Nakayama's lemma

Intuition [Beginner]

Imagine a module $M$ as a room full of objects, and the Jacobson radical $J$ of a ring as a collection of elements that act "almost like zero" on modules. Nakayama's lemma says: if you can rebuild every object in the room using only these almost-zero elements, then the room was empty. Formally, if $M=JM$ and $M$ is finitely generated, then $M=0$.

The "finitely generated" condition is essential. Without it, an infinitely generated module can satisfy $JM=M$ without being zero. The lemma applies precisely when the module admits a finite description — when finitely many elements suffice to generate all of $M$.

Why does this concept exist? Nakayama's lemma provides the single most important cancellation tool in commutative algebra: it turns module-theoretic equalities involving the radical into definitive conclusions about the module itself.

Visual [Beginner]

A rectangle labelled $M$ representing a module, with a shaded sub-region $JM$ inside it. An arrow from $JM$ wrapping around to cover all of $M$ illustrates the hypothesis $JM=M$. A second, collapsed diagram shows the conclusion: the rectangle shrinks to a single point $0$.

The diagram captures the collapse: when the radical-scaled part fills the whole module, the module must be zero.

Worked example [Beginner]

Take $R=\mathbb{Z}/4\mathbb{Z}$ with elements $\{0,1,2,3\}$. The unique maximal ideal is $\mathfrak{m}=\{0,2\}$, and the Jacobson radical is $J=\mathfrak{m}$. The module is $M=R$ (the ring viewed as a module over itself).

Step 1. Compute $JM$: multiply each element of $J=\{0,2\}$ by each element of $M=\{0,1,2,3\}$. The products are $0\cdot 0=0$, $0\cdot 1=0$, $0\cdot 2=0$, $0\cdot 3=0$, $2\cdot 0=0$, $2\cdot 1=2$, $2\cdot 2=0$, $2\cdot 3=2$. So $JM=\{0,2\}$.

Step 2. Compare $JM$ with $M$. We have $JM=\{0,2\}$ and $M=\{0,1,2,3\}$. The equality $JM=M$ fails because the elements $1$ and $3$ are missing from $JM$.

Step 3. Nakayama's lemma guarantees this failure: since $M\ne 0$ and $M$ is finitely generated (by the single element $1$), the radical-scaled part $JM$ must be strictly smaller than $M$.

What this tells us: the elements $1$ and $3$ are not "almost zero" — they cannot be produced from elements of the radical alone.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Throughout, $R$ denotes a commutative ring with identity. A maximal ideal $\mathfrak{m}$ of $R$ is a proper ideal such that no ideal $I$ satisfies $\mathfrak{m}⊊I⊊R$. Every commutative ring with identity has at least one maximal ideal (by Zorn's lemma). An $R$-module $M$ is an abelian group equipped with a scalar multiplication $R\times M\to M$ satisfying the usual axioms. A module $M$ is finitely generated if there exist $m_1,\dots ,m_n\in M$ such that every element of $M$ is an $R$-linear combination of $m_1,\dots ,m_n$.

Definition (Jacobson radical). The Jacobson radical of $R$, written $J(R)$, is the intersection of all maximal ideals of $R$: $$ J(R) = \bigcap_{\mathfrak{m} \text{ maximal}} \mathfrak{m}. $$ Equivalently, $x\in J(R)$ if and only if $1-xy$ is a unit of $R$ for every $y\in R$ ^{[Atiyah-Macdonald 1969]}.

If $(R,\mathfrak{m})$ is a local ring — a ring with exactly one maximal ideal $\mathfrak{m}$ — then $J(R)=\mathfrak{m}$. The residue field of a local ring is $k=R/\mathfrak{m}$.

Counterexamples to common slips [Intermediate+]

• The Jacobson radical need not be zero. In $R=\mathbb{Z}/4\mathbb{Z}$, the unique maximal ideal is $(2)=\{0,2\}$, so $J(R)=(2)\ne 0$. A ring has $J(R)=0$ precisely when the intersection of all maximal ideals is zero; this holds for $\mathbb{Z}$ and for any semilocal ring that is a finite product of fields.

