The Stacks project

10.109 Rings of finite global dimension

The following lemma is often used to compare different projective resolutions of a given module.

Lemma 10.109.1 (Schanuel's lemma). Let $R$ be a ring. Let $M$ be an $R$-module. Suppose that

\[ 0 \to K \xrightarrow {c_1} P_1 \xrightarrow {p_1} M \to 0 \quad \text{and}\quad 0 \to L \xrightarrow {c_2} P_2 \xrightarrow {p_2} M \to 0 \]

are two short exact sequences, with $P_ i$ projective. Then $K \oplus P_2 \cong L \oplus P_1$. More precisely, there exist a commutative diagram

\[ \xymatrix{ 0 \ar[r] & K \oplus P_2 \ar[r]_{(c_1, \text{id})} \ar[d] & P_1 \oplus P_2 \ar[r]_{(p_1, 0)} \ar[d] & M \ar[r] \ar@{=}[d] & 0 \\ 0 \ar[r] & P_1 \oplus L \ar[r]^{(\text{id}, c_2)} & P_1 \oplus P_2 \ar[r]^{(0, p_2)} & M \ar[r] & 0 } \]

whose vertical arrows are isomorphisms.

Proof. Consider the module $N$ defined by the short exact sequence $0 \to N \to P_1 \oplus P_2 \to M \to 0$, where the last map is the sum of the two maps $P_ i \to M$. It is easy to see that the projection $N \to P_1$ is surjective with kernel $L$, and that $N \to P_2$ is surjective with kernel $K$. Since $P_ i$ are projective we have $N \cong K \oplus P_2 \cong L \oplus P_1$. This proves the first statement.

To prove the second statement (and to reprove the first), choose $a : P_1 \to P_2$ and $b : P_2 \to P_1$ such that $p_1 = p_2 \circ a$ and $p_2 = p_1 \circ b$. This is possible because $P_1$ and $P_2$ are projective. Then we get a commutative diagram

\[ \xymatrix{ 0 \ar[r] & K \oplus P_2 \ar[r]_{(c_1, \text{id})} & P_1 \oplus P_2 \ar[r]_{(p_1, 0)} & M \ar[r] & 0 \\ 0 \ar[r] & N \ar[r] \ar[d] \ar[u] & P_1 \oplus P_2 \ar[r]_{(p_1, p_2)} \ar[d]_ S \ar[u]^ T & M \ar[r] \ar@{=}[d] \ar@{=}[u] & 0 \\ 0 \ar[r] & P_1 \oplus L \ar[r]^{(\text{id}, c_2)} & P_1 \oplus P_2 \ar[r]^{(0, p_2)} & M \ar[r] & 0 } \]

with $T$ and $S$ given by the matrices

\[ S = \left( \begin{matrix} \text{id} & 0 \\ a & \text{id} \end{matrix} \right) \quad \text{and}\quad T = \left( \begin{matrix} \text{id} & b \\ 0 & \text{id} \end{matrix} \right) \]

Then $S$, $T$ and the maps $N \to P_1 \oplus L$ and $N \to K \oplus P_2$ are isomorphisms as desired. $\square$

Definition 10.109.2. Let $R$ be a ring. Let $M$ be an $R$-module. We say $M$ has finite projective dimension if it has a finite length resolution by projective $R$-modules. The minimal length of such a resolution is called the projective dimension of $M$.

It is clear that the projective dimension of $M$ is $0$ if and only if $M$ is a projective module. The following lemma explains to what extent the projective dimension is independent of the choice of a projective resolution.

Lemma 10.109.3. Let $R$ be a ring. Suppose that $M$ is an $R$-module of projective dimension $d$. Suppose that $F_ e \to F_{e-1} \to \ldots \to F_0 \to M \to 0$ is exact with $F_ i$ projective and $e \geq d - 1$. Then the kernel of $F_ e \to F_{e-1}$ is projective (or the kernel of $F_0 \to M$ is projective in case $e = 0$).

Proof. We prove this by induction on $d$. If $d = 0$, then $M$ is projective. In this case there is a splitting $F_0 = \mathop{\mathrm{Ker}}(F_0 \to M) \oplus M$, and hence $\mathop{\mathrm{Ker}}(F_0 \to M)$ is projective. This finishes the proof if $e = 0$, and if $e > 0$, then replacing $M$ by $\mathop{\mathrm{Ker}}(F_0 \to M)$ we decrease $e$.

