04.02.07 · algebraic-geometry / schemes

Nullstellensatz and dimension theory

shipped3 tiersLean: none

Anchor (Master): Hilbert 1893 (originator); Krull 1928 (Hauptidealsatz); Zariski 1947 (modern Zariski-lemma proof); Eisenbud *Commutative Algebra with a View Toward Algebraic Geometry* Ch. 8–13; Matsumura *Commutative Ring Theory* Ch. 5, 14, 17; Stacks Project Tags 00FV (Nullstellensatz), 00KW (Krull dimension), 00KZ (Hauptidealsatz)

Intuition [Beginner]

Algebraic geometry begins with a dictionary. On one side: systems of polynomial equations. On the other: their common solution sets. The Nullstellensatz, Hilbert's theorem of zeros from 1893, says the dictionary is honest — for an algebraically closed field like the complex numbers, an ideal of polynomials and the geometric set it cuts out determine each other up to a precise correction. That correction is the radical of the ideal: the set of polynomials some power of which lies in the ideal. Once you account for it, ideals and their zero sets are two sides of the same coin.

Two payoff statements live inside this dictionary. First, a consistent system of polynomial equations — one whose ideal is a proper subset of the polynomial ring — always has a solution over an algebraically closed field. There is no way to write down equations that secretly have no answer. Second, every maximal ideal in $k [x_{1}, \dots, x_{n}]$ corresponds to a single point of $k^{n}$ — the most rigid algebraic structure matches the most basic geometric structure.

Once geometry and algebra are paired, you can ask how big the geometric object is. The dimension of an affine variety has three matching descriptions: the largest number of independent coordinates you can vary on it, the longest chain of nested irreducible subvarieties inside it, and the longest chain of prime ideals in its coordinate ring. They agree, and the agreement is the start of dimension theory.

Visual [Beginner]

Two parallel pictures. Left: an ideal in $k [x, y]$ generated by $y - x^{2}$ and $y$ . Right: the common solution set in $k^{2}$ , the single point at the origin. The Nullstellensatz says one side recovers the other up to taking radicals.

Worked example [Beginner]

Take the ideal $a = (x^{2}, x y)$ inside $C [x, y]$ . Its zero set $V (a)$ in $C^{2}$ is the set of points $(a, b)$ where $a^{2} = 0$ and $ab = 0$ — exactly the line $a = 0$ , that is, the $y$ -axis. The defining ideal of that line — the polynomials vanishing on the whole $y$ -axis — is $(x)$ , the principal ideal generated by $x$ .

So we have $V (a) = V (x)$ even though $a = (x^{2}, x y) \neq = (x)$ . The Nullstellensatz says these two ideals share the same zero set precisely because their radicals agree. And indeed $x^{2} \in a$ and $x y = x \cdot y \in a$ , so $x \in a$ ; conversely, $(x) \supseteq a$ contains $a$ , so $(x) = (x) \supseteq a$ . Therefore $a = (x)$ , and the dictionary checks out.

Now the dimension count. The $y$ -axis is a one-dimensional variety: you can parametrise it by the single coordinate $y$ , so it has one degree of freedom. On the algebra side, the coordinate ring is $C [x, y] / (x) = C [y]$ , a polynomial ring in one variable. Its prime ideals form the chain $(0) \subset (y - c)$ for any $c \in C$ , of length one. Its function field is $C (y)$ , with transcendence degree one over $C$ . All three counts give the same number.

What this tells us. The Nullstellensatz is the bridge that makes algebraic geometry possible: you can compute on either side of the dictionary and translate the answer. Dimension is the first invariant that the dictionary lets you read off, and the three descriptions agree precisely because the dictionary is honest.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Fix an algebraically closed field $k$ and the polynomial ring $R = k [x_{1}, \dots, x_{n}]$ . Two operations link ideals of $R$ to subsets of $k^{n}$ . For an ideal $a \subseteq R$ , the vanishing locus is

V (a) = {a = (a_{1}, \dots, a_{n}) \in k^{n} : f (a) = 0 for all f \in a} .

For a subset $S \subseteq k^{n}$ , the vanishing ideal is

I (S) = {f \in R : f (a) = 0 for all a \in S} .

The radical of an ideal $a$ is $a = {f \in R : f^{N} \in a for some N \geq 1}$ . An ideal is radical if $a = a$ .

Theorem (Hilbert's Nullstellensatz, three forms). Over an algebraically closed field $k$ and $R = k [x_{1}, \dots, x_{n}]$ :

(Weak form.) For every proper ideal $a ⊊ R$ , the locus $V (a) \subseteq k^{n}$ is non-empty.

(Strong form.) For every ideal $a \subseteq R$ , the equality $I (V (a)) = a$ holds. Equivalently, $f \in I (V (a))$ if and only if some power $f^{N}$ lies in $a$ .

(Maximal-ideal form.) Every maximal ideal $m \subset R$ has the form $m_{a} = (x_{1} - a_{1}, \dots, x_{n} - a_{n})$ for a unique point $a = (a_{1}, \dots, a_{n}) \in k^{n}$ .

The three forms are equivalent over an algebraically closed field. The maximal-ideal form is most often the working version: it identifies the closed points of $Spec (R)$ with $k^{n}$ and gives a clean correspondence between radical ideals and Zariski-closed subsets of $k^{n}$ .

