Gluing weak approximability along recollements

First, let \mathcal{D} be weakly approximable. Denote by \mathcal{D}^{\leq 0} the aisle compactly generated by G^\prime. Note that G^\prime\in \mathcal{D}^{\leq 0}. By [Burke/Neeman/Pauwels:2023, Lemma 3.7], there exists an integer A^\prime>0 such that (G^\prime,\mathcal{D}^{\leq 0},A^\prime) satisfies Definition 2.1darkblue(?). It follows that \operatorname{Hom}(G^\prime,\mathcal{D}^{\leq -A^\prime})=0. For n\geq A^\prime, G^\prime[n]\in \mathcal{D}^{\leq -n}\subseteq \mathcal{D}^{\leq -A^\prime}. Hence, for all n\geq A^\prime, \operatorname{Hom}(G^\prime,G^\prime[n])=0. This argument applies to every compact generator of \mathcal{D}.

Conversely, let \operatorname{Hom}(G^\prime,G^\prime[n])=0 for n\gg 0. Hence, we are in the setting of Setup 4.4, and we use its notation. By Lemma 4.3darkblue(?), G is a compact generator of \mathcal{D}. We show that (G,\mathcal{D}^{\leq 0},3A) is weak approximation data for \mathcal{D}.

We start with Definition 2.1darkblue(?). Since G\in \mathcal{D}^{\leq 0}, G\in \mathcal{D}^{\leq 3A}. As \mathcal{D}^{\leq -3A}\subseteq \mathcal{D}^{\leq -A}, Setup 4.4darkblue(?) implies \operatorname{Hom}(G,\mathcal{D}^{\leq -3A})=0.

Now we verify Definition 2.1darkblue(?). Fix X\in \mathcal{D}^{\leq 0}. We will approximate X in two stages: first its image i^\ast X in \mathcal{D}_Z, and then what remains in \mathcal{D}_U.

We start on \mathcal{D}_Z. By Lemma 4.2, i^\ast X\in \mathcal{D}_Z^{\leq 0}. By Setup 4.4darkblue(?), there exists a distinguished triangle

where E^\prime\in \overline{\langle i^\ast G\rangle}^{[-A,A]} and D^\prime\in \mathcal{D}_Z^{\leq -1}. By Proposition 4.8, E^\prime\in \mathcal{C}. Since X\in \mathcal{D}^{\leq 0}\subseteq \mathcal{D}^{\leq K}, f^\prime has a \mathcal{P}-lift (q_1\colon E_1\to X,\theta). Complete to a distinguished triangle

Since E_1\in \mathcal{P}\subseteq \mathcal{D}^{\leq K}, E_1[1]\in \mathcal{D}^{\leq K-1}\subseteq \mathcal{D}^{\leq K}. As aisles are closed under extensions, N\in X\star E_1[1]\subseteq \mathcal{D}^{\leq K}. Applying i^\ast to the distinguished triangle on q_1 gives the top row of the diagram

whose bottom row is the distinguished triangle on f^\prime. The left square commutes because i^\ast q_1 = f^\prime\circ \theta. By TR3, there exists \mu\colon i^\ast N\to D^\prime making the diagram commute. Since \theta and the identity are isomorphisms, so is \mu. By definition, aisles are strictly full, and so i^\ast N\in \mathcal{D}_Z^{\leq -1}.

Next we treat the case on \mathcal{D}_U. By Lemma 4.2, j^\ast N\in \mathcal{D}_U^{\leq K}, and so (j^\ast N)[K]\in \mathcal{D}_U^{\leq 0}. Apply Lemma 3.4 to the weak approximation data (G_U,\mathcal{D}_U^{\leq 0},A) of Setup 4.4darkblue(?) and the object (j^\ast N)[K], with K+1 in place of m. This gives a distinguished triangle

where V^\prime\in \overline{\langle G_U\rangle}^{[-K-A,A]} and R^\prime\in \mathcal{D}_U^{\leq -K-1}. Define V := V^\prime[-K] and R := R^\prime[-K]. Shifting by [-K] yields a distinguished triangle

By Lemma 3.1darkblue(?),

Moreover, R\in \mathcal{D}_U^{\leq -K-1}[-K] = \mathcal{D}_U^{\leq -1}.

Consider the composite

Complete to a distinguished triangle

We claim that D^{\prime\prime}\in \mathcal{D}^{\leq -1}. Since j_! is exact, applying it to the distinguished triangle on c gives a distinguished triangle

Hence, \operatorname{cone}(j_!(c)) \cong j_! R. By darkblue(4.1), \operatorname{cone}(\epsilon_N)\cong i_\ast i^\ast N. Applying Remark 3.2darkblue(?) to the composite above gives a distinguished triangle

By Lemma 4.2, R\in \mathcal{D}_U^{\leq -1} implies j_! R\in \mathcal{D}^{\leq -1}. Similarly, i^\ast N\in \mathcal{D}_Z^{\leq -1} implies i_\ast i^\ast N \in \mathcal{D}^{\leq -1}. As aisles are closed under extensions, D^{\prime\prime}\in j_! R\star i_\ast i^\ast N \subseteq \mathcal{D}^{\leq -1}.

