Loading web-font TeX/Caligraphic/Regular

The Stacks project

13.40 Admissible subcategories

A reference for this section is [Section 1, Bondal-Kapranov].

Definition 13.40.1. Let \mathcal{D} be an additive category. Let \mathcal{A} \subset \mathcal{D} be a full subcategory. The right orthogonal \mathcal{A}^\perp of \mathcal{A} is the full subcategory consisting of the objects X of \mathcal{D} such that \mathop{\mathrm{Hom}}\nolimits (A, X) = 0 for all A \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}). The left orthogonal {}^\perp \mathcal{A} of \mathcal{A} is the full subcategory consisting of the objects X of \mathcal{D} such that \mathop{\mathrm{Hom}}\nolimits (X, A) = 0 for all A \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}).

Lemma 13.40.2. Let \mathcal{D} be a triangulated category. Let \mathcal{A} \subset \mathcal{D} be a full subcategory invariant under all shifts. Consider a distinguished triangle

X \to Y \to Z \to X[1]

of \mathcal{D}. The following are equivalent

  1. Z is in \mathcal{A}^\perp , and

  2. \mathop{\mathrm{Hom}}\nolimits (A, X) = \mathop{\mathrm{Hom}}\nolimits (A, Y) for all A \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}).

Proof. By Lemma 13.4.2 the functor \mathop{\mathrm{Hom}}\nolimits (A, -) is homological and hence we get a long exact sequence as in (13.3.5.1). Assume (1) and let A \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}). Then we consider the exact sequence

\mathop{\mathrm{Hom}}\nolimits (A[1], Z) \to \mathop{\mathrm{Hom}}\nolimits (A, X) \to \mathop{\mathrm{Hom}}\nolimits (A, Y) \to \mathop{\mathrm{Hom}}\nolimits (A, Z)

Since A[1] \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}) we see that the first and last groups are zero. Thus we get (2). Assume (2) and let A \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}). Then we consider the exact sequence

\mathop{\mathrm{Hom}}\nolimits (A, X) \to \mathop{\mathrm{Hom}}\nolimits (A, Y) \to \mathop{\mathrm{Hom}}\nolimits (A, Z) \to \mathop{\mathrm{Hom}}\nolimits (A[-1], X) \to \mathop{\mathrm{Hom}}\nolimits (A[-1], Y)

and we conclude that \mathop{\mathrm{Hom}}\nolimits (A, Z) = 0 as desired. \square

Lemma 13.40.3. Let \mathcal{D} be a triangulated category. Let \mathcal{B} \subset \mathcal{D} be a full subcategory invariant under all shifts. Consider a distinguished triangle

X \to Y \to Z \to X[1]

of \mathcal{D}. The following are equivalent

  1. X is in {}^\perp \mathcal{B}, and

  2. \mathop{\mathrm{Hom}}\nolimits (Y, B) = \mathop{\mathrm{Hom}}\nolimits (Z, B) for all B \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{B}).

Proof. Dual to Lemma 13.40.2. \square

Lemma 13.40.4. Let \mathcal{D} be a triangulated category. Let \mathcal{A} \subset \mathcal{D} be a full subcategory invariant under all shifts. Then both the right orthogonal \mathcal{A}^\perp and the left orthogonal {}^\perp \mathcal{A} of \mathcal{A} are strictly full, saturated1, triangulated subcagories of \mathcal{D}.

Proof. It is immediate from the definitions that the orthogonals are preserved under taking shifts, direct sums, and direct summands. Consider a distinguished triangle

X \to Y \to Z \to X[1]

of \mathcal{D}. By Lemma 13.4.16 it suffices to show that if X and Y are in \mathcal{A}^\perp , then Z is in \mathcal{A}^\perp . This is immediate from Lemma 13.40.2. \square

Lemma 13.40.5. Let \mathcal{D} be a triangulated category. Let \mathcal{A} be a full triangulated subcategory of \mathcal{D}. For an object X of \mathcal{D} consider the property P(X): there exists a distinguished triangle A \to X \to B \to A[1] in \mathcal{D} with A in \mathcal{A} and B in \mathcal{A}^\perp .

  1. If X_1 \to X_2 \to X_3 \to X_1[1] is a distinguished triangle and P holds for two out of three, then it holds for the third.

  2. If P holds for X_1 and X_2, then it holds for X_1 \oplus X_2.

Proof. Let X_1 \to X_2 \to X_3 \to X_1[1] be a distinguished triangle and assume P holds for X_1 and X_2. Choose distinguished triangles

A_1 \to X_1 \to B_1 \to A_1[1] \quad \text{and}\quad A_2 \to X_2 \to B_2 \to A_2[1]

as in condition P. Since \mathop{\mathrm{Hom}}\nolimits (A_1, A_2) = \mathop{\mathrm{Hom}}\nolimits (A_1, X_2) by Lemma 13.40.2 there is a unique morphism A_1 \to A_2 such that the diagram