• The Jacobson radical is not the nilradical. The nilradical is the intersection of all prime ideals, while $J(R)$ is the intersection of all maximal ideals. Since every maximal ideal is prime, $J(R)$ contains the nilradical. The containment can be strict: in $\mathbb{Z}$, the nilradical is $0$ and $J(\mathbb{Z})=0$, but in $R=k[[x]]$ (formal power series), the nilradical is $0$ while $J(R)=(x)$.

• Nakayama fails for non-finitely-generated modules. Let $(R,\mathfrak{m})$ be a DVR (e.g., $R=k[[t]]$) with fraction field $K=\mathrm{Frac}(R)$. Then $\mathfrak{m}K=K$ (every Laurent series is $t$ times another Laurent series) and $K\ne 0$, but $K$ is not finitely generated over $R$.

Key theorem with proof [Intermediate+]

Theorem (Nakayama's lemma — determinant trick). Let $R$ be a commutative ring, $I$ an ideal, and $M$ a finitely generated $R$-module with $IM=M$. Then there exists $x\in I$ such that $(1-x)M=0$. In particular, if $I\subset J(R)$, then $1-x$ is a unit and $M=0$.

Proof. Let $m_1,\dots ,m_n$ generate $M$. Since $IM=M$, for each $i$ there exist $a_{i1},\dots ,a_{in}\in I$ with $$ m_i = \sum_{j=1}^{n} a_{ij}, m_j. $$ Rearranging gives ${\sum }_{j=1}^n({\delta }_{ij}-a_{ij})m_j=0$ for each $i$. Set $B=({\delta }_{ij}-a_{ij})$, an $n\times n$ matrix. The system above is $B\cdot \vec{m}=\vec{0}$ where $\vec{m}=(m_1,\dots ,m_n{)}^{\top }$.

Let $\mathrm{adj}(B)$ denote the classical adjugate (adjoint) matrix of $B$, satisfying $\mathrm{adj}(B)\cdot B=\det (B)\cdot I_n$. Multiplying by $\mathrm{adj}(B)$ yields $$ \det(B) \cdot m_j = 0 \qquad \text{for all } j = 1, \ldots, n. $$ Since the $m_j$ generate $M$, this gives $\det (B)\cdot M=0$.

Now examine $\det (B)=\det (I_n-A)$ where $A=(a_{ij})$ has all entries in $I$. Expanding the determinant: $$ \det(I_n - A) = \sum_{\sigma \in S_n} \operatorname{sgn}(\sigma) \prod_{i=1}^{n} (\delta_{i,\sigma(i)} - a_{i,\sigma(i)}). $$ The identity permutation $\sigma =\mathrm{id}$ contributes ${\prod }_{i=1}^n(1-a_{ii})$, which equals $1+c_0$ for some $c_0\in I$ (expand the product; every term beyond $1$ involves at least one $a_{ii}\in I$). Every other permutation $\sigma \ne \mathrm{id}$ satisfies $\sigma (i)\ne i$ for some $i$, so the factor $({\delta }_{i,\sigma (i)}-a_{i,\sigma (i)})=-a_{i,\sigma (i)}\in I$ for that index. Hence each non-identity term lies in $I$.

Therefore $\det (B)=1+c$ for some $c\in I$. Setting $x=-c\in I$, we have $(1-x)=(1+c)=\det (B)$, so $(1-x)M=0$.

