arXiv reader

Deconstructing Auslander’s formulas, I. Fundamental sequences associated with additive functors

Alex Martsinkovsky

arXiv
2410.14097v3 · PDF · 25 July 2026
Categories
math.RT, math.AT, math.CT, math.KT, math.RA
Licence
CC BY 4.0 The site shows the text.

Abstract

For any additive functor from modules (or, more generally, from an abelian category with enough projectives or injectives), we construct long sequences tying up together the derived functors, the satellites, and the stabilizations of the functor. For half-exact functors, the obtained sequences are exact. For general functors, nontrivial homology may only appear at the derived functors. Specializing to the familiar Hom and tensor product functors on finitely presented modules, we recover the classical formulas of Auslander. Unlike those formulas, our results hold for arbitrary rings and arbitrary modules, finite or infinite. The same formalism leads to universal coefficient theorems for homology and cohomology of arbitrary complexes. The new results are even more explicit for the cohomology of projective complexes and the homology of flat complexes.

Items 63

Example

Example 2.1

6GT5

The well-known Auslander-Reiten formula

D_{\mathbf{J}}\operatorname{\mathrm{Ext}}^{1}_{\Lambda}(A,B) \simeq \overline{\operatorname{\mathrm{Hom}}}_{\Lambda} (B, D_{\mathbf{J}}\operatorname{\mathrm{Tr}} A),

is a basic tool in Auslander-Reiten theory. Here \Lambda is an algebra over a commutative ring \Bbbk (which could be \mathbb{Z}), A and B are (left) \Lambda-modules, A is finitely presented, \mathbf{J} is an injective \Bbbk-module, D_{\mathbf{J}} := \operatorname{\mathrm{Hom}}_{\Bbbk}(\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} , \mathbf{J}), and the transpose is computed using any finite presentation of A. The overlined \operatorname{\mathrm{Hom}} on the right-hand side stands for \operatorname{\mathrm{Hom}} modulo injectives, i.e., the quotient of the usual \operatorname{\mathrm{Hom}} by the subgroup of all maps factoring through injectives. It “neutralizes” the projective ambiguity in the definition of the transpose, which is turned into an injective ambiguity by D_{\mathbf{J}}. If A is not finitely presented, then the left-hand side still makes sense, but the right-hand side is not well-defined. This may happen even when A is finitely presented but the transpose is computed via an infinite presentation. In fact, this may happen even for A \simeq 0. To see that, assume that \Lambda is not left-coherent. Then, by a theorem of Chase [Chase, Theorem 2.1], \Lambda_{\Lambda}^{N} is not flat for some N. Now choose the following projective resolution of the zero module

0 \longrightarrow \prescript{}{\Lambda}\Lambda^{(N)} \overset{\mathrm{id}}\longrightarrow \prescript{}{\Lambda}\Lambda^{(N)} \longrightarrow 0 \longrightarrow 0 \longrightarrow 0.

Computing the transpose with this resolution, we have \operatorname{\mathrm{Tr}} A \simeq \Lambda_{\Lambda}^{N}. Now choose \Bbbk : = \mathbb{Z} and \mathbf{J} := \mathbb{Q}/\mathbb{Z}. By a theorem of Lambek [L64, p. 238], the character module D_{\mathbf{J}}\operatorname{\mathrm{Tr}} A is not injective. Evaluating the right-hand side of the formula on B := D_{\mathbf{J}}\operatorname{\mathrm{Tr}} A we have a nonzero group, whereas the left-hand side is zero. Thus the formula fails.

Example

Example 2.2

90WS

Let A be a finitely presented \Lambda-module, and e_{A} : A \longrightarrow A^{\ast\ast} the canonical evaluation map. Then there is an exact sequence

\xymatrix { 0 \ar[r] & \operatorname{\mathrm{Ext}}^{1}(\operatorname{\mathrm{Tr}} A, \Lambda) \ar[r] & A \ar[r]^{e_{A}} & A^{\ast\ast} \ar[r] & \operatorname{\mathrm{Ext}}^{2}(\operatorname{\mathrm{Tr}} A, \Lambda) \ar[r] & 0. }

This is the \Lambda-component of the exact sequence of functors [AB, Proposition 2.6 (a)]

\xymatrix@R1pt@C15pt { 0 \ar[r] & \operatorname{\mathrm{Ext}}^{1}(\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \ar[r]^{\rho_{A}} & (A^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \operatorname{\mathrm{Ext}}^{2}(\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & 0. }\footnote{The reader may check that $\rho_{A}$ is the zeroth right-derived transformation for $A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} $.}

When A is not finitely presented, the sequence (2.2) fails to be exact already for vector spaces over a field. Indeed, in that case the end terms vanish, forcing \rho_{A}, and hence e_{A}, to be an isomorphism, which is impossible since A^{\ast\ast} has a larger dimension than A.

Example

Example 2.3

PPVP

Let A and B be left \Lambda-modules. Next we examine the canonical map

\varphi : A^{\ast} \otimes B \longrightarrow (A, B) : (f, b) \mapsto \big(a \mapsto f(a)b\big).

