Spherical generation · the thick span

Three spans of a set of objects

The thick span, the draft's span after Gaitsgory–Rozenblyum, and its double orthogonal after Mezer. The first two agree once the second is taken in the ind-completion; skyscrapers on \(\mathbb A^1\) show that the third can be strictly larger.

  1. DefinitionThree spans
  2. LemmaHow they compare
  3. PropositionSkyscrapers on the line
  4. LemmaThe missing test object
  5. ConsequenceThe Satake subcategory

1 · Three spans

Fix a small, idempotent-complete, stable \(\infty\)-category \(\mathcal D\) over \(k=\mathbb C\) and a set \(\mathcal S\) of its objects. The ind-completion is compactly generated, and its compact objects are exactly \(\mathcal D\) [HTT, Prop. 5.3.5.5, Lemma 5.4.2.4]:

$$\widehat{\mathcal D}=\operatorname{Ind}\mathcal D,\qquad\widehat{\mathcal D}^{\,\omega}=\mathcal D.$$

\(\Map\) is the derived mapping complex of the mapping convention: it vanishes iff \(\operatorname{Hom}(A,B[n])=0\) for every \(n\in\mathbb Z\).

(thick span) Definition

\(\operatorname{Thick}_{\mathcal D}(\mathcal S)\) is the smallest full subcategory of \(\mathcal D\) that contains \(0\) and \(\mathcal S\) and is closed under isomorphisms, shifts, cofibers and retracts [Stacks, Tag 05RB]. Dropping retracts gives the stable span \(\operatorname{Stab}_{\mathcal D}(\mathcal S)\).

(GR span) Definition

For a stable \(\infty\)-category \(\mathcal E\) with all small colimits, \(\langle\mathcal S\rangle^{\mathrm{GR}}_{\mathcal E}\) is the smallest full stable subcategory of \(\mathcal E\) that contains \(\mathcal S\) and is closed under small colimits. Gaitsgory–Rozenblyum define generation by vanishing of \(\operatorname{Maps}(s[-i],-)\), \(i\ge0\), and prove it equivalent to \(\langle\mathcal S\rangle^{\mathrm{GR}}_{\mathcal E}=\mathcal E\) for presentable \(\mathcal E\) [GR, Vol. I, Ch. 1, §5.4.1, Prop. 5.4.5].

(double orthogonal) [Mezer, Def. 5.1] Definition

$$\mathcal S^{\perp_{\mathcal D}}=\{M\in\mathcal D:\Map_{\mathcal D}(s,M)=0\ \ \forall s\in\mathcal S\},$$ $$\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D}={}^{\perp_{\mathcal D}}\big(\mathcal S^{\perp_{\mathcal D}}\big)=\{X\in\mathcal D:\Map_{\mathcal D}(X,M)=0\ \ \forall M\in\mathcal S^{\perp_{\mathcal D}}\}.$$
Table 1. Span and In the draft and Used there for
SpanIn the draftUsed there for
Thickproposed in Remark 3.29 as \(\mathcal D\cap\langle\mathcal S\rangle^{\mathrm{GR}}_{\widehat{\mathcal D}}\)—
GRDef. 3.7, after Gaitsgory–Rozenblyumthe Satake subcategory (Def. 3.28)
MDef. 3.32, after Mezerthe proof of Sat = Full (Prop. 3.47)