If $I\subset J(R)$, then $x\in J(R)$. For any $j\in J(R)$, the element $1-j$ is a unit: if $1-j$ is not a unit, then $(1-j)\subset \mathfrak{m}$ for some maximal ideal $\mathfrak{m}$, so $1=(1-j)+j\in \mathfrak{m}$ (since $j\in J(R)\subset \mathfrak{m}$), contradicting $\mathfrak{m}\ne R$. Hence $1-x$ is a unit, and $(1-x)M=0$ forces $M=0$. $\square$

Bridge. The determinant trick builds toward 01.02.22, where Nakayama's lemma controls the minimal number of generators of an ideal and therefore bounds the height of prime ideals via the principal ideal theorem. The foundational reason the proof works is that $1-j$ is invertible for every $j$ in the Jacobson radical, and this is exactly the algebraic property distinguishing the radical from an arbitrary ideal. The central insight is that a single determinant captures the entire module-theoretic relationship $IM=M$, and the pattern recurs in 01.02.17, where Nakayama's lemma combines with the Noetherian property to establish the Krull intersection theorem and the structure theory of finitely generated modules.

Exercises [Intermediate+]

Advanced results [Master]

Theorem 1 (Krull intersection theorem — local case). Let $(R,\mathfrak{m})$ be a Noetherian local ring and $M$ a finitely generated $R$-module. Then ${\bigcap }_{n=1}^{\infty }{\mathfrak{m}}^{n}M=0$.

Set $N={\bigcap }_{n\ge 1}{\mathfrak{m}}^{n}M$. By the Artin-Rees lemma, there exists an integer $r$ such that ${\mathfrak{m}}^{n}M\cap N={\mathfrak{m}}^{n-r}({\mathfrak{m}}^{r}M\cap N)\subset {\mathfrak{m}}^{n-r}N$ for all $n>r$. Setting $n$ large gives $N=\mathfrak{m}N$. Since $M$ is finitely generated over a Noetherian ring, so is $N$ (as a submodule of $M$). Nakayama's lemma gives $N=0$.

Theorem 2 (Surjective endomorphism theorem). Let $M$ be a finitely generated module over a commutative ring $R$. Every surjective $R$-linear endomorphism $\phi :M\to M$ is an isomorphism.

This follows from the determinant trick applied to $M$ as an $R[x]$-module with $x$ acting via $\phi$. Surjectivity gives $(x)M=M$, so the determinant trick produces $p(x)\in (x)$ with $(1-p(x))M=0$, yielding a right inverse for $\phi$. Since a surjective map with a right inverse is bijective, $\phi$ is an isomorphism.

Theorem 3 (Minimum-generator dimension). Let $(R,\mathfrak{m},k)$ be a local ring and $M$ a finitely generated $R$-module. The minimum number of generators of $M$ equals ${\dim }_k(M/\mathfrak{m}M)$. Any set of elements whose images form a $k$-basis of $M/\mathfrak{m}M$ is a minimal generating set for $M$ ^{[Matsumura 1989]}.

Theorem 4 (Krull intersection theorem — general Noetherian case). Let $R$ be a Noetherian ring, $I$ an ideal of $R$, and $M$ a finitely generated $R$-module. Then ${\bigcap }_{n=1}^{\infty }{I}^{n}M=0$ if and only if no element of $1+I$ is a zero-divisor on $M$.

This generalises the local case: when $I\subset J(R)$, every element of $1+I$ is a unit, hence not a zero-divisor, and the intersection vanishes.

Theorem 5 (Graded Nakayama lemma). Let $R={⨁}_{n\ge 0}{R}_n$ be a graded ring with $R_0$ a field, and let $R_{+}={⨁}_{n>0}{R}_n$ be the irrelevant ideal. If $M={⨁}_{n\in \mathbb{Z}}{M}_n$ is a graded $R$-module that is bounded below ($M_n=0$ for $n\ll 0$) and generated in degree $\le d$, then homogeneous elements $m_1,\dots ,m_k$ generate $M$ if and only if their images generate $M/R_{+}M$ as an $R_0$-module.

Theorem 6 (Completion is faithfully flat). Let $(R,\mathfrak{m})$ be a Noetherian local ring with $\mathfrak{m}$-adic completion $\hat{R}$. The natural map $R\to \hat{R}$ is faithfully flat: a sequence of $R$-modules is exact if and only if its tensor product with $\hat{R}$ is exact.