\xymatrix{ A_1 \ar[d] \ar[r] & X_1 \ar[d] \\ A_2 \ar[r] & X_2 }

commutes. Choose an extension of this to a diagram

\xymatrix{ A_1 \ar[r] \ar[d] & X_1 \ar[r] \ar[d] & Q_1 \ar[r] \ar[d] & A_1[1] \ar[d] \\ A_2 \ar[r] \ar[d] & X_2 \ar[r] \ar[d] & Q_2 \ar[r] \ar[d] & A_2[1] \ar[d] \\ A_3 \ar[r] \ar[d] & X_3 \ar[r] \ar[d] & Q_3 \ar[r] \ar[d] & A_3[1] \ar[d] \\ A_1[1] \ar[r] & X_1[1] \ar[r] & Q_1[1] \ar[r] & A_1[2] }

as in Proposition 13.4.23. By TR3 we see that Q_1 \cong B_1 and Q_2 \cong B_2 and hence Q_1, Q_2 \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}^\perp ). As Q_1 \to Q_2 \to Q_3 \to Q_1[1] is a distinguished triangle we see that Q_3 \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}^\perp ) by Lemma 13.40.4. Since \mathcal{A} is a full triangulated subcategory, we see that A_3 is isomorphic to an object of \mathcal{A}. Thus X_3 satisfies P. The other cases of (1) follow from this case by translation. Part (2) is a special case of (1) via Lemma 13.4.11. \square

Lemma 13.40.6. Let \mathcal{D} be a triangulated category. Let \mathcal{B} be a full triangulated subcategory of \mathcal{D}. For an object X of \mathcal{D} consider the property P(X): there exists a distinguished triangle A \to X \to B \to A[1] in \mathcal{D} with B in \mathcal{B} and A in {}^\perp \mathcal{B}.

  1. If X_1 \to X_2 \to X_3 \to X_1[1] is a distinguished triangle and P holds for two out of three, then it holds for the third.

  2. If P holds for X_1 and X_2, then it holds for X_1 \oplus X_2.

Proof. Dual to Lemma 13.40.5. \square

Lemma 13.40.7. Let \mathcal{D} be a triangulated category. Let \mathcal{A} \subset \mathcal{D} be a full triangulated subcategory. The following are equivalent

  1. the inclusion functor \mathcal{A} \to \mathcal{D} has a right adjoint, and

  2. for every X in \mathcal{D} there exists a distinguished triangle

    A \to X \to B \to A[1]

    in \mathcal{D} with A \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}) and B \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}^\perp ).

If this holds, then \mathcal{A} is saturated (Definition 13.6.1) and if \mathcal{A} is strictly full in \mathcal{D}, then \mathcal{A} = {}^\perp (\mathcal{A}^\perp ).

Proof. Assume (1) and denote v : \mathcal{D} \to \mathcal{A} the right adjoint. Let X \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{D}). Set A = v(X). We may extend the adjunction mapping A \to X to a distinguished triangle A \to X \to B \to A[1]. Since

\mathop{\mathrm{Hom}}\nolimits _\mathcal {A}(A', A) = \mathop{\mathrm{Hom}}\nolimits _\mathcal {A}(A', v(X)) = \mathop{\mathrm{Hom}}\nolimits _\mathcal {D}(A', X)

for A' \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}), we conclude that B \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{A}^\perp ) by Lemma 13.40.2.

Assume (2). We will construct the adjoint v explicitly. Let X \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{D}). Choose A \to X \to B \to A[1] as in (2). Set v(X) = A. Let f : X \to Y be a morphism in \mathcal{D}. Choose A' \to Y \to B' \to A'[1] as in (2). Since \mathop{\mathrm{Hom}}\nolimits (A, A') = \mathop{\mathrm{Hom}}\nolimits (A, Y) by Lemma 13.40.2 there is a unique morphism f' : A \to A' such that the diagram

\xymatrix{ A \ar[d]_{f'} \ar[r] & X \ar[d]^ f \\ A' \ar[r] & Y }

commutes. Hence we can set v(f) = f' to get a functor. To see that v is adjoint to the inclusion morphism use Lemma 13.40.2 again.

Proof of the final statement. In order to prove that \mathcal{A} is saturated we may replace \mathcal{A} by the strictly full subcategory having the same isomorphism classes as \mathcal{A}; details omitted. Assume \mathcal{A} is strictly full. If we show that \mathcal{A} = {}^\perp (\mathcal{A}^\perp ), then \mathcal{A} will be saturated by Lemma 13.40.4. Since the incusion \mathcal{A} \subset {}^\perp (\mathcal{A}^\perp ) is clear it suffices to prove the other inclusion. Let X be an object of {}^\perp (\mathcal{A}^\perp ). Choose a distinguished triangle A \to X \to B \to A[1] as in (2). As \mathop{\mathrm{Hom}}\nolimits (X, B) = 0 by assumption we see that A \cong X \oplus B[-1] by Lemma 13.4.11. Since \mathop{\mathrm{Hom}}\nolimits (A, B[-1]) = 0 as B \in \mathcal{A}^\perp this implies B[-1] = 0 and A \cong X as desired. \square

Lemma 13.40.8. Let \mathcal{D} be a triangulated category. Let \mathcal{B} \subset \mathcal{D} be a full triangulated subcategory. The following are equivalent

  1. the inclusion functor \mathcal{B} \to \mathcal{D} has a left adjoint, and

  2. for every X in \mathcal{D} there exists a distinguished triangle

    A \to X \to B \to A[1]

    in \mathcal{D} with B \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{B}) and A \in \mathop{\mathrm{Ob}}\nolimits ({}^\perp \mathcal{B}).

