Processing math: 100%

The Stacks project

10.69 Quasi-regular sequences

We introduce the notion of quasi-regular sequence which is slightly weaker than that of a regular sequence and easier to use. Let R be a ring and let f_1, \ldots , f_ c \in R. Set J = (f_1, \ldots , f_ c). Let M be an R-module. Then there is a canonical map

10.69.0.1
\begin{equation} \label{algebra-equation-quasi-regular} M/JM \otimes _{R/J} R/J[X_1, \ldots , X_ c] \longrightarrow \bigoplus \nolimits _{n \geq 0} J^ nM/J^{n + 1}M \end{equation}

of graded R/J[X_1, \ldots , X_ c]-modules defined by the rule

\overline{m} \otimes X_1^{e_1} \ldots X_ c^{e_ c} \longmapsto f_1^{e_1} \ldots f_ c^{e_ c} m \bmod J^{e_1 + \ldots + e_ c + 1}M.

Note that (10.69.0.1) is always surjective.

Definition 10.69.1. Let R be a ring. Let M be an R-module. A sequence of elements f_1, \ldots , f_ c of R is called M-quasi-regular if (10.69.0.1) is an isomorphism. If M = R, we call f_1, \ldots , f_ c simply a quasi-regular sequence.

So if f_1, \ldots , f_ c is a quasi-regular sequence, then

R/J[X_1, \ldots , X_ c] = \bigoplus \nolimits _{n \geq 0} J^ n/J^{n + 1}

where J = (f_1, \ldots , f_ c). It is clear that being a quasi-regular sequence is independent of the order of f_1, \ldots , f_ c.

Lemma 10.69.2. Let R be a ring.

  1. A regular sequence f_1, \ldots , f_ c of R is a quasi-regular sequence.

  2. Suppose that M is an R-module and that f_1, \ldots , f_ c is an M-regular sequence. Then f_1, \ldots , f_ c is an M-quasi-regular sequence.

Proof. Set J = (f_1, \ldots , f_ c). We prove the first assertion by induction on c. We have to show that given any relation \sum _{|I| = n} a_ I f^ I \in J^{n + 1} with a_ I \in R we actually have a_ I \in J for all multi-indices I. Since any element of J^{n + 1} is of the form \sum _{|I| = n} b_ I f^ I with b_ I \in J we may assume, after replacing a_ I by a_ I - b_ I, the relation reads \sum _{|I| = n} a_ I f^ I = 0. We can rewrite this as

\sum \nolimits _{e = 0}^ n \left( \sum \nolimits _{|I'| = n - e} a_{I', e} f^{I'} \right) f_ c^ e = 0

Here and below the “primed” multi-indices I' are required to be of the form I' = (i_1, \ldots , i_{c - 1}, 0). We will show by descending induction on l \in \{ 0, \ldots , n\} that if we have a relation

\sum \nolimits _{e = 0}^ l \left( \sum \nolimits _{|I'| = n - e} a_{I', e} f^{I'} \right) f_ c^ e = 0

then a_{I', e} \in J for all I', e. Namely, set J' = (f_1, \ldots , f_{c-1}). Observe that \sum \nolimits _{|I'| = n - l} a_{I', l} f^{I'} is mapped into (J')^{n - l + 1} by f_ c^{l}. By induction hypothesis (for the induction on c) we see that f_ c^ l a_{I', l} \in J'. Because f_ c is not a zerodivisor on R/J' (as f_1, \ldots , f_ c is a regular sequence) we conclude that a_{I', l} \in J'. This allows us to rewrite the term (\sum \nolimits _{|I'| = n - l} a_{I', l} f^{I'})f_ c^ l in the form (\sum \nolimits _{|I'| = n - l + 1} f_ c b_{I', l - 1} f^{I'})f_ c^{l-1}. This gives a new relation of the form

\left(\sum \nolimits _{|I'| = n - l + 1} (a_{I', l-1} + f_ c b_{I', l - 1}) f^{I'}\right)f_ c^{l-1} + \sum \nolimits _{e = 0}^{l - 2} \left( \sum \nolimits _{|I'| = n - e} a_{I', e} f^{I'} \right) f_ c^ e = 0

Now by the induction hypothesis (on l this time) we see that all a_{I', l-1} + f_ c b_{I', l - 1} \in J and all a_{I', e} \in J for e \leq l - 2. This, combined with a_{I', l} \in J' \subset J seen above, finishes the proof of the induction step.

