92.5 Constructing a resolution
In the Noetherian finite type case we can construct a “small” simplicial resolution for finite type ring maps.
Lemma 92.5.1. Let A be a Noetherian ring. Let A \to B be a finite type ring map. Let \mathcal{A} be the category of A-algebra maps C \to B. Let n \geq 0 and let P_\bullet be a simplicial object of \mathcal{A} such that
P_\bullet \to B is a trivial Kan fibration of simplicial sets,
P_ k is finite type over A for k \leq n,
P_\bullet = \text{cosk}_ n \text{sk}_ n P_\bullet as simplicial objects of \mathcal{A}.
Then P_{n + 1} is a finite type A-algebra.
Proof.
Although the proof we give of this lemma is straightforward, it is a bit messy. To clarify the idea we explain what happens for low n before giving the proof in general. For example, if n = 0, then (3) means that P_1 = P_0 \times _ B P_0. Since the ring map P_0 \to B is surjective, this is of finite type over A by More on Algebra, Lemma 15.5.1.
If n = 1, then (3) means that
P_2 = \{ (f_0, f_1, f_2) \in P_1^3 \mid d_0f_0 = d_0f_1,\ d_1f_0 = d_0f_2,\ d_1f_1 = d_1f_2 \}
where the equalities take place in P_0. Observe that the triple
(d_0f_0, d_1f_0, d_1f_1) = (d_0f_1, d_0f_2, d_1f_2)
is an element of the fibre product P_0 \times _ B P_0 \times _ B P_0 over B because the maps d_ i : P_1 \to P_0 are morphisms over B. Thus we get a map
\psi : P_2 \longrightarrow P_0 \times _ B P_0 \times _ B P_0
The fibre of \psi over an element (g_0, g_1, g_2) \in P_0 \times _ B P_0 \times _ B P_0 is the set of triples (f_0, f_1, f_2) of 1-simplices with (d_0, d_1)(f_0) = (g_0, g_1), (d_0, d_1)(f_1) = (g_0, g_2), and (d_0, d_1)(f_2) = (g_1, g_2). As P_\bullet \to B is a trivial Kan fibration the map (d_0, d_1) : P_1 \to P_0 \times _ B P_0 is surjective. Thus we see that P_2 fits into the cartesian diagram
\xymatrix{ P_2 \ar[d] \ar[r] & P_1^3 \ar[d] \\ P_0 \times _ B P_0 \times _ B P_0 \ar[r] & (P_0 \times _ B P_0)^3 }
By More on Algebra, Lemma 15.5.2 we conclude. The general case is similar, but requires a bit more notation.
The case n > 1. By Simplicial, Lemma 14.19.14 the condition P_\bullet = \text{cosk}_ n \text{sk}_ n P_\bullet implies the same thing is true in the category of simplicial A-algebras and hence in the category of sets (as the forgetful functor from A-algebras to sets commutes with limits). Thus
P_{n + 1} = \mathop{\mathrm{Mor}}\nolimits (\Delta [n + 1], P_\bullet ) = \mathop{\mathrm{Mor}}\nolimits (\text{sk}_ n \Delta [n + 1], \text{sk}_ n P_\bullet )
by Simplicial, Lemma 14.11.3 and Equation (14.19.0.1). We will prove by induction on 1 \leq k < m \leq n + 1 that the ring
Q_{k, m} = \mathop{\mathrm{Mor}}\nolimits (\text{sk}_ k \Delta [m], \text{sk}_ k P_\bullet )
is of finite type over A. The case k = 1, 1 < m \leq n + 1 is entirely similar to the discussion above in the case n = 1. Namely, there is a cartesian diagram
\xymatrix{ Q_{1, m} \ar[d] \ar[r] & P_1^ N \ar[d] \\ P_0 \times _ B \ldots \times _ B P_0 \ar[r] & (P_0 \times _ B P_0)^ N }
where N = {m + 1 \choose 2}. We conclude as before.