2 · How the spans compare

(comparison chain) Proved; (1) cited

$$\begin{aligned}\operatorname{Thick}_{\mathcal D}(\mathcal S)&\overset{(1)}{=}\mathcal D\cap\langle\mathcal S\rangle^{\mathrm{GR}}_{\widehat{\mathcal D}}\\&\overset{(2)}{=}\mathcal D\cap{}^{\perp_{\widehat{\mathcal D}}}\big(\mathcal S^{\perp_{\widehat{\mathcal D}}}\big)\\&\overset{(3)}{\subseteq}\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D}.\end{aligned}$$
Proof
  1. (1) Write \(\mathcal L=\langle\mathcal S\rangle^{\mathrm{GR}}_{\widehat{\mathcal D}}\). For the compact objects \(s[n]\) (\(s\in\mathcal S\), \(n\in\mathbb Z\)), Neeman's lemma says the objects compact in \(\mathcal L\) form their thick span [Neeman 1992, Lemma 2.2]. Coproducts in \(\mathcal L\) are computed in \(\widehat{\mathcal D}\), so these are exactly the objects of \(\mathcal D\cap\mathcal L\).
  2. (2) Let \(\iota:\mathcal L\hookrightarrow\widehat{\mathcal D}\). The left orthogonal \({}^{\perp}(\mathcal S^{\perp})\) contains \(\mathcal S\) and is closed under colimits, so it contains \(\mathcal L\). Conversely, \(\mathcal L\) is the smallest full subcategory containing every \(s[n]\) and closed under extensions and small colimits (that subcategory is closed under \([-1]\), hence stable), so it is presentable [HA, Prop. 1.4.4.11(2)] and \(\iota\) has a right adjoint \(R\) [HTT, Cor. 5.5.2.9]. The counit of \(X\in\widehat{\mathcal D}\) has cofiber \(C\in\mathcal S^{\perp}\): $$\iota RX\longrightarrow X\longrightarrow C,\qquad\Map(s,\iota RX)\xrightarrow{\ \sim\ }\Map(s,X)\quad(s\in\mathcal S).$$ If \(X\in{}^{\perp}(\mathcal S^{\perp})\), then \(\Map(X,C)=0=\Map(\iota RX,C)\), so \(\Map(C,C)=0\). Hence \(C=0\) and \(X\simeq\iota RX\in\mathcal L\).
  3. (3) \(\mathcal S^{\perp_{\mathcal D}}=\mathcal S^{\perp_{\widehat{\mathcal D}}}\cap\mathcal D\) has fewer test objects, so its left orthogonal in \(\mathcal D\) is larger.

(GR on a small category) Proved here

A small stable \(\infty\)-category with all small colimits is zero. So if \(\mathcal S\not\subseteq\{0\}\), no full stable subcategory of the small \(\mathcal D\) with all small colimits contains \(\mathcal S\), and \(\langle\mathcal S\rangle^{\mathrm{GR}}_{\mathcal D}\) is undefined.

Proof
  1. Let \(\mathcal A\) be such a category, \(X\in\mathcal A\), and \(I\) a set larger than the set of morphisms of the homotopy category \(h\mathcal A\). Coproducts in \(\mathcal A\) are coproducts in \(h\mathcal A\): $$\operatorname{Hom}_{h\mathcal A}\Big(\textstyle\bigoplus_I X,\,X\Big)\cong\operatorname{End}_{h\mathcal A}(X)^I.$$
  2. If \(\operatorname{End}(X)\) had two elements, the right side would have at least \(2^{|I|}\gt|I|\) elements, more than \(h\mathcal A\) has morphisms. So \(\operatorname{End}(X)=\{0\}\), \(\mathrm{id}_X=0\) and \(X\simeq0\). This is the dual of Freyd's argument that a small complete category is a preorder [Freyd, Ch. 3, Exercise D].

Reading "cocomplete" as "closed under the finite colimits of \(\mathcal D\)" gives \(\operatorname{Stab}_{\mathcal D}(\mathcal S)\) instead, which differs from the thick span by \(K_0\) [Thomason 1997, Thm. 2.1]:

$$\operatorname{Stab}_{\mathcal D}(\mathcal S)=\mathcal D\iff\operatorname{Thick}_{\mathcal D}(\mathcal S)=\mathcal D\ \text{ and }\ \langle[s]:s\in\mathcal S\rangle=K_0(\mathcal D).$$

Proposition I proves only the first condition for \(\mathcal C\): its step 6 extracts \(\Delta_\mu\) from \(\bigoplus_s\Delta_\mu[d_s]\) as a retract.

3 · Skyscrapers on the line

Can inclusion (3) of be strict? Yes, already for points on \(\mathbb A^1\).

(constructible complexes on the line) Definition

\(D^b_c(\mathbb A^1)\) is the stable \(\infty\)-category of bounded complexes \(M\) of sheaves of \(k\)-vector spaces on \(\mathbb C=\mathbb A^1(\mathbb C)\) for which some finite \(\Sigma_M\subset\mathbb C\) has two properties: on \(\mathbb C\setminus\Sigma_M\) every cohomology sheaf \(\mathcal H^nM\) is a local system of finite rank (\(M\) is lisse there), and at every \(b\in\Sigma_M\) every stalk \((\mathcal H^nM)_b\) is finite-dimensional. For \(a\in\mathbb C\), with \(i_a:\{a\}\hookrightarrow\mathbb A^1\):

$$\delta_a=i_{a*}k,\qquad M_a=i_a^*M\in D^b(k),\qquad\operatorname{supp}M=\overline{\{x\in\mathbb C:M_x\ne0\}}.$$

(skyscrapers) Proved here

Let \(\mathcal D=D^b_c(\mathbb A^1)\) and \(\mathcal S=\{\delta_a:a\in\mathbb C\}\).