Definition (Krull dimension). For a commutative ring $R$ , the Krull dimension $dim R$ is the supremum of lengths of chains of prime ideals $p_{0} ⊊ p_{1} ⊊ \dots ⊊ p_{d}$ in $R$ . The height of a prime $p$ is $dim R_{p}$ , equivalently the supremum of lengths of chains of primes ending at $p$ .

Definition (transcendence-degree dimension). Let $X \subseteq k^{n}$ be an irreducible affine variety with coordinate ring $k [X] = R / I (X)$ and function field $k (X) = Frac (k [X])$ . The transcendence-degree dimension of $X$ is $tr.deg_{k} k (X)$ — the size of any transcendence basis of $k (X)$ over $k$ .

Theorem (equivalence of dimension invariants). For a finitely generated $k$ -algebra $A = k [X]$ with $X$ irreducible over the algebraically closed field $k$ :

dim X = dim_{Krull} A = tr.deg_{k} Frac (A) .

Moreover, for a maximal ideal $m \subset A$ corresponding to a closed point of $X$ :

dim A_{m} = dim A

when $A$ is the coordinate ring of an irreducible affine variety; the Krull dimension is the same at every closed point.

Theorem (Krull's Hauptidealsatz, 1928). Let $R$ be a noetherian ring and $f \in R$ . Every minimal prime $p$ over the principal ideal $(f)$ has height $ht (p) \leq 1$ . If $f$ is a non-zero-divisor and not a unit, $ht (p) = 1$ . More generally, every minimal prime over an ideal $(f_{1}, \dots, f_{r}) \subseteq R$ has height at most $r$ .

In geometric language, intersecting an irreducible variety $X$ with the zero locus of a single equation $f = 0$ either gives back $X$ (when $f$ vanishes identically) or cuts the dimension down by at most one. Cutting with $r$ equations can drop dimension by at most $r$ .

Counterexamples to common slips

The Nullstellensatz fails over non-algebraically-closed fields. Over $R$ , the ideal $(x^{2} + 1) \subset R [x]$ is proper but $V (x^{2} + 1) = \emptyset$ in $R$ .
The strong form requires radical, not just inclusion: $V ((x^{2})) = V ((x)) = {0} \subset A^{1}$ , and $I ({0}) = (x) = (x^{2}) \neq = (x^{2})$ .
Krull dimension is not always finite. The ring $k [x_{1}, x_{2}, \dots]$ in countably many variables has infinite Krull dimension. Even noetherian rings can have infinite Krull dimension — Nagata exhibited such a ring in 1962.
For non-irreducible varieties the equality $dim A_{m} = dim A$ fails: the ring $k [x, y] / (x y)$ has two minimal primes corresponding to the two coordinate axes, both of height $0$ but each contributing dimension $1$ .
The Hauptidealsatz needs the noetherian hypothesis. In a non-noetherian ring, principal ideals can have minimal primes of arbitrarily high height.

Key theorem with proof [Intermediate+]

Theorem (Strong Nullstellensatz via the Rabinowitsch trick). Let $k$ be algebraically closed, $R = k [x_{1}, \dots, x_{n}]$ , $a \subseteq R$ an ideal, and $f \in R$ . If $f$ vanishes on $V (a) \subseteq k^{n}$ , then $f^{N} \in a$ for some $N \geq 1$ .

The argument splits in two: first prove the weak form (every proper ideal has a zero), then derive the strong form by Rabinowitsch's auxiliary-variable trick.

Step 1 (Weak Nullstellensatz via the Zariski lemma). If $a ⊊ R$ is a proper ideal, then $V (a) \neq = \emptyset$ .

Proof of Step 1. Pick a maximal ideal $m$ containing $a$ , which exists by Zorn's lemma since $a$ is proper. The quotient $K = R / m$ is a field, and $K$ is finitely generated as a $k$ -algebra because $R$ is.

Apply the Zariski lemma: if $K$ is a field finitely generated as a $k$ -algebra, then $K$ is a finite algebraic extension of $k$ . Since $k$ is algebraically closed, $K = k$ .

The composite $R \to K = k$ is a $k$ -algebra homomorphism, so it sends each $x_{i}$ to some $a_{i} \in k$ . Set $a = (a_{1}, \dots, a_{n}) \in k^{n}$ . Then the kernel of $R \to k$ is exactly $m_{a} = (x_{1} - a_{1}, \dots, x_{n} - a_{n})$ , hence $m = m_{a}$ and $a \in V (m) \subseteq V (a)$ . $□$

The Zariski lemma itself is proved by Noether normalisation: a finitely generated $k$ -algebra that is also a domain admits algebraically independent elements $y_{1}, \dots, y_{d}$ such that the algebra is module-finite over $k [y_{1}, \dots, y_{d}]$ . If the algebra is a field, the polynomial ring $k [y_{1}, \dots, y_{d}]$ must also be a field — which forces $d = 0$ , so the algebra is module-finite over $k$ , hence algebraic. The reference is Atiyah-Macdonald §5 ^{[Atiyah-Macdonald Ch. 5]} and Eisenbud §13 ^{[Eisenbud §13]}.

Step 2 (strong form via Rabinowitsch). Suppose $f \in R$ vanishes on $V (a)$ . Introduce a new variable $y$ and form the ideal

b = (a, 1 - y f) \subseteq R [y] = k [x_{1}, \dots, x_{n}, y] .