As B is canonically isomorphic to (\Lambda, B), we interpret \varphi as the map

(A, \Lambda) \otimes (\Lambda, B) \to (A,B)

given by the “composition" of the arguments. Therefore the image of \varphi consists of all maps A \to B factoring through finitely generated projectives.

Under the assumption that A is finitely presented, we have an exact sequence of functors [A66, Proposition 7.1].

\xymatrix@R1pt@C15pt { 0 \ar[r] & \operatorname{\mathrm{Tor}}_{2}(\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & A^{\ast} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \ar[r]^{\lambda_{A}} & (A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \operatorname{\mathrm{Tor}}_{1}(\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & 0. }\footnote{The reader may check that $\lambda_{A}$ is the zeroth left-derived transformation for $(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )$.}

However, when A is not finitely generated, this sequence fails to be exact already for vector spaces over a field. In that case, the end terms vanish, and the exactness of the sequence would force \lambda_{A} to be an isomorphism, which is impossible since, evaluating it on B : = A, we would have the identity map on A factor through a finite-dimensional vector space, a contradiction.

Remark

Remark

CG03

The Parallelogram Lemma Suppose, in the commutative diagram

\begin{tikzcd} A \ar[r] \ar[rd] & B \ar[d, "f"'] \ar[rd] \\ & C \ar[r] & D \end{tikzcd}

the two peripheral compositions are zero.

  1. a) If f is epic and the upper peripheral sequence A \to B \to D is exact at B, then the lower peripheral sequence A \to C \to D is exact at C.

  2. b) If f is monic and the lower peripheral sequence A \to C \to D is exact at C, then the upper peripheral sequence A \to B \to D is exact at B.

Proof of Remark

Obvious.

Theorem

Theorem 3.2

DCPH

  1. The right fundamental sequence (?) is a complex of functors, whose homology is zero except, possibly, at the terms R^{i}F.

  2. If F is half-exact, then the sequence is exact.

  3. If F is epi-preserving, then \beta_{F} is epic.

  4. If F is mono-preserving, then \overline{F} = 0 and \rho_{F} is monic.

  5. \rho_{F} is an isomorphism if and only if F is left-exact.

Used by Example 3.5
Proof of Theorem 3.2

(1) The claim that  (?) is a complex is easily checked by a diagram chase of arrows (with or without elements) in (?) and is left to the reader as an exercise.

For the exactness claim, we have to show the exactness at the first two terms of each row. For the first row, this is true by the definition of the injective stabilization.

For the second row, since \kappa_{1} is monic, it is easy to check that the fragment of (?) connecting R^{0}F and R^{1}F is part of the circular sequence of the triangle F(I^{0}), F(\Sigma B), Z_{1} in the diagram (?), and we are done by the exactness of the circular sequence.

For the third row, since \kappa_{2} is monic, we have the fragment

Z_{1} \longrightarrow \overline{F}(\Sigma^{2} B) \longrightarrow S^{1}F(\Sigma B) \longrightarrow R^{2}F(B)

of the circular sequence of the triangle F(I^{1}), F(\Sigma^{2} B), Z^{2}. Since Z_{1} \longrightarrow R^{1}F(B) is epic, these two modules have the same image in \overline{F}(\Sigma^{2} B), and this yields the desired result for the third row. The same argument applies to the remaining rows.

(2) The top part of (?) yields the commutative diagram

\begin{tikzcd} F(B) \ar[r] \ar[rd] & R^{0}F(B) \ar[d, tail] \ar[rd, "\iota \beta_{F}"] \ar[r, "\beta_{F}"] & \overline{F}(\Sigma B) \ar[d, "\iota", tail] \\ & F(I^{0}) \ar[r] & F(\Sigma B) \end{tikzcd}

If F is half-exact, the lower peripheral sequence is exact at F(I^{0}). By the parallelogram lemma, b), the top peripheral sequence is exact at R^{0}F(B). Since \iota is monic, the top row is also exact at that spot, which is the desired claim for R^{0}F.

To prove the exactness at R^{1}F, we examine the commutative diagram

\begin{tikzcd} F(\Sigma B) \ar[r] \ar[rd, "\eta \pi"] \ar[d, two heads, "\pi"'] & Z_{1} \ar[d, two heads] \ar[rd] \\ S^{1}F(B) \ar[r, "\eta"'] & R^{1}F(B) \ar[r] & \overline{F}(\Sigma^{2} B) \end{tikzcd}

The circular sequence for the composable pair F(e_{1}), F(m_{1}) produces a sequence

\operatorname{\mathrm{Ker}} F(e_{1}) \longrightarrow Z_{1} \longrightarrow \overline{F}(\Sigma^{2} B),

which is exact at Z_{1}. If F is half-exact, then \operatorname{\mathrm{Ker}} F(e_{1}) = \operatorname{\mathrm{Im}} F(m), which shows that the top peripheral sequence in the parallelogram above becomes exact at Z_{1}. By the parallelogram lemma, a), the lower peripheral sequence is exact at R^{1}F(B). Since \pi is epic, the bottom row is exact at the same spot, which is the desired claim for R^{1}F. The same argument applies to the remaining R^{i}F.

