> for some <math>a_1,\ldots,a_n \in K</math>.

As another example, an algebraic subset W in Kⁿ is irreducible (in the Zariski topology) if and only if <math>I(W)</math> is a prime ideal.

ProofsEdit

There are many known proofs of the theorem. Some are non-constructive, such as the first one. Others are constructive, as based on algorithms for expressing Template:Math or Template:Math as a linear combination of the generators of the ideal.

Using Zariski's lemmaEdit

Zariski's lemma asserts that if a field is finitely generated as an associative algebra over a field K, then it is a finite field extension of K (that is, it is also finitely generated as a vector space). If <math>\mathfrak{m}</math> is a maximal ideal of <math>K[X_1,\ldots,X_n]</math> for algebraically closed K, then Zariski's lemma implies that <math>K[X_1,\ldots,X_n]/ \mathfrak{m}</math> is a finite field extension of K, and thus, by algebraic closure, must be K. From this, it follows that there is an <math>a = (a_1,\dots,a_n)\in K^n</math> such that <math>X_i-a_i\in\mathfrak{m}</math> for <math>i=1,\dots, n</math>. In other words,

<math>\mathfrak{m} \supseteq \mathfrak{m}_a=(X_1 - a_1, \ldots, X_n - a_n)</math>

for some <math>a = (a_1,\dots,a_n)\in K^n</math>. But <math>\mathfrak{m}_a</math> is clearly maximal, so <math>\mathfrak{m}=\mathfrak{m}_a</math>. This is the weak Nullstellensatz: every maximal ideal of <math>K[X_1,\ldots,X_n]</math> for algebraically closed K is of the form <math>\mathfrak{m}_a=(X_1 - a_1, \ldots, X_n - a_n)</math> for some <math>a = (a_1,\dots,a_n)\in K^n</math>. Because of this close relationship, some texts refer to Zariski's lemma as the weak Nullstellensatz or as the 'algebraic version' of the weak Nullstellensatz.<ref>Template:Cite book</ref><ref>Template:Cite book</ref>

The full Nullstellensatz can also be proved directly from Zariski's lemma without employing the Rabinowitsch trick. Here is a sketch of a proof using this lemma.<ref>Template:Harvnb, Exercise 7.14.</ref>

Let <math>A = K[X_1, \ldots, X_n]</math> (K an algebraically closed field), J an ideal of A, and <math>V=\mathrm{V}(J)</math> the common zeros of J in <math>K^n</math>. Clearly, <math>\sqrt{J} \subseteq \mathrm{I}(V)</math>, where <math>\mathrm{I}(V)</math> is the ideal of polynomials in A vanishing on V. To show the opposite inclusion, let <math>f \not\in \sqrt{J}</math>. Then <math>f \not\in \mathfrak{p}</math> for some prime ideal <math>\mathfrak{p} \supseteq J</math> in A. Let <math>R = (A/\mathfrak{p}) [1/\bar{f}]</math>, where <math>\bar{f}</math>is the image of f under the natural map <math>A \to A/\mathfrak{p}</math>, and <math>\mathfrak{m}</math> be a maximal ideal in R. By Zariski's lemma, <math>R/\mathfrak{m}</math> is a finite extension of K, and thus, is K since K is algebraically closed. Let <math>x_i</math> be the images of <math>X_i</math> under the natural map <math>A \to A/\mathfrak{p}\to R \to R/\mathfrak{m}\cong K</math>. It follows that, by construction, <math>x = (x_1, \ldots, x_n) \in V</math> but <math>f(x) \ne 0</math>, so <math>f \notin \mathrm{I}(V)</math>.

Using resultantsEdit

The following constructive proof of the weak form is one of the oldest proofs (the strong form results from the Rabinowitsch trick, which is also constructive).

The resultant of two polynomials depending on a variable Template:Mvar and other variables is a polynomial in the other variables that is in the ideal generated by the two polynomials, and has the following properties: if one of the polynomials is monic in Template:Mvar, every zero (in the other variables) of the resultant may be extended into a common zero of the two polynomials.

The proof is as follows.