Consider the locus $V (b) \subseteq k^{n + 1}$ . A point $(a_{1}, \dots, a_{n}, b)$ lies in $V (b)$ if and only if $(a_{1}, \dots, a_{n}) \in V (a)$ and $1 - b f (a) = 0$ . Since $f$ vanishes on $V (a)$ by hypothesis, $f (a) = 0$ at every $a \in V (a)$ , and the equation $1 - b f (a) = 1 = 0$ has no solution. Therefore $V (b) = \emptyset$ .

By the weak form (Step 1), $b$ is not proper, so $b = R [y]$ . Write $1$ as a $R [y]$ -linear combination of generators:

1 = i \sum h_{i} (x, y) g_{i} (x) + h_{0} (x, y) (1 - y f (x)),

where $g_{1}, \dots, g_{m}$ generate $a$ and the $h_{i}$ lie in $R [y]$ .

Now substitute $y = 1/ f$ — formally, work in the localisation $R [1/ f]$ . The factor $(1 - y f)$ becomes $0$ , so

1 = i \sum h_{i} (x, 1/ f) g_{i} (x) in R [1/ f] .

Multiply through by a high enough power $f^{N}$ to clear all denominators on the right. The result is an identity $f^{N} = \sum_{i} \tilde{h}_{i} (x) g_{i} (x)$ in $R$ , where $\tilde{h}_{i} (x) = f^{N_{i}} h_{i} (x, 1/ f) \in R$ for sufficiently large $N$ . Thus $f^{N} \in a$ . $□$

The Rabinowitsch trick converts vanishing on the zero set to algebraic membership in the ideal. The trick is a recurrent technique in commutative algebra — adjoining a free variable to invert an element creates an empty locus precisely when the element vanishes everywhere on the original locus.

Bridge. This dictionary identifies radical ideals with Zariski-closed sets and builds toward 04.02.04 (morphism of schemes), where the same correspondence becomes the contravariant equivalence between affine schemes and commutative rings, and appears again in 04.05.01 (Weil divisor) where height-one primes correspond to codimension-one irreducible subvarieties via the Hauptidealsatz. The dimension equivalences put down here generalise to schemes through Krull-dimension localisation, and dimension theory is dual to depth theory — Cohen-Macaulay rings are exactly those for which the two agree. Putting these together, the foundational reason a noetherian local ring has a finite, well-defined dimension is that its maximal ideal can be cut down to a parameter system whose length is exactly the Krull dimension, and that length is precisely what the Hauptidealsatz controls.

Exercises [Intermediate+]

Exercise 1 (easy, symbolic).

Compute $a$ for $a = (x^{3}, x^{2} y, x y^{2}) \subseteq k [x, y]$ . Then identify $V (a) \subseteq k^{2}$ and confirm $I (V (a)) = a$ .

Hint

Each generator is $x^{a} y^{b}$ with $a \geq 1$ , so the ideal contains $x^{k}$ for some $k$ but not necessarily $x$ or $y$ directly. Look at which monomials are forced into the radical.

Answer

The ideal $a$ contains $x^{3}$ and $x y^{2}$ . From $x^{3} \in a$ and the radical's defining property, $x \in a$ . From $x y^{2} \in a$ and $x \in a$ alone we cannot deduce $y \in a$ — but $V (a) = {(a, b) \in k^{2} : a^{3} = a^{2} b = a b^{2} = 0}$ , which forces $a = 0$ , so $V (a) = {(0, b) : b \in k}$ is the $y$ -axis.

The ideal of the $y$ -axis is $(x)$ . Therefore $I (V (a)) = (x)$ , and the strong Nullstellensatz says $a = (x)$ . Direct check: every generator of $a$ has $x$ as a factor, so $a \subseteq (x)$ and hence $a \subseteq (x)$ ; conversely $x \in a$ since $x^{3} \in a$ , giving the reverse inclusion. $a = (x)$ .

Exercise 3 (medium, symbolic).

Show that the ideal of the twisted cubic in $A_{k}^{3}$ is $I = (y - x^{2}, z - x^{3})$ — verifying this is the ideal of the variety, not merely a defining one.

Hint

Show the quotient $k [x, y, z] / (y - x^{2}, z - x^{3})$ is isomorphic to $k [x]$ , hence a domain. The radical of a prime ideal is itself.

Answer

Set $J = (y - x^{2}, z - x^{3})$ . Define the $k$ -algebra homomorphism $ϕ : k [x, y, z] \to k [t]$ by $ϕ (x) = t$ , $ϕ (y) = t^{2}$ , $ϕ (z) = t^{3}$ . Then $ϕ (y - x^{2}) = t^{2} - t^{2} = 0$ and $ϕ (z - x^{3}) = t^{3} - t^{3} = 0$ , so $J \subseteq ker ϕ$ .

Conversely, given $f \in ker ϕ$ , use the relations $y = x^{2} + (y - x^{2})$ and $z = x^{3} + (z - x^{3})$ to rewrite $f$ modulo $J$ as a polynomial in $x$ alone. If this $\overset{ˉ}{f} \in k [x]$ satisfies $ϕ (\overset{ˉ}{f}) = \overset{ˉ}{f} (t) = 0$ , then $\overset{ˉ}{f} = 0$ in $k [x]$ , so $f \in J$ . Therefore $J = ker ϕ$ , the quotient $k [x, y, z] / J ≅ k [t]$ is a domain, and $J$ is prime, hence radical.