(3) If F is epi-preserving, then S^{1}F = 0, and we are done by the exactness of the circular sequence of the diagram (?) at \overline{F}(\Sigma B).

(4) If F is mono-preserving, then \overline{F} = 0, and we are done by (1).

(5) This is well-known.

Corollary

Corollary 3.3

XW6Z

If F is a right-exact functor, then the canonical natural transformation R^{i}F \longrightarrow \overline{F} \circ \Sigma^{i+1} is an isomorphism for each i \geq 1.

Proof of Corollary 3.3

Since F is right-exact, S^{i}F = 0 for each i \geq 1. The result now follows from the exactness of the fundamental sequence.

Remark

Remark 3.4

MNQA

Suppose F is left-exact but not exact. Then the fundamental sequence is exact by (2) and \rho_{F} is an isomorphism by (5). It follows that the converse of (3) is not true in general.

Example

Example 3.5

RE5A

Continue to assume that \mathscr{A} is the category of left \Lambda-modules over a ring \Lambda. Let A be a right \Lambda-module and F := A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm}. The sub-stabilization of A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} is denoted by A \operatorname{\overset{\rightharpoondown}{\otimes}} \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} (see [MR-3, MR-2, MR-1], where it was called the injective stabilization of the tensor product, for an extensive treatment of this concept). The initial segment of the right fundamental sequence (?) becomes

{0 \longrightarrow A \operatorname{\overset{\rightharpoondown}{\otimes}} \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow R^{0}(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow A \operatorname{\overset{\rightharpoondown}{\otimes}} (\Sigma \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow 0.}

By the preceding theorem, it is exact. Evaluating it on \prescript{}{\Lambda}\Lambda, we have another exact sequence (in the category of right \Lambda-modules)

0 \longrightarrow \mathfrak{s}(A) \longrightarrow A \longrightarrow R^{0}(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )(\Lambda) \longrightarrow A \operatorname{\overset{\rightharpoondown}{\otimes}} \Sigma \Lambda \longrightarrow 0,

where \mathfrak{s} denotes the torsion radical introduced in [MR-2].

Uses Theorem 3.2 · Used by Corollary 3.6
Corollary

Corollary 3.6

Q6AR

Suppose \Lambda is left hereditary. Then R^{0}(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \simeq I(A\otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ), the quotient functor of A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} by the largest subfunctor vanishing on injectives, and the sequence (3.5) becomes

0 \longrightarrow A \operatorname{\overset{\rightharpoondown}{\otimes}} \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow I(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow 0.

In particular, I(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is an exact functor.(The functor I(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is not as mysterious as it may seem. Without any assumption on \Lambda or on A, its value on a left module B can be computed by applying A\otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} to a cosyzygy sequence 0 \to B \to I \to \Sigma B \to 0 and taking the kernel of the induced map A \otimes I \to A \otimes \Sigma B.)

Uses Example 3.5
Proof of Corollary 3.6

Since the cosyzygy module of any module is injective, A \operatorname{\overset{\rightharpoondown}{\otimes}} (\Sigma_{\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) = 0. Since the fundamental sequence is exact, we then have R^{0}(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \simeq I(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ). To show that I(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is exact, we first notice that it preserves epimorphisms because it is a quotient of the right-exact functor A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm}. Being a zeroth right-derived functor, it is also left-exact, and therefore exact.

Definition

Definition 3.7

7R7T

w(F) := \operatorname{\mathrm{Ker}} f is called the defect of F.

Proposition

Proposition 3.8

N5NE

If F is a finitely presented covariant functor, then

R^{i} \rho_{F} : R^{i}F \longrightarrow \operatorname{\mathrm{Ext}}^{i}(w(F), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )

is an isomorphism for all i.

Example

Example 3.9

LM3G

We again set F := A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} and assume that A is finitely presented, i.e., there is an exact sequence

P_{1} \overset{d}\longrightarrow P_{0} \overset{p}\longrightarrow A \longrightarrow 0,

with P_{0} and P_{1} finitely generated projectives. This sequence gives rise to an exact sequence of functors

P_{1} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \overset{d \otimes \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} }\longrightarrow P_{0} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow A\otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow 0,

which is isomorphic to the sequence

(P_{1}^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \overset{(d^{\ast}, \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} )}\longrightarrow (P_{0}^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow 0.

Thus A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} is finitely presented. To compute its defect, we use the defining exact sequence for the transpose of A

0 \longrightarrow A^{\ast} \overset{p^{\ast}}\longrightarrow P_{0}^{\ast} \overset{d^{\ast}}\longrightarrow P_{1}^{\ast} \longrightarrow \operatorname{\mathrm{Tr}} A \longrightarrow 0,

which shows that w(A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \simeq A^{\ast}. As a result, the exact sequence (3.5) becomes

\xymatrix { 0 \ar[r] & A \operatorname{\overset{\rightharpoondown}{\otimes}} \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \ar[r] & A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \ar[r]^{\rho_{A \otimes \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} }} & (A^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & A \operatorname{\overset{\rightharpoondown}{\otimes}} \Sigma \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \ar[r] & 0. }

One can also check that \rho_{A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} } evaluates on a module B according to the rule

a \otimes b \mapsto l : f \mapsto f(a)b.