Let 1 \leq k_0 \leq n and assume Q_{k, m} is of finite type over A for all 1 \leq k \leq k_0 and k < m \leq n + 1. For k_0 + 1 < m \leq n + 1 we claim there is a cartesian square
\xymatrix{ Q_{k_0 + 1, m} \ar[d] \ar[r] & P_{k_0 + 1}^ N \ar[d] \\ Q_{k_0, m} \ar[r] & Q_{k_0, k_0 + 1}^ N }
where N is the number of nondegenerate (k_0 + 1)-simplices of \Delta [m]. Namely, to see this is true, think of an element of Q_{k_0 + 1, m} as a function f from the (k_0 + 1)-skeleton of \Delta [m] to P_\bullet . We can restrict f to the k_0-skeleton which gives the left vertical map of the diagram. We can also restrict to each nondegenerate (k_0 + 1)-simplex which gives the top horizontal arrow. Moreover, to give such an f is the same thing as giving its restriction to k_0-skeleton and to each nondegenerate (k_0 + 1)-face, provided these agree on the overlap, and this is exactly the content of the diagram. Moreover, the fact that P_\bullet \to B is a trivial Kan fibration implies that the map
P_{k_0} \to Q_{k_0, k_0 + 1} = \mathop{\mathrm{Mor}}\nolimits (\partial \Delta [k_0 + 1], P_\bullet )
is surjective as every map \partial \Delta [k_0 + 1] \to B can be extended to \Delta [k_0 + 1] \to B for k_0 \geq 1 (small argument about constant simplicial sets omitted). Since by induction hypothesis the rings Q_{k_0, m}, Q_{k_0, k_0 + 1} are finite type A-algebras, so is Q_{k_0 + 1, m} by More on Algebra, Lemma 15.5.2 once more.
\square
Proposition 92.5.2. Let A be a Noetherian ring. Let A \to B be a finite type ring map. There exists a simplicial A-algebra P_\bullet with an augmentation \epsilon : P_\bullet \to B such that each P_ n is a polynomial algebra of finite type over A and such that \epsilon is a trivial Kan fibration of simplicial sets.
Proof.
Let \mathcal{A} be the category of A-algebra maps C \to B. In this proof our simplicial objects and skeleton and coskeleton functors will be taken in this category.
Choose a polynomial algebra P_0 of finite type over A and a surjection P_0 \to B. As a first approximation we take P_\bullet = \text{cosk}_0(P_0). In other words, P_\bullet is the simplicial A-algebra with terms P_ n = P_0 \times _ A \ldots \times _ A P_0. (In the final paragraph of the proof this simplicial object will be denoted P^0_\bullet .) By Simplicial, Lemma 14.32.3 the map P_\bullet \to B is a trivial Kan fibration of simplicial sets. Also, observe that P_\bullet = \text{cosk}_0 \text{sk}_0 P_\bullet .
Suppose for some n \geq 0 we have constructed P_\bullet (in the final paragraph of the proof this will be P^ n_\bullet ) such that
P_\bullet \to B is a trivial Kan fibration of simplicial sets,
P_ k is a finitely generated polynomial algebra for 0 \leq k \leq n, and
P_\bullet = \text{cosk}_ n \text{sk}_ n P_\bullet
By Lemma 92.5.1 we can find a finitely generated polynomial algebra Q over A and a surjection Q \to P_{n + 1}. Since P_ n is a polynomial algebra the A-algebra maps s_ i : P_ n \to P_{n + 1} lift to maps s'_ i : P_ n \to Q. Set d'_ j : Q \to P_ n equal to the composition of Q \to P_{n + 1} and d_ j : P_{n + 1} \to P_ n. We obtain a truncated simplicial object P'_\bullet of \mathcal{A} by setting P'_ k = P_ k for k \leq n and P'_{n + 1} = Q and morphisms d'_ i = d_ i and s'_ i = s_ i in degrees k \leq n - 1 and using the morphisms d'_ j and s'_ i in degree n. Extend this to a full simplicial object P'_\bullet of \mathcal{A} using \text{cosk}_{n + 1}. By functoriality of the coskeleton functors there is a morphism P'_\bullet \to P_\bullet of simplicial objects extending the given morphism of (n + 1)-truncated simplicial objects. (This morphism will be denoted P^{n + 1}_\bullet \to P^ n_\bullet in the final paragraph of the proof.)