Next assume $d > 0$. Let $0 \to P_ d \to P_{d-1} \to \ldots \to P_0 \to M \to 0$ be a minimal length finite resolution with $P_ i$ projective. According to Schanuel's Lemma 10.109.1 we have $P_0 \oplus \mathop{\mathrm{Ker}}(F_0 \to M) \cong F_0 \oplus \mathop{\mathrm{Ker}}(P_0 \to M)$. This proves the case $d = 1$, $e = 0$, because then the right hand side is $F_0 \oplus P_1$ which is projective. Hence now we may assume $e > 0$. The module $F_0 \oplus \mathop{\mathrm{Ker}}(P_0 \to M)$ has the finite projective resolution

\[ 0 \to P_ d \to P_{d-1} \to \ldots \to P_2 \to P_1 \oplus F_0 \to \mathop{\mathrm{Ker}}(P_0 \to M) \oplus F_0 \to 0 \]

of length $d - 1$. By induction applied to the exact sequence

\[ F_ e \to F_{e-1} \to \ldots \to F_2 \to P_0 \oplus F_1 \to P_0 \oplus \mathop{\mathrm{Ker}}(F_0 \to M) \to 0 \]

of length $e - 1$ we conclude $\mathop{\mathrm{Ker}}(F_ e \to F_{e - 1})$ is projective (if $e \geq 2$) or that $\mathop{\mathrm{Ker}}(F_1 \oplus P_0 \to F_0 \oplus P_0)$ is projective. This implies the lemma. $\square$

Lemma 10.109.4. Let $R$ be a ring. Let $M$ be an $R$-module. Let $d \geq 0$. The following are equivalent

  1. $M$ has projective dimension $\leq d$,

  2. there exists a resolution $0 \to P_ d \to P_{d - 1} \to \ldots \to P_0 \to M \to 0$ with $P_ i$ projective,

  3. for some resolution $\ldots \to P_2 \to P_1 \to P_0 \to M \to 0$ with $P_ i$ projective we have $\mathop{\mathrm{Ker}}(P_{d - 1} \to P_{d - 2})$ is projective if $d \geq 2$, or $\mathop{\mathrm{Ker}}(P_0 \to M)$ is projective if $d = 1$, or $M$ is projective if $d = 0$,

  4. for any resolution $\ldots \to P_2 \to P_1 \to P_0 \to M \to 0$ with $P_ i$ projective we have $\mathop{\mathrm{Ker}}(P_{d - 1} \to P_{d - 2})$ is projective if $d \geq 2$, or $\mathop{\mathrm{Ker}}(P_0 \to M)$ is projective if $d = 1$, or $M$ is projective if $d = 0$.

Proof. The equivalence of (1) and (2) is the definition of projective dimension, see Definition 10.109.2. We have (2) $\Rightarrow $ (4) by Lemma 10.109.3. The implications (4) $\Rightarrow $ (3) and (3) $\Rightarrow $ (2) are immediate. $\square$

Lemma 10.109.5. Let $R$ be a local ring. Let $M$ be an $R$-module. Let $d \geq 0$. The equivalent conditions (1) – (4) of Lemma 10.109.4 are also equivalent to

  1. there exists a resolution $0 \to P_ d \to P_{d - 1} \to \ldots \to P_0 \to M \to 0$ with $P_ i$ free.

Lemma 10.109.6. Let $R$ be a Noetherian ring. Let $M$ be a finite $R$-module. Let $d \geq 0$. The equivalent conditions (1) – (4) of Lemma 10.109.4 are also equivalent to

  1. there exists a resolution $0 \to P_ d \to P_{d - 1} \to \ldots \to P_0 \to M \to 0$ with $P_ i$ finite projective.

Proof. Choose a resolution $\ldots \to F_2 \to F_1 \to F_0 \to M \to 0$ with $F_ i$ finite free (Lemma 10.71.1). By Lemma 10.109.4 we see that $P_ d = \mathop{\mathrm{Ker}}(F_{d - 1} \to F_{d - 2})$ is projective at least if $d \geq 2$. Then $P_ d$ is a finite $R$-module as $R$ is Noetherian and $P_ d \subset F_{d - 1}$ which is finite free. Whence $0 \to P_ d \to F_{d - 1} \to \ldots \to F_1 \to F_0 \to M \to 0$ is the desired resolution. $\square$

Lemma 10.109.7. Let $R$ be a local Noetherian ring. Let $M$ be a finite $R$-module. Let $d \geq 0$. The equivalent conditions (1) – (4) of Lemma 10.109.4, condition (5) of Lemma 10.109.5, and condition (6) of Lemma 10.109.6 are also equivalent to

  1. there exists a resolution $0 \to F_ d \to F_{d - 1} \to \ldots \to F_0 \to M \to 0$ with $F_ i$ finite free.