To further explicate the structure of (3.9), we compute A \operatorname{\overset{\rightharpoondown}{\otimes}} \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm}. The foregoing calculations produce a commutative diagram with an exact row

\xymatrix { 0 \ar[r] & (\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & (P_{1}^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r]^{(d^{\ast}, \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} )} & (P_{0}^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r]^{\pi} \ar[rd]_{(p^{\ast}, \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} )} & A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \ar[r] \ar[d]^{\rho_{A \otimes \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} }} & 0 \\ & & & & (A^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) }

The circular sequence for the composable morphisms \pi and \rho_{A \otimes \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} } yields a short exact sequence

0 \longrightarrow \operatorname{\mathrm{Ker}} \pi \longrightarrow \operatorname{\mathrm{Ker}} (p^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow \operatorname{\mathrm{Ker}} \rho_{A \otimes \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} } \longrightarrow 0

Notice that \operatorname{\mathrm{Ker}} \pi = \operatorname{\mathrm{Im}} (d^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) and \operatorname{\mathrm{Ker}} \rho_{A \otimes \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} } = A \operatorname{\overset{\rightharpoondown}{\otimes}} \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm}. To identify the middle term, choose an epimorphism q : Q \twoheadrightarrow A^{\ast} with Q projective. This gives the first three terms of a projective resolution

\ldots \longrightarrow Q \overset{p^{\ast}q}\longrightarrow P_{0}^{\ast} \overset{d^{\ast}}\longrightarrow P_{1}^{\ast} \longrightarrow \operatorname{\mathrm{Tr}} A \longrightarrow 0

of \operatorname{\mathrm{Tr}} A, which is enough for computing \operatorname{\mathrm{Ext}}^{1}(\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ). Since (q, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is monic, \operatorname{\mathrm{Ker}}(p^{\ast}q, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) = \operatorname{\mathrm{Ker}}(p^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ), and we recover the well-known isomorphism.

\operatorname{\mathrm{Ext}}^{1}(\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \simeq \operatorname{\mathrm{Ker}} (p^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )/\operatorname{\mathrm{Im}} (d^{\ast}, \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} ) \simeq A \operatorname{\overset{\rightharpoondown}{\otimes}} \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} .

Finally, using dimension shift on the covariant argument of \operatorname{\mathrm{Ext}}, we rewrite (3.9) in a more explicit form

0 \longrightarrow \operatorname{\mathrm{Ext}}^{1}(\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \overset{\rho_{A \otimes \hspace{0.04cm} \rule{1.5mm}{.4pt} \hspace{0.04cm} }}\longrightarrow (A^{\ast}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow \operatorname{\mathrm{Ext}}^{2}(\operatorname{\mathrm{Tr}} A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow 0,

thus recovering Auslander’s exact sequence (2.2).

Proposition

Proposition 4.1

7YS3

Let \mathscr{T} be the class of injectively stable covariant functors and \mathscr{F} the class of mono-preserving covariant functors. Then (\mathscr{T}, \mathscr{F}) is a hereditary torsion theory in the (large) category of additive covariant functors.

Remark

Remark 4.2

39D6

Returning to an arbitrary (i.e., not necessarily left-hereditary) ring, we have, by [A66, Proposition 4.3], that defect-zero half-exact finitely presented functors on the module category are precisely the injective objects in the category of defect-zero finitely presented functors. We also have [idem.] that such functors are precisely direct summands of the functors \operatorname{\mathrm{Ext}}^{1}(D, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ). It is interesting to revisit this result from a constructive point of view. Over a left hereditary ring, if we know a short exact sequence 0 \to A \to B \to C \to 0 which gives rise to a defining exact sequence for F,

0 \longrightarrow (C, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow (B, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow (A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow F \longrightarrow 0,

then [A66, Lemma 4.1 and Proposition 4.3] show that we can set D := C. In general, we may ask if there is a “coordinate-free” description of D. For example, the foregoing argument and the universal coefficient formula for cohomology (see Section ? below) show that, for F := \mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ), we can set

D:= \mathrm{H}_{n-1}(\mathbf{C}) \simeq w(\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )).

While this formula may be viewed as being coordinate-free, it requires the use of a different functor, i.e., \mathrm{H}_{n-1}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ).

Recall that a direct summand of a covariant \operatorname{\mathrm{Ext}} need not be an \operatorname{\mathrm{Ext}}, as was shown by Auslander in [A69], refuting an earlier conjecture of his own. The conjecture is, however, true if the domain category of F has denumerable sums [Fr66], or if it is of finite global dimension [A66, Theorem 4.8], or if the fixed argument of \operatorname{\mathrm{Ext}} is artinian (or, more generally, is noetherian and has the descending chain condition for the images of nested endomorphisms) [M-2016, Proposition 11]. In those cases, F \simeq \operatorname{\mathrm{Ext}}^{1}(D, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ), and one may ask how to recover D from the functor \operatorname{\mathrm{Ext}}^{1}(D, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ). As D is defined only up to projectively stable equivalence, one has to look for an answer in the stable category. In that category, the isomorphism type of D is uniquely determined by the isomorphism type of the corresponding representable functor (\underline{D, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }), and this immediately leads to the desired solution:

(\underline{D, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \simeq S_{1}\operatorname{\mathrm{Ext}}^{1}(D, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \simeq S_{1}F,\footnote{The reader who feels the given answer is not ``constructive'' enough for computational purposes may rethink what it means to compute and start thinking of computing with functors.}

where S_{1} denotes the left satellite.