Note that conditions (b) and (c) are satisfied for P'_\bullet with n replaced by n + 1. We claim the map P'_\bullet \to P_\bullet satisfies assumptions (1), (2), (3), and (4) of Simplicial, Lemmas 14.32.1 with n + 1 instead of n. Conditions (1) and (2) hold by construction. By Simplicial, Lemma 14.19.14 we see that we have P_\bullet = \text{cosk}_{n + 1}\text{sk}_{n + 1}P_\bullet and P'_\bullet = \text{cosk}_{n + 1}\text{sk}_{n + 1}P'_\bullet not only in \mathcal{A} but also in the category of A-algebras, whence in the category of sets (as the forgetful functor from A-algebras to sets commutes with all limits). This proves (3) and (4). Thus the lemma applies and P'_\bullet \to P_\bullet is a trivial Kan fibration. By Simplicial, Lemma 14.30.4 we conclude that P'_\bullet \to B is a trivial Kan fibration and (a) holds as well.
To finish the proof we take the inverse limit P_\bullet = \mathop{\mathrm{lim}}\nolimits P^ n_\bullet of the sequence of simplicial algebras
\ldots \to P^2_\bullet \to P^1_\bullet \to P^0_\bullet
constructed above. The map P_\bullet \to B is a trivial Kan fibration by Simplicial, Lemma 14.30.5. However, the construction above stabilizes in each degree to a fixed finitely generated polynomial algebra as desired.
\square
Lemma 92.5.3. Let A be a Noetherian ring. Let A \to B be a finite type ring map. Let \pi , \underline{B} be as in (92.4.0.1). If \mathcal{F} is an \underline{B}-module such that \mathcal{F}(P, \alpha ) is a finite B-module for all \alpha : P = A[x_1, \ldots , x_ n] \to B, then the cohomology modules of L\pi _!(\mathcal{F}) are finite B-modules.
Proof.
By Lemma 92.4.1 and Proposition 92.5.2 we can compute L\pi _!(\mathcal{F}) by a complex constructed out of the values of \mathcal{F} on finite type polynomial algebras.
\square
Lemma 92.5.4. Let A be a Noetherian ring. Let A \to B be a finite type ring map. Then H^ n(L_{B/A}) is a finite B-module for all n \in \mathbf{Z}.
Proof.
Apply Lemmas 92.4.3 and 92.5.3.
\square
Lemma 92.5.6. Let A \to B be a ring map. Let \pi , \mathcal{O}, \underline{B} be as in (92.4.0.1). For any \mathcal{O}-module \mathcal{F} we have
L\pi _!(\mathcal{F}) = L\pi _!(Li^*\mathcal{F}) = L\pi _!(\mathcal{F} \otimes _\mathcal {O}^\mathbf {L} \underline{B})
in D(\textit{Ab}).
Proof.
It suffices to verify the assumptions of Cohomology on Sites, Lemma 21.39.12 hold for \mathcal{O} \to \underline{B} on \mathcal{C}_{B/A}. We will use the results of Remark 92.5.5 without further mention. Choose a resolution P_\bullet of B over A to get a suitable cosimplicial object U_\bullet of \mathcal{C}_{B/A}. Since P_\bullet \to B induces a quasi-isomorphism on associated complexes of abelian groups we see that L\pi _!\mathcal{O} = B. On the other hand L\pi _!\underline{B} is computed by \underline{B}(U_\bullet ) = B. This verifies the second assumption of Cohomology on Sites, Lemma 21.39.12 and we are done with the proof.
\square
Lemma 92.5.7. Let A \to B be a ring map. Let \pi , \mathcal{O}, \underline{B} be as in (92.4.0.1). We have
L\pi _!(\mathcal{O}) = L\pi _!(\underline{B}) = B \quad \text{and}\quad L_{B/A} = L\pi _!(\Omega _{\mathcal{O}/A} \otimes _\mathcal {O} \underline{B}) = L\pi _!(\Omega _{\mathcal{O}/A})
in D(\textit{Ab}).
Proof.
This is just an application of Lemma 92.5.6 (and the first equality on the right is Lemma 92.4.3).
\square
Here is a special case of the fundamental triangle that is easy to prove.
Lemma 92.5.8. Let A \to B \to C be ring maps. If B is a polynomial algebra over A, then there is a distinguished triangle L_{B/A} \otimes _ B^\mathbf {L} C \to L_{C/A} \to L_{C/B} \to L_{B/A} \otimes _ B^\mathbf {L} C[1] in D(C).
Proof.