Proof. This follows from Lemmas 10.109.4, 10.109.5, and 10.109.6 and because a finite projective module over a local ring is finite free, see Lemma 10.78.2. $\square$

Lemma 10.109.8. Let $R$ be a ring. Let $M$ be an $R$-module. Let $n \geq 0$. The following are equivalent

  1. $M$ has projective dimension $\leq n$,

  2. $\mathop{\mathrm{Ext}}\nolimits ^ i_ R(M, N) = 0$ for all $R$-modules $N$ and all $i \geq n + 1$, and

  3. $\mathop{\mathrm{Ext}}\nolimits ^{n + 1}_ R(M, N) = 0$ for all $R$-modules $N$.

Proof. Assume (1). Choose a free resolution $F_\bullet \to M$ of $M$. Denote $d_ e : F_ e \to F_{e - 1}$. By Lemma 10.109.3 we see that $P_ e = \mathop{\mathrm{Ker}}(d_ e)$ is projective for $e \geq n - 1$. This implies that $F_ e \cong P_ e \oplus P_{e - 1}$ for $e \geq n$ where $d_ e$ maps the summand $P_{e - 1}$ isomorphically to $P_{e - 1}$ in $F_{e - 1}$. Hence, for any $R$-module $N$ the complex $\mathop{\mathrm{Hom}}\nolimits _ R(F_\bullet , N)$ is split exact in degrees $\geq n + 1$. Whence (2) holds. The implication (2) $\Rightarrow $ (3) is trivial.

Assume (3) holds. If $n = 0$ then $M$ is projective by Lemma 10.77.2 and we see that (1) holds. If $n > 0$ choose a free $R$-module $F$ and a surjection $F \to M$ with kernel $K$. By Lemma 10.71.7 and the vanishing of $\mathop{\mathrm{Ext}}\nolimits _ R^ i(F, N)$ for all $i > 0$ by part (1) we see that $\mathop{\mathrm{Ext}}\nolimits _ R^ n(K, N) = 0$ for all $R$-modules $N$. Hence by induction we see that $K$ has projective dimension $\leq n - 1$. Then $M$ has projective dimension $\leq n$ as any finite projective resolution of $K$ gives a projective resolution of length one more for $M$ by adding $F$ to the front. $\square$

Lemma 10.109.9. Let $R$ be a ring. Let $0 \to M' \to M \to M'' \to 0$ be a short exact sequence of $R$-modules.

  1. If $M$ has projective dimension $\leq n$ and $M''$ has projective dimension $\leq n + 1$, then $M'$ has projective dimension $\leq n$.

  2. If $M'$ and $M''$ have projective dimension $\leq n$ then $M$ has projective dimension $\leq n$.

  3. If $M'$ has projective dimension $\leq n$ and $M$ has projective dimension $\leq n + 1$ then $M''$ has projective dimension $\leq n + 1$.

Proof. Combine the characterization of projective dimension in Lemma 10.109.8 with the long exact sequence of ext groups in Lemma 10.71.7. $\square$

Definition 10.109.10. Let $R$ be a ring. The ring $R$ is said to have finite global dimension if there exists an integer $n$ such that every $R$-module has a resolution by projective $R$-modules of length at most $n$. The minimal such $n$ is then called the global dimension of $R$.

The argument in the proof of the following lemma can be found in the paper [Auslander] by Auslander.

Lemma 10.109.11. Let $R$ be a ring. Suppose we have a module $M = \bigcup _{e \in E} M_ e$ where the $M_ e$ are submodules well-ordered by inclusion. Assume the quotients $M_ e/\bigcup \nolimits _{e' < e} M_{e'}$ have projective dimension $\leq n$. Then $M$ has projective dimension $\leq n$.

Proof. We will prove this by induction on $n$.