  1. The skyscrapers detect zero, so the double orthogonal is everything: $$\mathcal S^{\perp_{\mathcal D}}=0,\qquad\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D}=\mathcal D.$$
  2. The thick span consists of the objects with finite support, so it misses the constant sheaf: $$\operatorname{Thick}_{\mathcal D}(\mathcal S)=\{M\in\mathcal D:\operatorname{supp}M\ \text{finite}\},\qquad\underline k_{\mathbb A^1}\notin\operatorname{Thick}_{\mathcal D}(\mathcal S).$$

Hence inclusion (3) is strict:

$$\operatorname{Thick}_{\mathcal D}(\mathcal S)\subsetneq\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D}=\mathcal D.$$

Why the constant sheaf is missing. Stalks are exact and keep retracts, so no operation of the thick closure enlarges supports beyond the union of the inputs' supports. A skyscraper has one point of support; the constant sheaf has stalk \(k\) everywhere:

$$\operatorname{supp}\delta_a=\{a\},\qquad\operatorname{supp}\underline k_{\mathbb A^1}=\mathbb C.$$
ℂa(a) \(\delta_a\)
ℂabc(b) \(\delta_a\oplus\delta_b[3]\oplus\delta_c^{\oplus2}\)
ℂ(c) \(\underline k_{\mathbb A^1}\)
Figure 1. Supports, shaded: objects built from skyscrapers, as in (a) and (b), have finitely many points of support, while (c) has all of \(\mathbb C\).

(costalk at a lisse point) Proved here

If \(M\in D^b_c(\mathbb A^1)\) and \(a\notin\Sigma_M\), then

$$i_a^!M\simeq M_a[-2].$$
Proof
  1. Choose an open disc \(B\ni a\) with \(B\cap\Sigma_M=\varnothing\), and let \(p:B\to\mathrm{pt}\). The counit is an isomorphism, so \(M|_B\) is constant: $$p^*p_*(M|_B)\xrightarrow{\ \sim\ }M|_B,\qquad M|_B\simeq p^*M_a.$$ Both sides are exact in \(M\), so by induction on the number of nonzero \(\mathcal H^nM\) it suffices to treat one local system \(L\). On the contractible \(B\), \(L\) is constant and \(H^{\gt0}(B,L)=0\), so \(p_*L=L_a\) and the counit is an isomorphism.
  2. \(i_a^!\) depends only on \(M|_B\), and \(M_a\simeq\bigoplus_nH^n(M_a)[-n]\) is a finite sum, so it suffices to compute \(i_a^!\underline k_B\). It is local cohomology at \(a\); since \(B\simeq\mathrm{pt}\), \(B\setminus a\simeq S^1\) and restriction is an isomorphism on \(H^0\): $$i_a^!\underline k_B=\operatorname{fib}\big(R\Gamma(B,k)\to R\Gamma(B\setminus a,k)\big)=\operatorname{fib}\big(k\to k\oplus k[-1]\big)=k[-2].$$
Proof