Lemma

Lemma 5.1

HKXE

Let \mathbf{C} : \ldots \longrightarrow \mathrm{C}_{n+1} \overset{d_{n+1}}\longrightarrow \mathrm{C}_{n} \overset{d_{n}}\longrightarrow \mathrm{C}_{n-1} \longrightarrow \ldots be an arbitrary complex. Then

w(\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )) \simeq \operatorname{\mathrm{Ker}} (\mathrm{C}_{n}/\mathrm{B}_{n} \to \mathrm{C}_{n-1}) \simeq \mathrm{H}_{n}(\mathbf{C}).

As a consequence,

R^{i}(\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )) \simeq \operatorname{\mathrm{Ext}}^{i}(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )

for all integers i.

Remark

Remark

YFCC

Universal Coefficient Theorem for Cohomology of Arbitrary Complexes Let \mathbf{C} be an arbitrary complex. Then the right fundamental sequence for \mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is of the form

\begin{tikzcd} 0 \ar[r] & \overline{\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \ar[r] & \mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=Z}] & (\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (Z) [pos=0.3]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ & \overline{\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \circ \Sigma \ar[r] & S^{1} (\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )) \ar[r] \ar[d, phantom, ""{coordinate, name=Y}] & \operatorname{\mathrm{Ext}}^{1}(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (Y) [near end]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ & \overline{\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \circ \Sigma^{2} \ar[r] & S^{2}(\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )) \ar[r] \ar[d, phantom, ""{coordinate, name=X}] & \operatorname{\mathrm{Ext}}^{2}(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (X) [near end]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ & \overline{\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \circ \Sigma^{3} \ar[r] & S^{3}(\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )) \ar[r] & \ldots \end{tikzcd}

where \overline{\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) can be computed by (?). This sequence is exact except, possibly, at the rightmost terms. If \mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is half-exact(This happens, for example, when \mathbf{C} is a complex of projectives, see Lemma 5.3 below.), then this sequence is exact.

Lemma

Lemma 5.2

TJVG

Suppose (F^{i})_{i \in \mathbb{Z}} is a cohomological functor, i.e., an exact \delta-functor ([Gr57, 2.1]).(In this lemma, our definition of exact \delta-functor is a little more general than the quoted one. To wit, the integer n can vary over the entire \mathbb{Z}.) Then, for each integer n, the natural transformation \theta^{n} : S^{1}F^{n} \longrightarrow \overline{F^{n + 1}} induced by \delta^{n} is an isomorphism.

Used by Proof
Proof of Lemma 5.2

For a module B, choose an exact sequence 0 \to B \to I \to \Sigma B \to 0 with an injective I, apply F, and pass to the corresponding long exact sequence.

Lemma

Lemma 5.3

OJMS

Suppose \mathbf{C} is a projective complex. Then (\mathrm{H}^{i}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ))_{i \in \mathbb{Z}} is a cohomological functor, and, for each integer n, the canonical natural transformation

\xi^{n-1} : \overline{\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \longrightarrow \operatorname{\mathrm{Ext}}^{1}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )

is an isomorphism.

Used by Proof
Proof of Lemma 5.3

Since \mathbf{C} is projective, applying the functor (\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) to a short exact sequence of modules produces a short exact sequence of complexes. The first result now follows from the fact that cohomology is a cohomological functor on complexes.

To prove the second assertion, apply the Yoneda embedding to the short exact sequence (?), pass to the cohomology long exact sequence, and compare the result with (?).

Proposition

Proposition 5.4

7MFN

Suppose \mathbf{C} is a projective complex. Then the natural transformation \theta^{n} : S^{1}(\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )) \longrightarrow \overline{\mathrm{H}^{n+1}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) induced by \delta^{n} is a functor isomorphism. The composition

\xi^{n}\theta^{n} : S^{1}(\mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )) \longrightarrow \operatorname{\mathrm{Ext}}^{1}(\mathrm{C}_{n}/\mathrm{B}_{n},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )

is an isomorphism.

Proof of Proposition 5.4

By Lemma 5.3, (\mathrm{H}^{i}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ))_{i \in \mathbb{Z}} is a cohomological functor. The first assertion is then a consequence of Lemma 5.2. The second assertion follows from Lemma 5.3.