If this holds, then \mathcal{B} is saturated (Definition 13.6.1) and if \mathcal{B} is strictly full in \mathcal{D}, then \mathcal{B} = ({}^\perp \mathcal{B})^\perp .

Proof. Dual to Lemma 13.40.7. \square

Definition 13.40.9. Let \mathcal{D} be a triangulated category. A right admissible subcategory of \mathcal{D} is a strictly full triangulated subcategory satisfying the equivalent conditions of Lemma 13.40.7. A left admissible subcategory of \mathcal{D} is a strictly full triangulated subcategory satisfying the equivalent conditions of Lemma 13.40.8. A two-sided admissible subcategory is one which is both right and left admissible.

Let \mathcal{A} be a right admissible subcategory of the triangulated category \mathcal{D}. Then we observe that for X \in \mathcal{D} the distinguished triangle

A \to X \to B \to A[1]

with A \in \mathcal{A} and B \in \mathcal{A}^\perp is canonical in the following sense: for any other distinguished triangle A' \to X \to B' \to A'[1] with A' \in \mathcal{A} and B' \in \mathcal{A}^\perp there is an isomorphism (\alpha , \text{id}_ X, \beta ) : (A, X, B) \to (A', X, B') of triangles. The following proposition summarizes what was said above.

Proposition 13.40.10. Let \mathcal{D} be a triangulated category. Let \mathcal{A} \subset \mathcal{D} and \mathcal{B} \subset \mathcal{D} be subcategories. The following are equivalent

  1. \mathcal{A} is right admissible and \mathcal{B} = \mathcal{A}^\perp ,

  2. \mathcal{B} is left admissible and \mathcal{A} = {}^\perp \mathcal{B},

  3. \mathop{\mathrm{Hom}}\nolimits (A, B) = 0 for all A \in \mathcal{A} and B \in \mathcal{B} and for every X in \mathcal{D} there exists a distinguished triangle A \to X \to B \to A[1] in \mathcal{D} with A \in \mathcal{A} and B \in \mathcal{B}.

If this is true, then \mathcal{A} \to \mathcal{D}/\mathcal{B} and \mathcal{B} \to \mathcal{D}/\mathcal{A} are equivalences of triangulated categories, the right adjoint to the inclusion functor \mathcal{A} \to \mathcal{D} is \mathcal{D} \to \mathcal{D}/\mathcal{B} \to \mathcal{A}, and the left adjoint to the inclusion functor \mathcal{B} \to \mathcal{D} is \mathcal{D} \to \mathcal{D}/\mathcal{A} \to \mathcal{B}.

Proof. The equivalence between (1), (2), and (3) follows in a straightforward manner from Lemmas 13.40.7 and 13.40.8 (small detail omitted). Denote v : \mathcal{D} \to \mathcal{A} the right adjoint of the inclusion functor i : \mathcal{A} \to \mathcal{D}. It is immediate that \mathop{\mathrm{Ker}}(v) = \mathcal{A}^\perp = \mathcal{B}. Thus v factors over a functor \overline{v} : \mathcal{D}/\mathcal{B} \to \mathcal{A} by the universal property of the quotient. Since v \circ i = \text{id}_\mathcal {A} by Categories, Lemma 4.24.4 we see that \overline{v} is a left quasi-inverse to \overline{i} : \mathcal{A} \to \mathcal{D}/\mathcal{B}. We claim also the composition \overline{i} \circ \overline{v} is isomorphic to \text{id}_{\mathcal{D}/\mathcal{B}}. Namely, suppose we have X fitting into a distinguished triangle A \to X \to B \to A[1] as in (3). Then v(X) = A as was seen in the proof of Lemma 13.40.7. Viewing X as an object of \mathcal{D}/\mathcal{B} we have \overline{i}(\overline{v}(X)) = A and there is a functorial isomorphism \overline{i}(\overline{v}(X)) = A \to X in \mathcal{D}/\mathcal{B}. Thus we find that indeed \overline{v} : \mathcal{D}/\mathcal{B} \to \mathcal{A} is an equivalence. To show that \mathcal{B} \to \mathcal{D}/\mathcal{A} is an equivalence and the left adjoint to the inclusion functor \mathcal{B} \to \mathcal{D} is \mathcal{D} \to \mathcal{D}/\mathcal{A} \to \mathcal{B} is dual to what we just said. \square

[1] Definition 13.6.1.

Comments (3)

Comment #6021 by Sandeep on

Two typos: In the statements of Lemmas 13.39.5 and 13.39.6, the distinguished triangles should be and respectively.

Comment #6022 by Sandeep on

Sorry, I mean .


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.