Faithful flatness of $R\to \hat{R}$ rests on Nakayama's lemma through the Krull intersection theorem: the $\mathfrak{m}$-adic topology on any finitely generated $R$-module is separated ($\bigcap {\mathfrak{m}}^{n}M=0$), which guarantees that $M\to M{\otimes }_R\hat{R}$ is injective.

Synthesis. Nakayama's lemma is the foundational reason that the structure theory of finitely generated modules over local rings reduces to linear algebra over the residue field. The central insight is the radical's inability to contribute new generators: if $M=N+JM$, the part $JM$ is redundant. Putting these together with the Krull intersection theorem, which iterates Nakayama's cancellation to show that the $\mathfrak{m}$-adic topology separates points, the bridge is between the algebraic statement $M=JM\Rightarrow M=0$ and the topological statement $\bigcap {\mathfrak{m}}^{n}M=0$. This is exactly the mechanism that identifies completions of local rings with inverse limits of quotient systems $R/{\mathfrak{m}}^n$, and the pattern generalises to graded rings via the graded Nakayama lemma and to filtered modules via the Artin-Rees lemma.

Full proof set [Master]

Proposition 1 (Krull intersection theorem — local case). Let $(R,\mathfrak{m})$ be a Noetherian local ring and $M$ a finitely generated $R$-module. Then ${\bigcap }_{n=1}^{\infty }{\mathfrak{m}}^{n}M=0$.

Proof. Set $N={\bigcap }_{n\ge 1}{\mathfrak{m}}^{n}M$. Since $R$ is Noetherian and $M$ is finitely generated, $N$ is a finitely generated $R$-module (being a submodule of $M$). By the Artin-Rees lemma ^{[Atiyah-Macdonald 1969]}, there exists an integer $c$ such that for all $n\ge c$: $$ \mathfrak{m}^n M \cap N = \mathfrak{m}^{n-c} (\mathfrak{m}^c M \cap N). $$ Since $N\subset {\mathfrak{m}}^{n}M$ for every $n$, the left side equals $N$. The right side is ${\mathfrak{m}}^{n-c}\cdot N\subset \mathfrak{m}N$ for $n>c+1$. Taking $n$ sufficiently large gives $N={\mathfrak{m}}^{n-c}N\subset \mathfrak{m}N$. Therefore $N=\mathfrak{m}N$.

Since $R$ is local, $\mathfrak{m}=J(R)$. Nakayama's lemma applies: $N$ is finitely generated, $\mathfrak{m}N=N$, and $\mathfrak{m}=J(R)$, so $N=0$. $\square$

Proposition 2 (Surjective endomorphisms are isomorphisms). Let $M$ be a finitely generated module over a commutative ring $R$ and let $\phi :M\to M$ be a surjective $R$-linear map. Then $\phi$ is injective, hence an isomorphism.

Proof. View $M$ as a module over $S=R[x]$, where $x$ acts on $M$ via $\phi$. Concretely, for $p(x)=a_0+a_{1}x+\cdots +a_d{x}^d\in R[x]$ and $m\in M$, define $p(x)\cdot m=a_{0}m+a_1\phi (m)+\cdots +a_d{\phi }^d(m)$.

Since $M$ is finitely generated over $R$ (by $m_1,\dots ,m_n$), it is finitely generated over $S$ (by the same elements). Surjectivity of $\phi$ gives $xM=M$, so $IM=M$ for the ideal $I=(x)\subset S$.