Uses Lemma 5.3, Lemma 5.2
Remark

Remark

DC9L

Universal Coefficient Theorem for Cohomology of Projective Complexes Let \mathbf{C} be a projective complex. Then the right fundamental sequence for \mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is exact and of the form

\begin{tikzcd}[cramped, sep=small] 0 \ar[r] & \operatorname{\mathrm{Ext}}^{1}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \mathrm{H}^{n}(\mathbf{C}, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=Z}] & (\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (Z) [pos=0.3]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ & \operatorname{\mathrm{Ext}}^{2}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \operatorname{\mathrm{Ext}}^{1}(\mathrm{C}_{n}/\mathrm{B}_{n},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=Y}] & \operatorname{\mathrm{Ext}}^{1}(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (Y) [near end]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ & \operatorname{\mathrm{Ext}}^{3}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \operatorname{\mathrm{Ext}}^{2}(\mathrm{C}_{n}/\mathrm{B}_{n},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=X}] & \operatorname{\mathrm{Ext}}^{2}(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (X) [near end]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ & \operatorname{\mathrm{Ext}}^{4}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \operatorname{\mathrm{Ext}}^{3}(\mathrm{C}_{n}/\mathrm{B}_{n},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \ldots \end{tikzcd}
Remark

Remark 5.5

S8VQ

Yuya Otake informed me (October 30, 2024) he had also obtained the UCT for cohomology of projective complexes, see [O-25, Corollary 3.1].

Definition

Definition 6.1

YRST

We shall refer to (?) as the left fundamental sequence of F.

Theorem

Theorem 6.2

ATAS

  1. The left fundamental sequence (?) is a complex, whose homology is zero except possibly, at the terms L_{i}F.

  2. If F is half-exact, then the sequence is exact.

  3. If F is mono-preserving, then \alpha_{F} is monic.

  4. If F is epi-preserving, then \underline{F} = 0 and \lambda_{F} is epic.

  5. \lambda_{F} is an isomorphism if and only if F is right-exact.

Proof of Theorem 6.2

Similar to the proof of Theorem 3.2.

Corollary

Corollary 6.3

2LZX

Suppose F is a left-exact functor. Then the canonical natural transformation \underline{F} \circ \Omega^{i+1} \longrightarrow L_{i}F is an isomorphism for each i \geq 1.

Remark

Remark 6.4

54MX

Assume \mathscr{A} is the category of left \Lambda-modules over a ring \Lambda. Recall that the canonical natural transformation F(\Lambda) \otimes _{\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} } \longrightarrow F is an isomorphism on \Lambda. If F preserves direct sums, then this transformation is an isomorphism on all projectives, and therefore it is isomorphic to the zeroth left-derived transformation for F. In particular, L_{i}F \simeq \operatorname{\mathrm{Tor}}_{i}(F(\Lambda), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) for all i.

Example

Example 6.5

HGTX

Let F := (A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ). It is easy to see that the quot-stabilization of F is (A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) modulo projectives, i.e., \underline{F} \simeq (\underline{A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) in the classical notation.(This is not a tautology but a fortuitous (if not unfortunate) coincidence; the underline notation for the Hom modulo projectives seems to predate the underline notation for the quot-stabilization. As we shall see later, the situation is dramatically different for contravariant functors.) The left fundamental sequence then becomes

\xymatrix { 0 \ar[r] & (\underline{A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \circ \Omega \ar[r]^{\alpha_{F}} & L_{0}(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r]^{\lambda_{A}} & (A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & (\underline{A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \ar[r] & 0. }

Since (A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is left-exact, this sequence is exact.

Corollary

Corollary 6.6

M0P1

Suppose \Lambda is left hereditary. Then L_{0}(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \simeq P(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ), the functor of maps with domain A factoring through projectives, and the sequence (6.5) becomes

\xymatrix { 0 \ar[r] & P(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r]^{\lambda_{A}} & (A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & (\underline{A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \ar[r] & 0. }

In particular, P(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is an exact functor.

Proof of Corollary 6.6

Since the syzygy module of any module is projective, the first term of the fundamental sequence vanishes. Since the fundamental sequence is exact, we also have L_{0}(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \simeq P(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ). To show that P(A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is exact, we first notice that it preserves monomorphisms because it is a subfunctor of the left-exact functor (A, \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ). Being a zeroth left-derived functor, it is also right-exact, and therefore exact.

Definition

Definition 7.1

MX20

An additive functor from (left) \Lambda-modules to abelian groups is said to be tensor-copresented (TC, for short) if it is isomorphic to the kernel of a natural transformation between univariate tensor product functors. In other words, F is TC if it is defined by an exact sequence

0 \longrightarrow F \longrightarrow A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow B \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ,

where A and B are right \Lambda-modules.

Remark

Remark 7.2

PQLX

The reader who is familiar with the Auslander-Gruson-Jensen (AGJ) duality, will immediately see that TC functors on left modules are precisely the AGJ transforms of finitely presented functors on right modules.(The functor A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} is not in general an injective object in a functor category. Indeed, suppose the ring \Lambda is von Neumann regular but not semisimple. Then there is a nonsplit exact sequence 0 \to A \to B \to C \to 0, and it must be pure. Apply the tensor embedding to get a short exact sequence of functors 0 \to A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \to B \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \to C \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \to 0 (because the original sequence is pure exact). If A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} were injective, then this sequence would split, which would imply that the original sequence is split, too, a contradiction.