Proof of the Proposition

  1. Probes are costalks. \(i_{a*}\) is left adjoint to \(i_a^!\), so $$\Map_{\mathcal D}(\delta_a,M)\simeq\Map_{D(k)}(k,i_a^!M)\simeq i_a^!M.$$
  2. Finite supports split. Let \(\Sigma\subset\mathbb C\) be finite with \(M_x=0\) for \(x\notin\Sigma\). Then \(M\) vanishes on \(\mathbb C\setminus\Sigma\), and the localization triangle and discreteness of \(\Sigma\) give $$M\simeq\bigoplus_{b\in\Sigma}i_{b*}M_b,\qquad i_a^!\,i_{b*}V\simeq\begin{cases}V,&a=b,\\0,&a\ne b.\end{cases}$$
  3. Part (a). Let \(M\in\mathcal S^{\perp_{\mathcal D}}\), so \(i_a^!M=0\) for every \(a\) by step 1. gives \(M_a=0\) for \(a\notin\Sigma_M\). Step 2 with \(\Sigma=\Sigma_M\) then gives \(M_a\simeq i_a^!M=0\) for \(a\in\Sigma_M\). So \(M=0\), and $$\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D}={}^{\perp_{\mathcal D}}0=\mathcal D.$$
  4. Supports are thick. The stalk functor \(M\mapsto M_x=i_x^*M\) is exact and carries retracts to retracts. So for a cofiber sequence \(A\to B\to C\) and a retract \(M\) of \(N\): $$\operatorname{supp}M[1]=\operatorname{supp}M,\qquad\operatorname{supp}C\subseteq\operatorname{supp}A\cup\operatorname{supp}B,\qquad\operatorname{supp}M\subseteq\operatorname{supp}N.$$ The objects of finite support therefore form a thick subcategory containing \(\mathcal S\): $$\operatorname{Thick}_{\mathcal D}(\mathcal S)\subseteq\{M\in\mathcal D:\operatorname{supp}M\ \text{finite}\}.$$
  5. Part (b). Conversely, if \(\operatorname{supp}M\) is finite, step 2 writes \(M\) as a finite sum of objects \(i_{b*}M_b\simeq\bigoplus_n\delta_b^{\oplus\dim H^n(M_b)}[-n]\), and a finite sum is the cofiber of a zero map. Finally \(\underline k_x=k\ne0\) for every \(x\): $$\operatorname{supp}\underline k_{\mathbb A^1}=\mathbb C\ \text{ is infinite},\qquad\underline k_{\mathbb A^1}\notin\operatorname{Thick}_{\mathcal D}(\mathcal S).$$

4 · The test object the small category misses

By (2) of , \(\underline k_{\mathbb A^1}\notin{}^{\perp}(\mathcal S^{\perp_{\widehat{\mathcal D}}})\), so some nonzero ind-object is orthogonal to every skyscraper. It punctures the line everywhere. For a finite \(T\subset\mathbb C\) write

$$U_T=\mathbb A^1\setminus T,\qquad j_T:U_T\hookrightarrow\mathbb A^1.$$

(the missing test object) Proved here

Let \(T\) run over the finite subsets of \(\mathbb C\), ordered by inclusion, with the restriction maps \(j_{T*}\underline k_{U_T}\to j_{T'*}\underline k_{U_{T'}}\) for \(T\subseteq T'\). In \(\widehat{\mathcal D}=\operatorname{Ind}D^b_c(\mathbb A^1)\),

$$F=\operatorname*{colim}_T\,j_{T*}\underline k_{U_T}:\qquad\Map_{\widehat{\mathcal D}}(\delta_a,F)=0\ \ (a\in\mathbb C),\qquad H^0\Map_{\widehat{\mathcal D}}(\underline k_{\mathbb A^1},F)=k.$$
Proof
  1. Objects of \(D^b_c(\mathbb A^1)\) are compact in \(\widehat{\mathcal D}\), so \(\Map(\delta_a,-)\) and \(\Map(\underline k_{\mathbb A^1},-)\) commute with the filtered colimit.
  2. The sets \(T\ni a\) are cofinal, and for them \(j_T^*\delta_a=0\): $$\Map(\delta_a,j_{T*}\underline k_{U_T})\simeq\Map(j_T^*\delta_a,\underline k_{U_T})=0.$$
  3. \(U_T\) is the plane minus \(|T|\) points, and restriction is an isomorphism on \(H^0\): $$\Map(\underline k_{\mathbb A^1},j_{T*}\underline k_{U_T})\simeq R\Gamma(U_T,k)\simeq k\oplus k^{|T|}[-1].$$
point of \(U_{T_n}\)puncture, in \(T_n\)probe point \(a\); click a point to move it

Figure 2. Every skyscraper probe dies once its point is punctured, while the constant sheaf's probe keeps \(H^0=k\) at every stage; so the colimit \(F\) is nonzero and orthogonal to all of \(\mathcal S\).