Apply the determinant trick: there exists $f(x)\in I=(x)$ with $(1-f(x))M=0$. Write $f(x)=xg(x)$ for some $g(x)\in R[x]$. The relation $(1-xg(x))\cdot m=0$ for all $m\in M$ means: $$ m = \varphi(g(\varphi)(m)) \quad \text{for all } m \in M. $$ Define $\psi =g(\phi ):M\to M$. Then $\phi \circ \psi ={\mathrm{id}}_M$. Since $\phi$ has a right inverse and is surjective, $\phi$ is bijective: if $\phi (m)=0$, then $m=\psi (\phi (m))=\psi (0)=0$, so $\ker \phi =0$. $\square$

Connections [Master]

• Hilbert basis theorem and Noetherian rings 01.02.17. Nakayama's lemma is most powerful in the Noetherian setting, where every submodule of a finitely generated module is again finitely generated. The Krull intersection theorem and the faithful flatness of completions both require the Noetherian hypothesis supplied by 01.02.17. Conversely, Nakayama's lemma provides the cancellation step in many finiteness arguments that build on the Hilbert basis theorem.

• Localisation of commutative rings 01.02.08. The local rings introduced via localisation are the natural home for Nakayama's lemma: in a local ring $(R,\mathfrak{m})$, the Jacobson radical equals $\mathfrak{m}$, and the minimum-generator theorem reduces module structure to linear algebra over the residue field $R/\mathfrak{m}$. Localisation at a prime ideal produces a local ring, so Nakayama applies after localising.

• Krull dimension and the principal ideal theorem 01.02.22. Nakayama's lemma controls the number of generators of an ideal in a local ring, which directly bounds the height of a prime ideal via the principal ideal theorem. The dimension equality $\dim R={\dim }_{R/\mathfrak{m}}(\mathfrak{m}/{\mathfrak{m}}^2)$ for regular local rings — a cornerstone of dimension theory — is proved using Nakayama's lemma to compare generators of $\mathfrak{m}$ with the $R/\mathfrak{m}$-vector space $\mathfrak{m}/{\mathfrak{m}}^2$.

Historical & philosophical context [Master]

Krull 1938 proved the intersection theorem that now bears his name, showing that ${\bigcap }_{n\ge 1}{\mathfrak{m}}^n=0$ in a Noetherian local ring ^{[Krull 1938]}. The general module-theoretic statement — if $M$ is finitely generated and $IM=M$ for an ideal $I$ contained in the Jacobson radical, then $M=0$ — appeared in Nakayama 1951 ^{[Nakayama 1951]}, who used the determinant trick to give the proof in its modern form. Azumaya 1950 independently developed equivalent machinery in the context of maximally central algebras ^{[Azumaya 1950]}.

The determinant trick itself has roots in the Cayley-Hamilton theorem: both produce an annihilating polynomial from the finite-generation hypothesis. The modern standard reference is Atiyah-Macdonald 1969, Chapter 2, Propositions 2.4--2.8 ^{[Atiyah-Macdonald 1969]}, which streamlines the proof and emphasises the role of the Jacobson radical. Matsumura 1989, §2, gives the definitive treatment for local rings ^{[Matsumura 1989]}.

Bibliography [Master]

@article{Nakayama1951,
  author = {Nakayama, Tadasi},
  title = {A remark on finitely generated modules},
  journal = {Nagoya Mathematical Journal},
  volume = {3},
  year = {1951},
  pages = {139--140},
}

@article{Krull1938,
  author = {Krull, Wolfgang},
  title = {Dimensionstheorie in Stellenringen},
  journal = {Journal f\"ur die reine und angewandte Mathematik},
  volume = {179},
  year = {1938},
  pages = {204--226},
}

@article{Azumaya1950,
  author = {Azumaya, Goro},
  title = {On maximally central algebras},
  journal = {Nagoya Mathematical Journal},
  volume = {2},
  year = {1950},
  pages = {119--150},
}

@book{AtiyahMacdonald1969,
  author = {Atiyah, Michael F. and Macdonald, Ian G.},
  title = {Introduction to Commutative Algebra},
  publisher = {Addison-Wesley},
  year = {1969},
}

@book{Matsumura1989,
  author = {Matsumura, Hideyuki},
  title = {Commutative Ring Theory},
  publisher = {Cambridge University Press},
  year = {1989},
  series = {Cambridge Studies in Advanced Mathematics 8},
}