Now we can also show that the AGJ transformation is not always exact on tensor-copresented functors. Start with the same sequence as before and apply the tensor embedding to get a short exact sequence in the functor category. Now assume the AGJ transformation is exact and apply it to the obtained sequence. Get a short exact sequence of representable functors. It must split because representables are projective. Now applying the AGJ transformation again, we have the split exact sequence of tensor product functors we started with. But then the original sequence must split, a contradiction.)

Lemma

Lemma 7.3

PCHR

Let A be a right module and G an additive functor from left modules to abelian groups. If G preserves direct sums, then any natural transformation \phi : A \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \longrightarrow G is uniquely determined by its \prescript{}{\Lambda}\Lambda-component.

Proof of Lemma 7.3

It follows from the assumptions that \phi is uniquely determined on free left modules. To see that the same is true for any left module X, take a free presentation of X, apply \phi, and use the naturality of \phi together with the fact that the tensor product is right-exact.

Theorem

Theorem 7.5

D6C3

In the above diagram, K \simeq \underline{F} and the bottom row is a tensor copresentation of \underline{F}.

Proof of Theorem 7.5

We only need to prove the first assertion. We view the first three columns and the maps between them as a short exact sequence of complexes. Since the third column is exact, the map between the first two is a quasi-isomorphism (i.e., induces an isomorphism on the homology). In particular, the first column is exact at F and p = \operatorname{\mathrm{coker}} \lambda_{F}. Furthermore, \lambda_{F} is clearly an isomorphism on projectives. Together with the fact that K_{f} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} is right-exact, it shows that \lambda_{F} is the zeroth left-derived transformation for F, and therefore K is the quot-stabilization of F.

Remark

Remark

C4UQ

Universal Coefficient Theorem for Homology of Arbitrary Complexes Let \mathbf{C} be an arbitrary complex. Then the left fundamental sequence for \mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is of the form

\begin{tikzcd}[cramped, sep=small] \ldots \ar[r] & S_{3}\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=Z}] & \underline{\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )}\circ \Omega^{3} \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (Z) [pos=0.3]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ \operatorname{\mathrm{Tor}}_{2}\big(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \big) \ar[r] & S_{2}\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=Y}] & \underline{\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )}\circ \Omega^{2} \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (Y) [near end]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ \operatorname{\mathrm{Tor}}_{1}\big(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \big) \ar[r] & S_{1}\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=X}] & \underline{\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )}\circ \Omega \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (X) [pos=0.3]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ \mathrm{H}_{n}(\mathbf{C}) \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \ar[r] & \mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \underline{\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )} \ar[r] & 0, \end{tikzcd}

where \underline{\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )} can be computed by (?). This sequence is exact except, possibly, at the leftmost terms. If \mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is half-exact(This happens, for example, when \mathbf{C} is a complex of flats, see Lemma 8.2 below.), then this sequence is exact.

Lemma

Lemma 8.1

B6MR

Suppose (F_{i})_{i \in \mathbb{Z}} is a (co)homological functor, i.e., an exact \delta-functor. Then, for each integer n, the natural transformation \eta_{n} : \underline{F_{n+1}} \longrightarrow S_{1}F_{n} induced by \delta_{n+1} is an isomorphism.

Lemma

Lemma 8.2

T1C9

Suppose \mathbf{C} is a flat complex. Then (\mathrm{H}_{i}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ))_{i \in \mathbb{Z}} is cohomological functor, and, for each integer n, the canonical natural transformation

\tau_{n-1} : \operatorname{\mathrm{Tor}}_{1}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow \underline{\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} })

is an isomorphism.

Proposition

Proposition 8.3

AHEI

Suppose \mathbf{C} is a flat complex. Then the natural transformation \eta_{n} : \underline{\mathrm{H}_{n+1}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} }) \longrightarrow S_{1}(\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} )) induced by \delta_{n+1} is a functor isomorphism. The composition

\eta_{n} \tau_{n} : \operatorname{\mathrm{Tor}}_{1}(\mathrm{C}_{n}/\mathrm{B}_{n},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \longrightarrow S_{1}(\mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ))

is an isomorphism.

Remark

Remark

UTZG

Universal Coefficient Theorem for Homology of Flat Complexes Let \mathbf{C} be a flat complex. Then the left fundamental sequence for \mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) is exact and of the form

\begin{tikzcd}[cramped, sep=small] \ldots \ar[r] & \operatorname{\mathrm{Tor}}_{3}(\mathrm{C}_{n}/\mathrm{B}_{n},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=Z}] & \operatorname{\mathrm{Tor}}_{4}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (Z) [pos=0.3]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ \operatorname{\mathrm{Tor}}_{2}\big(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \big) \ar[r] & \operatorname{\mathrm{Tor}}_{2}(\mathrm{C}_{n}/\mathrm{B}_{n},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=Y}] & \operatorname{\mathrm{Tor}}_{3}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (Y) [near end]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ \operatorname{\mathrm{Tor}}_{1}\big(\mathrm{H}_{n}(\mathbf{C}), \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \big) \ar[r] & \operatorname{\mathrm{Tor}}_{1}(\mathrm{C}_{n}/\mathrm{B}_{n},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] \ar[d, phantom, ""{coordinate, name=X}] & \operatorname{\mathrm{Tor}}_{2}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[dll, rounded corners, to path={ -- ([xshift=2ex]\tikztostart.east) |- (X) [pos=0.3]\tikztonodes -| ([xshift=-2ex]\tikztotarget.west) -- (\tikztotarget)}] \\ \mathrm{H}_{n}(\mathbf{C}) \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} \ar[r] & \mathrm{H}_{n}(\mathbf{C} \otimes \hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & \operatorname{\mathrm{Tor}}_{1}(\mathrm{C}_{n-1}/\mathrm{B}_{n-1},\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} ) \ar[r] & 0 \end{tikzcd}
Remark