By the strong Nullstellensatz, $J = J = I (V (J)) = I (C)$ . The image of $ϕ$ is exactly the coordinate ring of $C$ , confirming $J$ is the ideal of the variety.

Exercise 4 (medium, numeric).

Let $X = V (x z - y^{2}, x^{3} - y z) \subseteq A_{k}^{3}$ . Determine $dim X$ and write down a chain of prime ideals of maximal length in the coordinate ring.

Hint

The variety is the closure of the rational normal curve image of the parametrisation $t \mapsto (t^{3}, t^{4}, t^{5})$ . Compare with the twisted cubic.

Answer

$1$ . Define $ϕ : k [x, y, z] \to k [t]$ by $x \mapsto t^{3}$ , $y \mapsto t^{4}$ , $z \mapsto t^{5}$ . Then $ϕ (x z - y^{2}) = t^{8} - t^{8} = 0$ and $ϕ (x^{3} - y z) = t^{9} - t^{9} = 0$ , so the image lies in $V (x z - y^{2}, x^{3} - y z)$ and $X$ contains the curve ${(t^{3}, t^{4}, t^{5})}$ .

The image of $ϕ$ is the subring $k [t^{3}, t^{4}, t^{5}] \subseteq k [t]$ , a one-dimensional domain. After taking the radical of the defining ideal we get exactly the ideal $ker ϕ$ , so $dim X = dim k [t^{3}, t^{4}, t^{5}] = 1$ .

A chain of prime ideals of length $1$ in $k [t^{3}, t^{4}, t^{5}]$ is $(0) ⊊ (t^{3} - c, t^{4} - c^{4/3}, t^{5} - c^{5/3})$ for some non-zero $c \in k$ — that is, the prime corresponding to a closed point on the curve. No longer chain exists because the dimension is $1$ .

Exercise 6 (hard, proof).

Prove Krull's Hauptidealsatz for principal ideals: in a noetherian ring $R$ , every minimal prime $p$ over a principal ideal $(f)$ has height at most $1$ .

Hint

Localise at $p$ to reduce to a noetherian local ring with a single maximal ideal. Use symbolic powers and the descending chain condition on a quotient ring.

Answer

Replace $R$ by $R_{p}$ , so we may assume $R$ is a noetherian local ring with maximal ideal $p$ and $p$ is the unique minimal prime over $(f)$ . We must show every prime $q ⊊ p$ has $ht (q) = 0$ , i.e., $q$ is a minimal prime of $R$ .

Suppose for contradiction $q_{1} ⊊ q_{2} ⊊ p$ is a chain of primes of length $2$ . Pass to $R / q_{1}$ , again a noetherian local ring, where $p / q_{1}$ is the unique minimal prime over $(\overset{ˉ}{f})$ . So we may assume $R$ is also a domain (replacing $q_{1}$ by $(0)$ ).

In this local domain $R$ with $p$ minimal over $(f)$ , consider the ring $\overset{ˉ}{R} = R / (f)$ . Since $p$ is the unique minimal prime of $\overset{ˉ}{R}$ , the ring $\overset{ˉ}{R}$ has $Spec (\overset{ˉ}{R})$ a single point with $p / (f)$ as nilradical, so $\overset{ˉ}{R}$ is artinian (a noetherian ring with finitely many primes, all maximal, has the descending chain condition).

For each $n \geq 1$ , define the symbolic power $q_{2}^{(n)} = q_{2}^{n} R_{q_{2}} \cap R$ , the saturation of $q_{2}^{n}$ at $q_{2}$ . The chain $q_{2}^{(1)} \supseteq q_{2}^{(2)} \supseteq \dots$ has images in $\overset{ˉ}{R}$ which descend, hence stabilise: $(q_{2}^{(n)} + (f)) / (f) = (q_{2}^{(n + 1)} + (f)) / (f)$ for some $n$ , equivalently $q_{2}^{(n)} \subseteq q_{2}^{(n + 1)} + (f)$ .

Now $f \in / q_{2}$ (else $p$ would not be minimal over $(f)$ — a contradiction since $q_{2} ⊊ p$ ). Working in $R_{q_{2}}$ , the element $f$ is a unit, so $q_{2}^{(n)} R_{q_{2}} \subseteq q_{2}^{(n + 1)} R_{q_{2}}$ , forcing equality $q_{2}^{n} R_{q_{2}} = q_{2}^{n + 1} R_{q_{2}}$ . By Nakayama's lemma applied to the noetherian local ring $R_{q_{2}}$ , this gives $q_{2}^{n} R_{q_{2}} = 0$ , hence $q_{2} R_{q_{2}} = 0$ since $R_{q_{2}}$ is a domain — contradicting $q_{1} ⊊ q_{2}$ .

So no length- $2$ chain $q_{1} ⊊ q_{2} ⊊ p$ exists, hence $ht (p) \leq 1$ . $□$

Exercise 7 (hard, proof).

Show that for an irreducible affine variety $X$ over algebraically closed $k$ , the Krull dimension of the coordinate ring $A = k [X]$ equals the transcendence degree $tr.deg_{k} k (X)$ .

Hint

Use Noether normalisation to reduce to the polynomial-ring case, then combine the going-up theorem of Cohen-Seidenberg with the dimension formula for module-finite extensions.