(the double orthogonal depends on the ambient category) Proved here

Let \(\mathcal D'=\operatorname{Thick}_{\widehat{\mathcal D}}\big(D^b_c(\mathbb A^1)\cup\{F\}\big)\), a small stable category containing \(\mathcal D=D^b_c(\mathbb A^1)\). Then

$$\underline k_{\mathbb A^1}\in\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D},\qquad\underline k_{\mathbb A^1}\notin\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D'},$$

so the exact inclusion \(\mathcal D\hookrightarrow\mathcal D'\) does not carry \(\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D}\) into \(\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D'}\): (T2) fails for the double orthogonal.

Proof

The first is (a). By , \(F\in\mathcal S^{\perp_{\mathcal D'}}\) and \(\Map(\underline k_{\mathbb A^1},F)\ne0\).

5 · What this means for the Satake subcategory

Take \(\mathcal D=\mathcal C\) and \(\mathcal S=\{V_\lambda\ast\mathrm{IC}_0:\lambda\in\Lambda_T^+\}\), so that \(\operatorname{Thick}_{\mathcal C}(\mathcal S)=\mathcal C^{\mathrm{Sat}}\).

Table 2. Span and On the small \(\mathcal C\) and Compared with \(\mathcal C^{\mathrm{Sat}}\)
SpanOn the small \(\mathcal C\)Compared with \(\mathcal C^{\mathrm{Sat}}\)
Thickdefinedequal
GRundefined ()equal after \(\mathcal C\cap\langle\mathcal S\rangle^{\mathrm{GR}}_{\widehat{\mathcal C}}\) ()
Mdefinedcontains it, strictly in general ()

6 · Questions

Predict. Is \(j_!\underline k_{\mathbb A^1\setminus0}\), the constant sheaf off the origin extended by zero, in \(\operatorname{Thick}_{\mathcal D}(\mathcal S)\)?

Explain. The generic Euler characteristic \(\chi(M_x)\), \(x\notin\Sigma_M\), is additive on cofiber sequences, vanishes on every \(\delta_a\), and equals \(1\) on \(\underline k_{\mathbb A^1}\). What does it prove?

Transfer. An argument shows \(\Map_{\mathcal C}(\Delta_\mu,F)=0\) for every \(\mu\) and every \(F\in\mathcal C\) orthogonal to all Satake objects. The standards thickly generate \(\mathcal C\). What follows?

Where the three spans come from
Table 3. Year and Advance and Source
YearAdvanceSource
1996Verdier's thesis defines saturée triangulated subcategories: those containing every direct summand of their objects, the thick spans of .[Verdier, Ch. II, 2.1.6]
1992The objects compact in the localizing subcategory generated by a set of compact objects form its thick span: equality (1) of .[Neeman, Lemma 2.2]
1997Dense triangulated subcategories correspond to subgroups of \(K_0\): the gap between the stable and the thick span.[Thomason, Thm. 2.1]
1998Thick subcategories of finite spectra are classified; thick spans become the objects of classification theorems.[Hopkins–Smith, Def. 6, Thm. 7]
2003"Classically generates" (thick span is everything) is separated from "generates" (\(E^{\perp}=0\)); for compact objects of a compactly generated category the two agree after passing to compact objects.[Bondal–Van den Bergh, §2.1, Thm. 2.1.2]
2017Generation of a cocomplete stable category, by a Map criterion and by the absence of proper cocomplete subcategories: the GR span.[GR, Vol. I, Ch. 1, §5.4]
2026Generation in a stable category that need not be cocomplete, as the double orthogonal: the span (M).[Mezer, Def. 5.1]
Notation
Table 4. Symbol and Meaning and Defined in
SymbolMeaningDefined in
\(\widehat{\mathcal D}=\operatorname{Ind}\mathcal D\)ind-completion; its compact objects are \(\mathcal D\)§1
\(\operatorname{Thick}_{\mathcal D}(\mathcal S)\), \(\operatorname{Stab}_{\mathcal D}(\mathcal S)\)thick span; stable span (no retracts)
\(\langle\mathcal S\rangle^{\mathrm{GR}}_{\mathcal E}\)smallest colimit-closed stable subcategory containing \(\mathcal S\)
\(\mathcal S^{\perp_{\mathcal D}}\), \(\langle\mathcal S\rangle^{\mathrm M}_{\mathcal D}\)right orthogonal; double orthogonal
\(\Sigma_M\)finite set off which \(M\) is lisse
\(i_a\), \(\delta_a\), \(M_a\)inclusion of the point \(a\); skyscraper \(i_{a*}k\); stalk \(i_a^*M\)
\(\operatorname{supp}M\)closure of \(\{x:M_x\ne0\}\)
\(U_T\), \(j_T\), \(F\)line minus a finite \(T\); its inclusion; the colimit over all \(T\)
Sources