If the ideal is principal, generated by a non-constant polynomial Template:Mvar that depends on Template:Mvar, one chooses arbitrary values for the other variables. The fundamental theorem of algebra asserts that this choice can be extended to a zero of Template:Mvar.

In the case of several polynomials <math>p_1,\ldots, p_n,</math> a linear change of variables allows to suppose that <math>p_1</math> is monic in the first variable Template:Mvar. Then, one introduces <math>n-1</math> new variables <math>u_2, \ldots, u_n,</math> and one considers the resultant

<math>R=\operatorname{Res}_x(p_1,u_2p_2+\cdots +u_np_n).</math>

As Template:Mvar is in the ideal generated by <math>p_1,\ldots, p_n,</math> the same is true for the coefficients in Template:Mvar of the monomials in <math>u_2, \ldots, u_n.</math> So, if Template:Math is in the ideal generated by these coefficients, it is also in the ideal generated by <math>p_1,\ldots, p_n.</math> On the other hand, if these coefficients have a common zero, this zero can be extended to a common zero of <math>p_1,\ldots, p_n,</math> by the above property of the resultant.

This proves the weak Nullstellensatz by induction on the number of variables.

Using Gröbner basesEdit

A Gröbner basis is an algorithmic concept that was introduced in 1973 by Bruno Buchberger. It is presently fundamental in computational geometry. A Gröbner basis is a special generating set of an ideal from which most properties of the ideal can easily be extracted. Those that are related to the Nullstellensatz are the following:

An ideal contains Template:Math if and only if its reduced Gröbner basis (for any monomial ordering) is Template:Math.
The number of the common zeros of the polynomials in a Gröbner basis is strongly related to the number of monomials that are irreducibles by the basis. Namely, the number of common zeros is infinite if and only if the same is true for the irreducible monomials; if the two numbers are finite, the number of irreducible monomials equals the numbers of zeros (in an algebraically closed field), counted with multiplicities.
With a lexicographic monomial order, the common zeros can be computed by solving iteratively univariate polynomials (this is not used in practice since one knows better algorithms).
Strong Nullstellensatz: a power of Template:Mvar belongs to an ideal Template:Mvar if and only the saturation of Template:Mvar by Template:Mvar produces the Gröbner basis Template:Math. Thus, the strong Nullstellensatz results almost immediately from the definition of the saturation.

GeneralizationsEdit

The Nullstellensatz is subsumed by a systematic development of the theory of Jacobson rings, which are those rings in which every radical ideal is an intersection of maximal ideals. Given Zariski's lemma, proving the Nullstellensatz amounts to showing that if k is a field, then every finitely generated k-algebra R (necessarily of the form <math display="inline">R = k[t_1,\cdots,t_n]/I</math>) is Jacobson. More generally, one has the following theorem:

Let <math>R</math> be a Jacobson ring. If <math>S</math> is a finitely generated R-algebra, then <math>S</math> is a Jacobson ring. Furthermore, if <math>\mathfrak{n}\subseteq S</math> is a maximal ideal, then <math>\mathfrak{m} := \mathfrak{n} \cap R</math> is a maximal ideal of <math display="inline">R</math>, and <math>S/\mathfrak{n}</math> is a finite extension of <math>R/\mathfrak{m}</math>.<ref>{{#invoke:citation/CS1|citation

|CitationClass=web }}</ref>

Other generalizations proceed from viewing the Nullstellensatz in scheme-theoretic terms as saying that for any field k and nonzero finitely generated k-algebra R, the morphism <math display="inline">\mathrm{Spec} \, R \to \mathrm{Spec} \, k</math> admits a section étale-locally (equivalently, after base change along <math display="inline">\mathrm{Spec} \, L \to \mathrm{Spec} \, k</math> for some finite field extension <math display="inline">L/k</math>). In this vein, one has the following theorem:

Any faithfully flat morphism of schemes <math display="inline">f: Y \to X</math> locally of finite presentation admits a quasi-section, in the sense that there exists a faithfully flat and locally quasi-finite morphism <math display="inline">g: X' \to X</math> locally of finite presentation such that the base change <math display="inline">f': Y \times_X X' \to X'</math> of <math display="inline">f</math> along <math display="inline">g</math> admits a section.<ref>EGA §IV.17.16.2.</ref> Moreover, if <math display="inline">X</math> is quasi-compact (resp. quasi-compact and quasi-separated), then one may take <math display="inline">X'</math> to be affine (resp. <math display="inline">X'</math> affine and <math display="inline">g</math> quasi-finite), and if <math display="inline">f</math> is smooth surjective, then one may take <math display="inline">g</math> to be étale.<ref>EGA §IV.17.16.3(ii).</ref>

Serge Lang gave an extension of the Nullstellensatz to the case of infinitely many generators:

Let <math display="inline">\kappa</math> be an infinite cardinal and let <math display="inline">K</math> be an algebraically closed field whose transcendence degree over its prime subfield is strictly greater than <math>\kappa</math>. Then for any set <math display="inline">S</math> of cardinality <math display="inline">\kappa</math>, the polynomial ring <math display="inline">A = K[x_i]_{i \in S}</math> satisfies the Nullstellensatz, i.e., for any ideal <math display="inline">J \sub A</math> we have that <math>\sqrt{J} = \hbox{I} (\hbox{V} (J))</math>.<ref>Template:Cite journal</ref>

Effective NullstellensatzEdit

In all of its variants, Hilbert's Nullstellensatz asserts that some polynomial Template:Mvar belongs or not to an ideal generated, say, by Template:Math; we have Template:Math in the strong version, Template:Math in the weak form. This means the existence or the non-existence of polynomials Template:Math such that Template:Math. The usual proofs of the Nullstellensatz are not constructive, non-effective, in the sense that they do not give any way to compute the Template:Math.

It is thus a rather natural question to ask if there is an effective way to compute the Template:Math (and the exponent Template:Mvar in the strong form) or to prove that they do not exist. To solve this problem, it suffices to provide an upper bound on the total degree of the Template:Math: such a bound reduces the problem to a finite system of linear equations that may be solved by usual linear algebra techniques. Any such upper bound is called an effective Nullstellensatz.

A related problem is the ideal membership problem, which consists in testing if a polynomial belongs to an ideal. For this problem also, a solution is provided by an upper bound on the degree of the Template:Math. A general solution of the ideal membership problem provides an effective Nullstellensatz, at least for the weak form.

In 1925, Grete Hermann gave an upper bound for ideal membership problem that is doubly exponential in the number of variables. In 1982 Mayr and Meyer gave an example where the Template:Math have a degree that is at least double exponential, showing that every general upper bound for the ideal membership problem is doubly exponential in the number of variables.

Since most mathematicians at the time assumed the effective Nullstellensatz was at least as hard as ideal membership, few mathematicians sought a bound better than double-exponential. In 1987, however, W. Dale Brownawell gave an upper bound for the effective Nullstellensatz that is simply exponential in the number of variables.<ref>Template:Citation</ref> Brownawell's proof relied on analytic techniques valid only in characteristic 0, but, one year later, János Kollár gave a purely algebraic proof, valid in any characteristic, of a slightly better bound.

In the case of the weak Nullstellensatz, Kollár's bound is the following:<ref>Template:Citation</ref>

Let Template:Math be polynomials in Template:Math variables, of total degree Template:Math. If there exist polynomials Template:Math such that Template:Math, then they can be chosen such that

This bound is optimal if all the degrees are greater than 2.

If Template:Mvar is the maximum of the degrees of the Template:Math, this bound may be simplified to

An improvement due to M. Sombra is<ref>Template:Citation</ref>

His bound improves Kollár's as soon as at least two of the degrees that are involved are lower than 3.