Answer

Apply Noether normalisation to the finitely generated $k$ -algebra $A$ : there exist algebraically independent elements $y_{1}, \dots, y_{d} \in A$ over $k$ such that the inclusion $k [y_{1}, \dots, y_{d}] ↪ A$ is module-finite. The transcendence degree of $Frac (A)$ over $k$ equals $d$ , by definition of algebraic independence and the inclusion $k (y_{1}, \dots, y_{d}) \subseteq Frac (A)$ being algebraic (a module-finite extension is algebraic on fraction fields).

The polynomial ring $B = k [y_{1}, \dots, y_{d}]$ has Krull dimension $d$ — exhibit the chain $(0) ⊊ (y_{1}) ⊊ (y_{1}, y_{2}) ⊊ \dots ⊊ (y_{1}, \dots, y_{d})$ — and the upper bound $d$ follows from the Hauptidealsatz applied iteratively.

For the module-finite extension $B ↪ A$ , the going-up theorem of Cohen-Seidenberg lifts every chain of primes in $B$ to a chain in $A$ , and the incomparability of primes in module-finite extensions ensures lifted chains have the same length. The going-down theorem (when $B$ is normal, which it is for a polynomial ring) lifts chains in the other direction. Together, $dim A = dim B = d = tr.deg_{k} Frac (A)$ . $□$

Exercise 8 (hard, conceptual).

State the dimension formula for catenary rings and explain how it sharpens the equivalence of dimension invariants.

Hint

A noetherian ring is catenary when every two saturated chains of primes between the same endpoints have the same length. The dimension formula relates $ht (p) + dim (R / p)$ to $dim R$ .

Answer

A noetherian ring $R$ is catenary if for every pair of primes $p \subseteq q$ in $R$ , all saturated chains of primes from $p$ to $q$ have the same length. $R$ is universally catenary if every finitely generated $R$ -algebra is catenary.

For a universally catenary noetherian local domain $R$ with $p$ a prime, the dimension formula reads

ht (p) + dim (R / p) = dim R .

Geometrically, for an irreducible affine variety $X$ and an irreducible closed subvariety $Y \subseteq X$ , the codimension of $Y$ in $X$ plus the dimension of $Y$ equals the dimension of $X$ . This is the formula students expect from intuition — but it requires the catenary hypothesis. Nagata exhibited a noetherian local domain with a non-catenary chain in which the formula fails.

Finitely generated algebras over a field are universally catenary (Eisenbud Ch. 13), so for the affine varieties in this unit the formula applies without comment. The catenary hypothesis is the structural condition under which Krull dimension behaves additively over inclusion of closed subvarieties — a refinement of the equivalence-of-dimensions theorem that promotes the equality of three numbers to a coherent dimension theory across a stratification.

Lean formalization [Intermediate+]

lean_status: none — Mathlib has the Hilbert basis theorem, the Nullstellensatz over $C$ via the Zariski lemma, and Krull-dimension predicates, but the equivalence of the three classical dimension invariants and the Hauptidealsatz in its general form are not yet assembled.

[object Promise]

Advanced results [Master]

Going-up and going-down (Cohen-Seidenberg). Let $R \subseteq S$ be an integral extension of rings. Lying over: every prime $p \subset R$ has a prime $q \subset S$ with $q \cap R = p$ . Incomparability: if $q_{1} ⊊ q_{2}$ in $S$ then their contractions in $R$ are distinct. Going-up: for primes $p_{1} \subseteq p_{2}$ in $R$ and $q_{1}$ in $S$ with $q_{1} \cap R = p_{1}$ , there exists $q_{2} \supseteq q_{1}$ with $q_{2} \cap R = p_{2}$ . Going-down (when $R$ is a normal domain and $S$ is a domain): the analogous downward lifting. These theorems were established by Cohen and Seidenberg in the 1940s and form the structural foundation underneath the equivalence of Krull and transcendence-degree dimension.

Catenary and universally catenary rings. A noetherian ring $R$ is catenary if every saturated chain of primes between any two primes has the same length, and universally catenary if every finitely generated $R$ -algebra is catenary. Finitely generated algebras over a field, complete noetherian local rings, and Cohen-Macaulay rings are universally catenary. Nagata 1956 exhibited a noetherian local domain that is not catenary, showing the hypothesis is genuinely needed for the dimension formula $ht (p) + dim (R / p) = dim R$ .

Cohen-Macaulay rings. A noetherian local ring $(R, m)$ is Cohen-Macaulay when its depth equals its Krull dimension. Equivalently, every system of parameters is a regular sequence — generators of an $m$ -primary ideal that act non-zero-divisor-by-non-zero-divisor on each successive quotient. Cohen-Macaulay rings are universally catenary; their unmixedness theorem says every minimal prime over an ideal generated by a regular sequence has the same height. Hochster and Huneke developed tight closure as a structural extension of Cohen-Macaulay theory in characteristic $p$ .

Krull's intersection theorem. For a noetherian local ring $(R, m)$ and any ideal $I$ with $I \subseteq m$ , the intersection $⋂_{n \geq 1} I^{n}$ vanishes when $I = m$ and $R$ is a domain — and more generally equals the kernel of the $I$ -adic completion map. The proof goes through the Artin-Rees lemma. Krull's intersection theorem is the structural input that lets formal power series rings $R_{m}$ act as "infinitesimal completions" of $R$ at $m$ .