We will use the observations of Remark 92.5.5 without further mention. Choose a resolution \epsilon : P_\bullet \to C of C over B (for example the standard resolution). Since B is a polynomial algebra over A we see that P_\bullet is also a resolution of C over A. Hence L_{C/A} is computed by \Omega _{P_\bullet /A} \otimes _{P_\bullet , \epsilon } C and L_{C/B} is computed by \Omega _{P_\bullet /B} \otimes _{P_\bullet , \epsilon } C. Since for each n we have the short exact sequence 0 \to \Omega _{B/A} \otimes _ B P_ n \to \Omega _{P_ n/A} \to \Omega _{P_ n/B} \to 0 (Algebra, Lemma 10.138.9) and since L_{B/A} = \Omega _{B/A}[0] (Lemma 92.4.7) we obtain the result.
\square
Example 92.5.9. Let A \to B be a ring map. In this example we will construct an “explicit” resolution P_\bullet of B over A of length 2. To do this we follow the procedure of the proof of Proposition 92.5.2, see also the discussion in Remark 92.5.5.
We choose a surjection P_0 = A[u_ i] \to B where u_ i is a set of variables. Choose generators f_ t \in P_0, t \in T of the ideal \mathop{\mathrm{Ker}}(P_0 \to B). We choose P_1 = A[u_ i, x_ t] with face maps d_0 and d_1 the unique A-algebra maps with d_ j(u_ i) = u_ i and d_0(x_ t) = 0 and d_1(x_ t) = f_ t. The map s_0 : P_0 \to P_1 is the unique A-algebra map with s_0(u_ i) = u_ i. It is clear that
P_1 \xrightarrow {d_0 - d_1} P_0 \to B \to 0
is exact, in particular the map (d_0, d_1) : P_1 \to P_0 \times _ B P_0 is surjective. Thus, if P_\bullet denotes the 1-truncated simplicial A-algebra given by P_0, P_1, d_0, d_1, and s_0, then the augmentation \text{cosk}_1(P_\bullet ) \to B is a trivial Kan fibration. The next step of the procedure in the proof of Proposition 92.5.2 is to choose a polynomial algebra P_2 and a surjection
P_2 \longrightarrow \text{cosk}_1(P_\bullet )_2
Recall that
\text{cosk}_1(P_\bullet )_2 = \{ (g_0, g_1, g_2) \in P_1^3 \mid d_0(g_0) = d_0(g_1), d_1(g_0) = d_0(g_2), d_1(g_1) = d_1(g_2)\}
Thinking of g_ i \in P_1 as a polynomial in x_ t the conditions are
g_0(0) = g_1(0),\quad g_0(f_ t) = g_2(0),\quad g_1(f_ t) = g_2(f_ t)
Thus \text{cosk}_1(P_\bullet )_2 contains the elements y_ t = (x_ t, x_ t, f_ t) and z_ t = (0, x_ t, x_ t). Every element G in \text{cosk}_1(P_\bullet )_2 is of the form G = H + (0, 0, g) where H is in the image of A[u_ i, y_ t, z_ t] \to \text{cosk}_1(P_\bullet )_2. Here g \in P_1 is a polynomial with vanishing constant term such that g(f_ t) = 0 in P_0. Observe that
g = x_ t x_{t'} - f_ t x_{t'} and
g = \sum r_ t x_ t with r_ t \in P_0 if \sum r_ t f_ t = 0 in P_0
are elements of P_1 of the desired form. Let
Rel = \mathop{\mathrm{Ker}}(\bigoplus \nolimits _{t \in T} P_0 \longrightarrow P_0),\quad (r_ t) \longmapsto \sum r_ tf_ t
We set P_2 = A[u_ i, y_ t, z_ t, v_ r, w_{t, t'}] where r = (r_ t) \in Rel, with map
P_2 \longrightarrow \text{cosk}_1(P_\bullet )_2
given by y_ t \mapsto (x_ t, x_ t, f_ t), z_ t \mapsto (0, x_ t, x_ t), v_ r \mapsto (0, 0, \sum r_ t x_ t), and w_{t, t'} \mapsto (0, 0, x_ t x_{t'} - f_ t x_{t'}). A calculation (omitted) shows that this map is surjective. Our choice of the map displayed above determines the maps d_0, d_1, d_2 : P_2 \to P_1. Finally, the procedure in the proof of Proposition 92.5.2 tells us to choose the maps s_0, s_1 : P_1 \to P_2 lifting the two maps P_1 \to \text{cosk}_1(P_\bullet )_2. It is clear that we can take s_ i to be the unique A-algebra maps determined by s_0(x_ t) = y_ t and s_1(x_ t) = z_ t.
Comments (0)