Projective NullstellensatzEdit

We can formulate a certain correspondence between homogeneous ideals of polynomials and algebraic subsets of a projective space, called the projective Nullstellensatz, that is analogous to the affine one. To do that, we introduce some notations. Let <math>R = k[t_0, \ldots, t_n].</math> The homogeneous ideal,

<math>R_+ = \bigoplus_{d \geqslant 1} R_d</math>

is called the maximal homogeneous ideal (see also irrelevant ideal). As in the affine case, we let: for a subset <math>S \subseteq \mathbb{P}^n</math> and a homogeneous ideal I of R,

<math>\begin{align}

\operatorname{I}_{\mathbb{P}^n}(S) &= \{ f \in R_+ \mid f = 0 \text{ on } S \}, \\ \operatorname{V}_{\mathbb{P}^n}(I) &= \{ x \in \mathbb{P}^n \mid f(x) = 0 \text{ for all } f \in I \}. \end{align}</math>

By <math>f = 0 \text{ on } S</math> we mean: for every homogeneous coordinates <math>(a_0 : \cdots : a_n)</math> of a point of S we have <math>f(a_0,\ldots, a_n)=0</math>. This implies that the homogeneous components of f are also zero on S and thus that <math>\operatorname{I}_{\mathbb{P}^n}(S)</math> is a homogeneous ideal. Equivalently, <math>\operatorname{I}_{\mathbb{P}^n}(S)</math> is the homogeneous ideal generated by homogeneous polynomials f that vanish on S. Now, for any homogeneous ideal <math>I \subseteq R_+</math>, by the usual Nullstellensatz, we have:

<math>\sqrt{I} = \operatorname{I}_{\mathbb{P}^n}(\operatorname{V}_{\mathbb{P}^n}(I)),</math>

and so, like in the affine case, we have:<ref>This formulation comes from Milne, Algebraic geometry [1] and differs from Template:Harvnb</ref>

There exists an order-reversing one-to-one correspondence between proper homogeneous radical ideals of R and subsets of <math>\mathbb{P}^n</math> of the form <math>\operatorname{V}_{\mathbb{P}^n}(I).</math> The correspondence is given by <math>\operatorname{I}_{\mathbb{P}^n}</math> and <math>\operatorname{V}_{\mathbb{P}^n}.</math>

Analytic Nullstellensatz (Rückert’s Nullstellensatz)Edit

The Nullstellensatz also holds for the germs of holomorphic functions at a point of complex n-space <math>\Complex^n.</math> Precisely, for each open subset <math>U \subseteq \Complex^n,</math> let <math>\mathcal{O}_{\Complex^n}(U)</math> denote the ring of holomorphic functions on U; then <math>\mathcal{O}_{\Complex^n}</math> is a sheaf on <math>\Complex^n.</math> The stalk <math>\mathcal{O}_{\Complex^n, 0}</math> at, say, the origin can be shown to be a Noetherian local ring that is a unique factorization domain.

If <math>f \in \mathcal{O}_{\Complex ^n, 0}</math> is a germ represented by a holomorphic function <math>\widetilde{f}: U \to \Complex </math>, then let <math>V_0(f)</math> be the equivalence class of the set

<math>\left \{ z \in U \mid \widetilde{f}(z) = 0 \right\},</math>

where two subsets <math>X, Y \subseteq \Complex^n</math> are considered equivalent if <math>X \cap U = Y \cap U</math> for some neighborhood U of 0. Note <math>V_0(f)</math> is independent of a choice of the representative <math>\widetilde{f}.</math> For each ideal <math>I \subseteq \mathcal{O}_{\Complex^n,0},</math> let <math>V_0(I)</math> denote <math>V_0(f_1) \cap \dots \cap V_0(f_r)</math> for some generators <math>f_1, \ldots, f_r</math> of I. It is well-defined; i.e., is independent of a choice of the generators.

For each subset <math>X \subseteq \Complex ^n</math>, let

<math>I_0(X) = \left \{ f \in \mathcal{O}_{\Complex^n,0} \mid V_0(f) \supset X \right \}.</math>

It is easy to see that <math>I_0(X)</math> is an ideal of <math>\mathcal{O}_{\Complex ^n, 0}</math> and that <math>I_0(X) = I_0(Y)</math> if <math>X \sim Y</math> in the sense discussed above.

The analytic Nullstellensatz then states:<ref>Template:Harvnb</ref> for each ideal <math>I \subseteq \mathcal{O}_{\Complex ^n, 0}</math>,

where the left-hand side is the radical of I.

NotesEdit

Template:Reflist