Hilbert-Samuel polynomial. For a noetherian local ring $(R, m)$ and an $m$ -primary ideal $q$ , the function $n \mapsto length (R / q^{n})$ agrees with a polynomial $P_{q} (n)$ for $n ≫ 0$ . Its degree equals $dim R$ , and the leading coefficient times $dim (R)!$ is the multiplicity $e_{q} (R)$ . The dimension of $R$ admits three equivalent descriptions: Krull dimension, $de g P_{q}$ , and the minimum number of generators of an $m$ -primary ideal — the "system-of-parameters" definition.

Effective Nullstellensatz. For an algebraically closed field $k$ and polynomials $f, g_{1}, \dots, g_{m} \in k [x_{1}, \dots, x_{n}]$ with $max de g (g_{i}) \leq d$ and $f$ vanishing on the common zero set of the $g_{i}$ , there exist polynomials $h_{i}$ and an exponent $N$ with $f^{N} = \sum h_{i} g_{i}$ and explicit degree bounds. Brownawell 1987 proved $N \leq n d^{n}$ in characteristic zero; Kollár 1988 sharpened to the optimal $N \leq d^{n}$ for $d \geq 3$ , with a separate sharp bound at $d = 2$ . The effective bounds make the Nullstellensatz a quantitative tool in arithmetic geometry and complexity theory.

Real Nullstellensatz and Positivstellensatz. Stengle 1974 and independently Krivine 1964 proved a Nullstellensatz for the real-field case, replacing radicals by real radicals and using sums of squares. Stengle's Positivstellensatz characterises polynomials non-negative (resp. positive, vanishing) on a basic semialgebraic set ${g_{1} \geq 0, \dots, g_{m} \geq 0}$ via cone identities involving sums of squares. This is the foundation of real algebraic geometry — the Tarski-Seidenberg quantifier-elimination theorem and the modern theory of polynomial optimisation rest on it.

Arithmetic Nullstellensatz over $Z$ . A version over $Z$ replaces "no zeros over algebraic closure" by "the ideal contains $1$ in $Z [x_{1}, \dots, x_{n}]$ , possibly after passing to a localisation." The arithmetic Nullstellensatz of Masser-Wüstholz, Krick-Pardo and others gives explicit bounds combining the polynomial degrees and the heights of coefficients.

Model-theoretic content: quantifier elimination. The theory $ACF_{p}$ of algebraically closed fields of characteristic $p$ is complete and model-complete, and admits quantifier elimination. A first-order formula in $ACF$ is equivalent to a quantifier-free formula. The Nullstellensatz is the algebraic content: the set of points where a system of polynomial equations has a solution is itself defined by a system of polynomial equations — no genuine projection / existential quantifier is needed once the field is algebraically closed. The Ax-Grothendieck theorem (every injective polynomial map $C^{n} \to C^{n}$ is surjective) and Chevalley's theorem on constructible sets are downstream consequences.

Stillman's conjecture (Ananyan-Hochster theorem). Stillman conjectured in the early 2000s that the projective dimension of a homogeneous ideal in a polynomial ring is bounded by a function of the number and degrees of generators alone, independent of the ambient number of variables. Ananyan and Hochster 2020 proved the conjecture using small-subalgebra arguments, providing universal bounds. The result reframes ideal-theoretic dimension invariants as functions of the generator data, decoupled from the ambient ring size.

Synthesis. The Nullstellensatz is a translation theorem; dimension theory is what the translated language can say. The three forms — weak, strong ( $I (V (a)) = a$ ), maximal-ideal — are equivalent reformulations of the same data, each tuned to a different working setting. The Rabinowitsch trick is the bridge that makes weak ⇒ strong: introducing an auxiliary variable converts vanishing-on-the-zero-set into emptiness-of-an-extended-locus. Putting these together with Krull's Hauptidealsatz, the central insight is that codimension is genuinely well-defined: cutting an irreducible variety by a single non-vanishing equation drops dimension by exactly one, and dimension itself admits a triple description as Krull dimension, transcendence degree of the function field, and Hilbert-Samuel polynomial degree. This trichotomy generalises to noetherian local rings via the system-of-parameters characterisation, identifies dimension theory with the spectral theory of the Zariski topology, and is dual to depth theory through the Cohen-Macaulay condition. Dimension is an instance of the deeper pattern that algebraic and geometric structure on a variety are interchangeable through the coordinate-ring functor — a pattern that appears again in 04.05.01 (Weil divisor) when codimension-one subvarieties are matched against height-one primes, and in 04.02.04 (morphism of schemes) when morphisms of varieties become ring maps in the opposite direction.

Full proof set [Master]

The strong Nullstellensatz via the Rabinowitsch trick is proved in the key-theorem section. Detailed proofs of Krull's Hauptidealsatz (Exercise 6 sketches the principal-ideal case; the multivariable case follows by induction using minimal-prime decomposition), the equivalence of Krull, transcendence-degree, and Hilbert-Samuel-polynomial dimension for finitely generated algebras over a field (Exercise 7 for the first equivalence; the Hilbert-Samuel agreement via the standard noetherian-local argument with system of parameters), Cohen-Seidenberg going-up and going-down, and the dimension formula for universally catenary rings are referenced to Atiyah-Macdonald §11 ^{[Atiyah-Macdonald Ch. 11]}, Eisenbud Ch. 8–13 ^{[Eisenbud Ch. 8–13]}, and Matsumura Ch. 14, 17 ^{[Matsumura Ch. 14]}. The Brownawell-Kollár effective Nullstellensatz, Stengle's Positivstellensatz, and the Ananyan-Hochster theorem are stated without proof — see the bibliography for the original papers.