Remark 8.4

KGKA

Yuya Otake informed me (October 30, 2024) he had also obtained the UCT for homology of flat complexes, see [O-25, Corollary 3.2].

Definition

Definition 9.1

O0D8

We shall refer to (?) as the right fundamental sequence of F.

Theorem

Theorem 9.2

4UAN

  1. The right fundamental sequence (?) is a complex, whose homology is zero, except, possibly, at the terms R_{i}F.

  2. If F is half-exact, then the sequence is exact.

  3. If F carries monomorphisms to epimorphisms, then \beta_{F} is epic.

  4. If F carries epimorphisms to monomorphisms, then \overline{F} = 0 and \rho_{F} is monic.

  5. \rho_{F} is an isomorphism if and only if F is left-exact.

Proof of Theorem 9.2

Entirely analogous to that of Theorem 3.2.

Corollary

Corollary 9.3

K7V6

If F is a right-exact functor, then the canonical natural transformation R_{i}F \longrightarrow \overline{F} \circ \Omega^{i+1} is an isomorphism for each i \geq 1.

Remark

Remark 9.4

75KJ

Assume \mathscr{A} is the category of left \Lambda-modules over a ring \Lambda. Recall that for any contravariant functor F, the canonical natural transformation F \to \big(\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} , F(\Lambda)\big) is an isomorphism on \Lambda. If F carries coproducts to products, then this natural transformation is an isomorphism on all projectives, and therefore it is isomorphic to \rho_{F}. Hence, the canonical natural transformation

R_{i}\rho_{F} : R_{i}F \longrightarrow \operatorname{\mathrm{Ext}}^{i}(\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} , F(\Lambda))

is an isomorphism for all i, and each row of (?) rewrites as

\overline{F} \circ \Omega^{i} \longrightarrow S^{i}F \longrightarrow \operatorname{\mathrm{Ext}}^{i}(\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} , F(\Lambda)).

Recall also that if F is finitely presented, then the covariant defect v(F) of F is defined as F(\Lambda). If a finite presentation of F arises as the Yoneda embedding of f : B \longrightarrow A, then v(F) \simeq \operatorname{\mathrm{Coker}} f \simeq F(\Lambda), and each row of the left fundamental sequence is of the form

\overline{F} \circ \Omega^{i} \longrightarrow S^{i}F \longrightarrow \operatorname{\mathrm{Ext}}^{i}(\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} , v(F)).
Corollary

Corollary 9.5

KSHH

Suppose F is finitely presented and that \Lambda is left hereditary. Then the first row of the right fundamental sequence becomes

\begin{tikzcd} 0 \ar[r] & \overline{F} \ar[r] & F \ar[r, "\rho_{F}"] & \big(\hspace{0.04cm} \rule{2.4mm}{.4pt} \hspace{0.04cm} , v(F)\big) \ar[r] & 0. \end{tikzcd}

If F is half-exact, this sequence is exact, in fact, split exact.

Proof of Corollary 9.5

Since finitely presented contravariant functors carry coproducts to products, only the last claim needs a proof, but this is immediate since representable functors are projective objects.

Definition

Definition 10.1

HR1A

We shall refer to (?) as the left fundamental sequence of F.

Theorem

Theorem 10.2

ANNN

  1. The left fundamental sequence (?) is a complex, whose homology is zero except possibly, at the terms L^{i}F.

  2. If F is half-exact, then the sequence is exact.

  3. If F carries epimorphisms to monomorphisms, then \alpha_{F} is monic.

  4. If F carries monomorphisms to epimorphisms, then \underline{F} = 0 and \lambda_{F} is epic.

  5. \lambda_{F} is an isomorphism if and only if F is right-exact.

Proof of Theorem 10.2

Similar to the proof of Theorem 6.2.

Corollary

Corollary 10.3

NHTS

Suppose F is a left-exact functor. Then the canonical natural transformation \underline{F} \circ \Sigma^{i+1} \longrightarrow L^{i}F is an isomorphism for each i \geq 1.

Proposition

Proposition 11.1

HWJN

There is a functor isomorphism h_{J} \circ \overline{F} \simeq \overline{h}_{G(J)}.

Proposition

Proposition 11.2

T0DJ

There is a functor isomorphism h^{Q} \circ \underline{G} \simeq \underline{h}^{F(Q)}.

Source and licence

The text on this page is from arXiv:2410.14097v3 by Alex Martsinkovsky, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2410/2410.14097.jsonl.