Base case: $n = 0$. Then $P_ e = M_ e/\bigcup _{e' < e} M_{e'}$ is projective. Thus we may choose a section $P_ e \to M_ e$ of the projection $M_ e \to P_ e$. We claim that the induced map $\psi : \bigoplus _{e \in E} P_ e \to M$ is an isomorphism. Namely, if $x = \sum x_ e \in \bigoplus P_ e$ is nonzero, then we let $e_{max}$ be maximal such that $x_{e_{max}}$ is nonzero and we conclude that $y = \psi (x) = \psi (\sum x_ e)$ is nonzero because $y \in M_{e_{max}}$ has nonzero image $x_{e_{max}}$ in $P_{e_{max}}$. On the other hand, let $y \in M$. Then $y \in M_ e$ for some $e$. We show that $y \in \mathop{\mathrm{Im}}(\psi )$ by transfinite induction on $e$. Let $x_ e \in P_ e$ be the image of $y$. Then $y - \psi (x_ e) \in \bigcup _{e' < e} M_{e'}$. By induction hypothesis we conclude that $y - \psi (x_ e) \in \mathop{\mathrm{Im}}(\psi )$ hence $y \in \mathop{\mathrm{Im}}(\psi )$. Thus the claim is true and $\psi $ is an isomorphism. We conclude that $M$ is projective as a direct sum of projectives, see Lemma 10.77.4.

If $n > 0$, then for $e \in E$ we denote $F_ e$ the free $R$-module on the set of elements of $M_ e$. Then we have a system of short exact sequences

\[ 0 \to K_ e \to F_ e \to M_ e \to 0 \]

over the well-ordered set $E$. Note that the transition maps $F_{e'} \to F_ e$ and $K_{e'} \to K_ e$ are injective too. Set $F = \bigcup F_ e$ and $K = \bigcup K_ e$. Then

\[ 0 \to K_ e/\bigcup \nolimits _{e' < e} K_{e'} \to F_ e/\bigcup \nolimits _{e' < e} F_{e'} \to M_ e/\bigcup \nolimits _{e' < e} M_{e'} \to 0 \]

is a short exact sequence of $R$-modules too and $F_ e/\bigcup _{e' < e} F_{e'}$ is the free $R$-module on the set of elements in $M_ e$ which are not contained in $\bigcup _{e' < e} M_{e'}$. Hence by Lemma 10.109.9 we see that the projective dimension of $K_ e/\bigcup _{e' < e} K_{e'}$ is at most $n - 1$. By induction we conclude that $K$ has projective dimension at most $n - 1$. Whence $M$ has projective dimension at most $n$ and we win. $\square$

Lemma 10.109.12. Let $R$ be a ring. The following are equivalent

  1. $R$ has finite global dimension $\leq n$,

  2. every finite $R$-module has projective dimension $\leq n$, and

  3. every cyclic $R$-module $R/I$ has projective dimension $\leq n$.

Proof. It is clear that (1) $\Rightarrow $ (2) and (2) $\Rightarrow $ (3). Assume (3). Choose a set $E \subset M$ of generators of $M$. Choose a well ordering on $E$. For $e \in E$ denote $M_ e$ the submodule of $M$ generated by the elements $e' \in E$ with $e' \leq e$. Then $M = \bigcup _{e \in E} M_ e$. Note that for each $e \in E$ the quotient

\[ M_ e/\bigcup \nolimits _{e' < e} M_{e'} \]

is either zero or generated by one element, hence has projective dimension $\leq n$ by (3). By Lemma 10.109.11 this means that $M$ has projective dimension $\leq n$. $\square$

Lemma 10.109.13. Let $R$ be a ring. Let $M$ be an $R$-module. Let $S \subset R$ be a multiplicative subset.

  1. If $M$ has projective dimension $\leq n$, then $S^{-1}M$ has projective dimension $\leq n$ over $S^{-1}R$.

  2. If $R$ has finite global dimension $\leq n$, then $S^{-1}R$ has finite global dimension $\leq n$.

Proof. Let $0 \to P_ n \to P_{n - 1} \to \ldots \to P_0 \to M \to 0$ be a projective resolution. As localization is exact, see Proposition 10.9.12, and as each $S^{-1}P_ i$ is a projective $S^{-1}R$-module, see Lemma 10.94.1, we see that $0 \to S^{-1}P_ n \to \ldots \to S^{-1}P_0 \to S^{-1}M \to 0$ is a projective resolution of $S^{-1}M$. This proves (1). Let $M'$ be an $S^{-1}R$-module. Note that $M' = S^{-1}M'$. Hence we see that (2) follows from (1). $\square$

Comments (1)

Comment #689 by Keenan Kidwell on

In the paragraph following 00O4, "extend" should be "extent."

Post a comment

Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.




In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 00O2. Beware of the difference between the letter 'O' and the digit '0'.



View Section 10.109 as pdf