Connections [Master]

Affine scheme 04.02.02 — the Nullstellensatz makes the contravariant functor $Spec$ from commutative rings to affine schemes an equivalence on the affine-variety side, identifying radical ideals of $k [x_{1}, \dots, x_{n}]$ with closed subschemes of $A_{k}^{n}$ and maximal ideals with closed points.
Morphism of schemes 04.02.04 — the maximal-ideal correspondence is the reason morphisms of affine varieties over an algebraically closed field correspond to $k$ -algebra homomorphisms in the opposite direction, generalising to the anti-equivalence between commutative rings and affine schemes.
Weil divisor 04.05.01 — Krull's Hauptidealsatz is the structural input that makes Weil divisors well-defined: codimension-one irreducible subvarieties of an irreducible variety correspond exactly to height-one primes in the coordinate ring, and the divisor of a rational function counts vanishing orders along these height-one primes.
Smooth, étale, and unramified morphisms 04.02.05 — dimension theory underwrites the smoothness criterion, where a morphism $f : X \to Y$ is smooth of relative dimension $n$ when its fibres are regular of dimension $n$ . The cotangent-rank criterion ( $Ω_{X / Y}^{1}$ locally free of rank $n$ ) compares against the dimension of the fibre via the Jacobian.
Sheaf of differentials 04.08.01 — the dimension of a smooth variety equals the rank of its sheaf of Kähler differentials, tying the abstract dimension invariants of this unit to the geometric cotangent calculus on a variety.

Historical & philosophical context [Master]

David Hilbert's Ueber die vollen Invariantensysteme (1893, Math. Ann. 42, 313–373) introduced what is now called the Nullstellensatz in the appendix, where Hilbert used elimination theory and the resultant to establish the variety-ideal correspondence. The 1893 paper is also where the Hilbert basis theorem appears, foreshadowing the noetherian foundations. The proof technique was elementary but elimination-theoretic; the modern proof via the Zariski lemma is due to Oscar Zariski's 1947 Bull. AMS note A new proof of Hilbert's Nullstellensatz (53, 362–368).

Wolfgang Krull's Primidealketten in allgemeinen Ringbereichen (1928, Sitzungsberichte der Heidelberger Akademie der Wissenschaften) introduced the Hauptidealsatz — the height theorem — establishing that a principal ideal cuts dimension by at most one in any noetherian ring. Krull's contribution shifted dimension theory from a counting principle on varieties to a structural theorem on chains of primes in arbitrary noetherian rings, and the Hauptidealsatz is the algebraic engine behind Bezout-style intersection-counting in modern algebraic geometry. The auxiliary-variable proof of strong-from-weak is due to George Yuri Rainich, writing under the pen name "Rabinowitsch" (1929, Math. Ann. 102, 520) — a one-page note that is one of the more memorable two-line proofs in mathematics.

Emmy Noether's Idealtheorie in Ringbereichen (1921, Math. Ann. 83, 24–66) established the noetherian foundations on which the modern statements rest. Noether's ascending chain condition and her abstract treatment of ideals replaced the case-by-case constructive arguments of Hilbert and Lasker with a uniform structural theory. Emanuel Lasker's Zur Theorie der Moduln und Ideale (1905, Math. Ann. 60, 20–116), written during Lasker's career as world chess champion, gave primary decomposition for ideals in polynomial rings, generalised by Noether to all noetherian rings.

The model-theoretic perspective came later. Tarski, Seidenberg, Robinson, and Ax established that the theory of algebraically closed fields admits quantifier elimination and is model-complete, recasting the Nullstellensatz as the algebraic content of a logical reduction. Effective bounds in the Nullstellensatz were established by W. D. Brownawell (1987, Annals of Math. 126, 577–591) and sharpened by János Kollár (1988, Journal AMS 1, 963–975), giving polynomial-time-relevant degree bounds for the Bezout-style identity. The real-algebraic counterpart — the Positivstellensatz of Gilbert Stengle (1974, Math. Ann. 207, 87–97) and Jean-Louis Krivine — completes the picture for ordered fields and grounds the field of polynomial optimisation. Most recently, Tigran Ananyan and Melvin Hochster proved Stillman's conjecture (2020, Journal AMS 33, 291–309), confirming that projective dimension of a homogeneous ideal depends only on generator data, not on the ambient polynomial ring.

Bibliography [Master]

Originators.

Hilbert, Ueber die vollen Invariantensysteme, Math. Ann. 42 (1893), 313–373. The Nullstellensatz in the appendix, alongside the Hilbert basis theorem.
Lasker, Zur Theorie der Moduln und Ideale, Math. Ann. 60 (1905), 20–116. Primary decomposition.
Noether, Idealtheorie in Ringbereichen, Math. Ann. 83 (1921), 24–66. Noetherian foundations.
Krull, Primidealketten in allgemeinen Ringbereichen, S.-B. Heidelberg Akad. Wiss. (1928). Hauptidealsatz.
Rainich (Rabinowitsch), Zum Hilbertschen Nullstellensatz, Math. Ann. 102 (1929), 520. Auxiliary-variable trick.
Zariski, A new proof of Hilbert's Nullstellensatz, Bull. AMS 53 (1947), 362–368.

Structural follow-ups.