The second assertion means that given any formal expression F = \sum _{|I| = n} m_ I X^ I, m_ I \in M with \sum m_ I f^ I \in J^{n + 1}M, then all the coefficients m_ I are in J. This is proved in exactly the same way as we prove the corresponding result for the first assertion above. \square

Lemma 10.69.3. Let R \to R' be a flat ring map. Let M be an R-module. Suppose that f_1, \ldots , f_ r \in R form an M-quasi-regular sequence. Then the images of f_1, \ldots , f_ r in R' form a M \otimes _ R R'-quasi-regular sequence.

Proof. Set J = (f_1, \ldots , f_ r), J' = JR' and M' = M \otimes _ R R'. We have to show the canonical map \mu : R'/J'[X_1, \ldots X_ r] \otimes _{R'/J'} M'/J'M' \to \bigoplus (J')^ nM'/(J')^{n + 1}M' is an isomorphism. Because R \to R' is flat the sequences 0 \to J^ nM \to M and 0 \to J^{n + 1}M \to J^ nM \to J^ nM/J^{n + 1}M \to 0 remain exact on tensoring with R'. This first implies that J^ nM \otimes _ R R' = (J')^ nM' and then that (J')^ nM'/(J')^{n + 1}M' = J^ nM/J^{n + 1}M \otimes _ R R'. Thus \mu is the tensor product of (10.69.0.1), which is an isomorphism by assumption, with \text{id}_{R'} and we conclude. \square

Lemma 10.69.4. Let R be a Noetherian ring. Let M be a finite R-module. Let \mathfrak p be a prime. Let x_1, \ldots , x_ c be a sequence in R whose image in R_{\mathfrak p} forms an M_{\mathfrak p}-quasi-regular sequence. Then there exists a g \in R, g \not\in \mathfrak p such that the image of x_1, \ldots , x_ c in R_ g forms an M_ g-quasi-regular sequence.

Proof. Consider the kernel K of the map (10.69.0.1). As M/JM \otimes _{R/J} R/J[X_1, \ldots , X_ c] is a finite R/J[X_1, \ldots , X_ c]-module and as R/J[X_1, \ldots , X_ c] is Noetherian, we see that K is also a finite R/J[X_1, \ldots , X_ c]-module. Pick homogeneous generators k_1, \ldots , k_ t \in K. By assumption for each i = 1, \ldots , t there exists a g_ i \in R, g_ i \not\in \mathfrak p such that g_ i k_ i = 0. Hence g = g_1 \ldots g_ t works. \square

Lemma 10.69.5. Let R be a ring. Let M be an R-module. Let f_1, \ldots , f_ c \in R be an M-quasi-regular sequence. For any i the sequence \overline{f}_{i + 1}, \ldots , \overline{f}_ c of \overline{R} = R/(f_1, \ldots , f_ i) is an \overline{M} = M/(f_1, \ldots , f_ i)M-quasi-regular sequence.

Proof. It suffices to prove this for i = 1. Set \overline{J} = (\overline{f}_2, \ldots , \overline{f}_ c) \subset \overline{R}. Then

\begin{align*} \overline{J}^ n\overline{M}/\overline{J}^{n + 1}\overline{M} & = (J^ nM + f_1M)/(J^{n + 1}M + f_1M) \\ & = J^ nM / (J^{n + 1}M + J^ nM \cap f_1M). \end{align*}

Thus, in order to prove the lemma it suffices to show that J^{n + 1}M + J^ nM \cap f_1M = J^{n + 1}M + f_1J^{n - 1}M because that will show that \bigoplus _{n \geq 0} \overline{J}^ n\overline{M}/\overline{J}^{n + 1}\overline{M} is the quotient of \bigoplus _{n \geq 0} J^ nM/J^{n + 1}M \cong M/JM[X_1, \ldots , X_ c] by X_1. Actually, we have J^ nM \cap f_1M = f_1J^{n - 1}M. Namely, if m \not\in J^{n - 1}M, then f_1m \not\in J^ nM because \bigoplus J^ nM/J^{n + 1}M is the polynomial algebra M/J[X_1, \ldots , X_ c] by assumption. \square

Lemma 10.69.6. Let (R, \mathfrak m) be a local Noetherian ring. Let M be a nonzero finite R-module. Let f_1, \ldots , f_ c \in \mathfrak m be an M-quasi-regular sequence. Then f_1, \ldots , f_ c is an M-regular sequence.