References

  1. The project draft, sources/main-draft.pdf: Def. 3.7 (p. 8), Def. 3.8 (p. 9), Def. 3.28, Remark 3.29 and Conjecture 3.30 (p. 15), Def. 3.32 (p. 17), Prop. 3.47 (p. 22). Used in §1 and §5.
  2. A. Bondal, M. Van den Bergh, Generators and representability of functors in commutative and noncommutative geometry, Mosc. Math. J. 3 (2003), no. 1, 1–36. doi:10.17323/1609-4514-2003-3-1-1-36; open copy arXiv:math/0204218v2. Used: §2.1 ("classically generates", "generates"), Thm. 2.1.2. Numbers checked in arXiv v2. History.
  3. P. Freyd, Abelian Categories, Harper & Row, 1964; reprinted in Repr. Theory Appl. Categ. 3 (2003). Open copy TAC reprint. Used: Ch. 3, Exercise D, p. 78 ("Small complete categories are lattices"). .
  4. D. Gaitsgory, N. Rozenblyum, A Study in Derived Algebraic Geometry, Vol. I: Correspondences and Duality, Math. Surveys Monogr. 221, AMS, 2017. doi:10.1090/surv/221.1; authors' copy Vol1.pdf. Used: Part I, Ch. 1, §5.4.1 (definition of generation) and Prop. 5.4.5 (equivalence with the subcategory condition); cocomplete means presentable there (§5.1.5, footnote 7). Numbers checked in the authors' Dec 2021 PDF, whose pagination matches print. .
  5. J. Lurie, Higher Algebra, manuscript, version of 18 Sep 2017, HA.pdf. Used: Prop. 1.4.4.11(2), p. 160 (the closure of a set under extensions and small colimits in a presentable stable category is presentable). .
  6. M. J. Hopkins, J. H. Smith, Nilpotence and stable homotopy theory II, Ann. of Math. (2) 148 (1998), no. 1, 1–49. doi:10.2307/120991. Used: Def. 6 (thick subcategory), Thm. 7. History.
  7. J. Lurie, Higher Topos Theory, Ann. of Math. Stud. 170, Princeton Univ. Press, 2009. doi:10.1515/9781400830558; author's copy HTT.pdf. Used: Prop. 5.3.5.5 and Lemma 5.4.2.4 (compact objects of \(\operatorname{Ind}\mathcal D\)), Cor. 5.5.2.9 (adjoint functor theorem). Numbers checked in the April 2017 PDF. §1, .
  8. D. Mezer, Towards the relative Langlands duality for orthosymplectic pairs, arXiv:2606.03187v1 (2026). Used: Def. 5.1, p. 23 (the subcategory generated by a set, as a double orthogonal, with derived Hom). , §5.
  9. A. Neeman, The connection between the \(K\)-theory localization theorem of Thomason, Trobaugh and Yao and the smashing subcategories of Bousfield and Ravenel, Ann. Sci. École Norm. Sup. (4) 25 (1992), no. 5, 547–566. doi:10.24033/asens.1659; open copy Numdam. Used: Lemma 2.2, p. 555 (stated inside the proof of Thm. 2.1; the generating set is closed under suspension). .
  10. The Stacks Project, Tag 05RB (Definition 13.6.1: saturated, or thick, triangulated subcategory). .
  11. R. W. Thomason, The classification of triangulated subcategories, Compositio Math. 105 (1997), no. 1, 1–27. doi:10.1023/A:1017932514274. Used: Thm. 2.1, p. 5, with "dense" from 1.4. §2.
  12. J.-L. Verdier, Des catégories dérivées des catégories abéliennes, ed. G. Maltsiniotis, Astérisque 239 (1996). Open copy Numdam. Used: Ch. II, 2.1.6, p. 114 (saturée). History.

Every identifier above was resolved, and every cited number was read in the version named, on 2026-10-08 (Crossref, zbMATH, Numdam, arXiv).