Cohen-Seidenberg, Prime ideals and integral dependence, Bull. AMS 52 (1946), 252–261. Going-up and going-down.
Nagata, On the chain problem of prime ideals, Nagoya Math. J. 10 (1956), 51–64. Non-catenary noetherian local domain.
Stengle, A Nullstellensatz and a Positivstellensatz in semialgebraic geometry, Math. Ann. 207 (1974), 87–97.
Brownawell, Bounds for the degrees in the Nullstellensatz, Annals of Math. 126 (1987), 577–591.
Kollár, Sharp effective Nullstellensatz, Journal AMS 1 (1988), 963–975. Optimal bound $d^{n}$ .
Ananyan-Hochster, Small subalgebras of polynomial rings and Stillman's conjecture, Journal AMS 33 (2020), 291–309.

Textbook references.

Atiyah-Macdonald, Introduction to Commutative Algebra, Addison-Wesley, 1969. Ch. 5, 7, 11.
Hartshorne, Algebraic Geometry, Springer GTM 52, 1977. §I.1.
Eisenbud, Commutative Algebra with a View Toward Algebraic Geometry, Springer GTM 150, 1995. Ch. 1, 8–10, 13.
Matsumura, Commutative Ring Theory, Cambridge Studies in Advanced Mathematics 8, 1986. Ch. 5, 14, 17.
Reid, Undergraduate Commutative Algebra, LMS Student Texts 29, 1995. §3, §6, §8.
Stacks Project, https://stacks.math.columbia.edu. Tags 00FV, 00FW, 00KW, 00KZ.

Prerequisites

04.02.02
04.02.04

Tier anchors

beginner: Strogatz analogous level for the dictionary between polynomial equations and their common zero sets
intermediate: Atiyah-Macdonald *Introduction to Commutative Algebra* Ch. 5, 7, 11; Hartshorne §I.1; Eisenbud *Commutative Algebra* Ch. 1, 8–10; Reid *Undergraduate Commutative Algebra* §3, §6, §8
master: Hilbert 1893 (originator); Krull 1928 (Hauptidealsatz); Zariski 1947 (modern Zariski-lemma proof); Eisenbud *Commutative Algebra with a View Toward Algebraic Geometry* Ch. 8–13; Matsumura *Commutative Ring Theory* Ch. 5, 14, 17; Stacks Project Tags 00FV (Nullstellensatz), 00KW (Krull dimension), 00KZ (Hauptidealsatz)

References

TODO_REF
Hilbert 1893 — Ueber die vollen Invariantensysteme · Math. Ann. 42, 313–373; Nullstellensatz appears in the appendix via elimination theory
TODO_REF
Krull 1928 — Primidealketten in allgemeinen Ringbereichen · S.-B. Heidelberg Akad. Wiss. (Hauptidealsatz / height theorem)
TODO_REF
Noether 1921 — Idealtheorie in Ringbereichen · Math. Ann. 83, 24–66 (modern Noetherian foundations of ideal theory)
TODO_REF
Zariski 1947 — A new proof of Hilbert's Nullstellensatz · Bull. AMS 53, 362–368 (modern Zariski-lemma proof)
TODO_REF
Rainich (Rabinowitsch) 1929 — Zum Hilbertschen Nullstellensatz · Math. Ann. 102, 520 (the Rabinowitsch trick: weak ⇒ strong via auxiliary variable)
TODO_REF
Lasker 1905 — Zur Theorie der Moduln und Ideale · Math. Ann. 60, 20–116 (primary decomposition; companion to Nullstellensatz)
TODO_REF
Atiyah-Macdonald — Introduction to Commutative Algebra · Addison-Wesley 1969; Ch. 5 (integral dependence, Nullstellensatz), Ch. 11 (dimension theory)
TODO_REF
Hartshorne — Algebraic Geometry · §I.1 (Nullstellensatz and the variety/ideal correspondence)
TODO_REF
Eisenbud — Commutative Algebra with a View Toward Algebraic Geometry · Springer GTM 150, 1995; Ch. 1 (Nullstellensatz), Ch. 8 (integral extensions, going-up/going-down), Ch. 9–10 (dimension theory, Hilbert-Samuel polynomial), Ch. 13 (Cohen-Macaulay rings)
TODO_REF
Matsumura — Commutative Ring Theory · Cambridge Studies in Advanced Mathematics 8, 1986; Ch. 5 (dimension), Ch. 14 (Hauptidealsatz), Ch. 17 (Cohen-Macaulay rings, catenary rings)
TODO_REF
Stacks Project · Tags 00FV (Nullstellensatz), 00FW (Zariski lemma), 00KW (Krull dimension), 00KZ (Hauptidealsatz); https://stacks.math.columbia.edu
TODO_REF
Brownawell 1987 — Bounds for the degrees in the Nullstellensatz · Annals of Math. 126, 577–591 (effective Nullstellensatz, polynomial degree bounds)
TODO_REF
Kollár 1988 — Sharp effective Nullstellensatz · Journal AMS 1, 963–975 (sharpens Brownawell to optimal $d^n$)
TODO_REF
Stengle 1974 — A Nullstellensatz and a Positivstellensatz in semialgebraic geometry · Math. Ann. 207, 87–97 (real Nullstellensatz / Positivstellensatz)
TODO_REF
Ananyan-Hochster 2020 — Small subalgebras of polynomial rings and Stillman's conjecture · Journal AMS 33, 291–309 (Stillman's conjecture, formerly Ananyan-Hochster theorem)

Reviewer

TBD

Estimated time

beginner: 15m
intermediate: 45m
master: 70m