Proof. Set J = (f_1, \ldots , f_ c). Let us show that f_1 is a nonzerodivisor on M. Suppose x \in M is not zero. By Krull's intersection theorem there exists an integer r such that x \in J^ rM but x \not\in J^{r + 1}M, see Lemma 10.51.4. Then f_1 x \in J^{r + 1}M is an element whose class in J^{r + 1}M/J^{r + 2}M is nonzero by the assumed structure of \bigoplus J^ nM/J^{n + 1}M. Whence f_1x \not= 0.

Now we can finish the proof by induction on c using Lemma 10.69.5. \square

Remark 10.69.7 (Other types of regular sequences). In the paper [Kabele] the author discusses two more regularity conditions for sequences x_1, \ldots , x_ r of elements of a ring R. Namely, we say the sequence is Koszul-regular if H_ i(K_{\bullet }(R, x_{\bullet })) = 0 for i \geq 1 where K_{\bullet }(R, x_{\bullet }) is the Koszul complex. The sequence is called H_1-regular if H_1(K_{\bullet }(R, x_{\bullet })) = 0. One has the implications regular \Rightarrow Koszul-regular \Rightarrow H_1-regular \Rightarrow quasi-regular. By examples the author shows that these implications cannot be reversed in general even if R is a (non-Noetherian) local ring and the sequence generates the maximal ideal of R. We introduce these notions in more detail in More on Algebra, Section 15.30.

Remark 10.69.8. Let k be a field. Consider the ring

A = k[x, y, w, z_0, z_1, z_2, \ldots ]/ (y^2z_0 - wx, z_0 - yz_1, z_1 - yz_2, \ldots )

In this ring x is a nonzerodivisor and the image of y in A/xA gives a quasi-regular sequence. But it is not true that x, y is a quasi-regular sequence in A because (x, y)/(x, y)^2 isn't free of rank two over A/(x, y) due to the fact that wx = 0 in (x, y)/(x, y)^2 but w isn't zero in A/(x, y). Hence the analogue of Lemma 10.68.7 does not hold for quasi-regular sequences.

Lemma 10.69.9. Let R be a ring. Let J = (f_1, \ldots , f_ r) be an ideal of R. Let M be an R-module. Set \overline{R} = R/\bigcap _{n \geq 0} J^ n, \overline{M} = M/\bigcap _{n \geq 0} J^ nM, and denote \overline{f}_ i the image of f_ i in \overline{R}. Then f_1, \ldots , f_ r is M-quasi-regular if and only if \overline{f}_1, \ldots , \overline{f}_ r is \overline{M}-quasi-regular.

Proof. This is true because J^ nM/J^{n + 1}M \cong \overline{J}^ n\overline{M}/\overline{J}^{n + 1}\overline{M}. \square


Comments (5)

Comment #7891 by Laurent Berger on

In the first sentence "There is a notion of regular sequence which is slightly weaker than that of a regular sequence" the word "quasi" is presumably missing.

Comment #8824 by Ryo Suzuki on

It might be good to add this lemma:

Let be a ring and let . Set . Let be an -module. Suppose that, for any and a homogeneous polynomial of degree , if then the coefficients of is in . Then is a -quasi-regular sequence.

It is essentially proved in the proof of Lemma 00LN. It is also appear in the proof of Lemma 062I and its proof might become cleaner by using this lemma.

Comment #9266 by on

This is essentially just the definition of a quasi-regular sequence and I think we're just using this directly in the places you mention. Going to leave as is for now.

Comment #9788 by Oliver on

In the proof of 00LN, is the induction on descending or ascending?


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.