We now assemble the two stages. Consider the composite

Complete to a distinguished triangle

The cocones of r, \rho, and \rho\circ r are respectively E_1, j_! V, and E. Hence, Remark 3.2darkblue(?) gives a distinguished triangle E_1\to E\to j_! V\to E_1[1]. Note that E\in E_1\star j_! V.

Lastly we bound the range of E. Since E_1\in \mathcal{P} and B = 3A, E_1\in \overline{\langle G\rangle}^{[-3A,3A]}. Note that j_! is exact and preserves small coproducts. By Setup 4.4, j_! G_U\in \overline{\langle G\rangle}^{[0,0]}. By Lemma 3.1darkblue(?), j_! V\in \overline{\langle G\rangle}^{[-A,2A+1]}. As 2A+1\leq 3A, Lemma 3.1darkblue(?) gives j_! V\in \overline{\langle G\rangle}^{[-3A,3A]}. By Lemma 3.1darkblue(?), \overline{\langle G\rangle}^{[-3A,3A]} is closed under extensions, and so E\in \overline{\langle G\rangle}^{[-3A,3A]}.

Thus, E\to X\to D^{\prime\prime}\to E[1] is a distinguished triangle with E\in \overline{\langle G\rangle}^{[-3A,3A]} and D^{\prime\prime}\in \mathcal{D}^{\leq -1}. As X\in \mathcal{D}^{\leq 0} was arbitrary, (G,\mathcal{D}^{\leq 0},3A) is weak approximation data for \mathcal{D}. Therefore, \mathcal{D} is weakly approximable.

By [Neeman:2024, Theorem 3.2(iv)], every quasi-compact quasi-separated scheme satisfies W.A. Moreover, Proposition 5.5 asserts that W.A. is preserved under elementary distinguished squares. Consequently, the induction principle [StacksProject, Tag 09IT] yields that every quasi-compact quasi-separated algebraic space satisfies W.A. Taking S=|X| gives the second claim.

We prove the last claim. Here X is Noetherian. Fix a closed subset S\subseteq|X| with quasi-compact complement. Set \mathcal{T}:=D_{\operatorname{qc},S}(X). By Lemma 5.6, \mathcal{T}^-_c = D^-_{\operatorname{coh},S}(X) and \mathcal{T}^b_c = D^b_{\operatorname{coh},S}(X). The standard t-structure restricts to D^-_{\operatorname{coh},S}(X). Thus, we finish by [Canonaco/Haesemeyer/Neeman/Stellari:2024, Example 10.1.2] and the first claim. Indeed, by loc. cit., a weakly approximable category is coherent if some t-structure in the preferred class restricts to \mathcal{T}^-_c.

First we prove darkblue(?). By definition, \overline{\langle H\rangle}^{[a,b]} := \operatorname{smd}(\operatorname{Coprod}(H[a,b])). Let \Lambda be a set. Choose X_\lambda \in \overline{\langle H\rangle}^{[a,b]} where \lambda \in \Lambda. For each \lambda \in \Lambda, there exists X^\prime_\lambda \in \mathcal{T} such that X_\lambda \oplus X^\prime_\lambda\in \operatorname{Coprod}(H[a,b]). Since \operatorname{Coprod}(H[a,b]) is closed under small coproducts, \oplus_{\lambda\in \Lambda} (X_\lambda \oplus X^\prime_\lambda)\in \operatorname{Coprod}(H[a,b]). Note that \oplus_{\lambda\in \Lambda}X_\lambda is a direct summand of \oplus_{\lambda\in \Lambda} (X_\lambda \oplus X^\prime_\lambda). Hence, \overline{\langle H\rangle}^{[a,b]} is closed under small coproducts.

Consider a distinguished triangle

where A,C\in \overline{\langle H\rangle}^{[a,b]}. We can find A^\prime,C^\prime such that A\oplus A^\prime, C\oplus C^\prime \in \operatorname{Coprod}(H[a,b]). There exist distinguished triangles

and

Taking the direct sum of these distinguished triangles yields a distinguished triangle

See [Neeman:2001, Proposition 1.2.1]. Since A\oplus A^\prime, C\oplus C^\prime\in \operatorname{Coprod}(H[a,b]), it follows that B \oplus A^\prime \oplus C^\prime\in \operatorname{Coprod}(H[a,b]) because the category is closed under extensions. Consequently, B \in \overline{\langle H\rangle}^{[a,b]}.

To see that \overline{\langle H\rangle}^{[a,b]} is closed under direct summands, use the fact that a direct summand of a direct summand remains a direct summand.

Lastly, we show the minimality claim. Let \mathcal{B} be a strictly full subcategory of \mathcal{T} closed under small coproducts, extensions, and direct summands that contains H[a,b]. In this case, \operatorname{Coprod}(H[a,b])\subseteq \mathcal{B}. Indeed, \mathcal{B} is closed under small coproducts and extensions, and it contains H[a,b]. Moreover, \operatorname{smd}(\operatorname{Coprod}(H[a,b]))\subseteq \mathcal{B} because \mathcal{B} is closed under direct summands.

Second we prove darkblue(?). The autoequivalence [s]\colon \mathcal{T} \to \mathcal{T} preserves extensions, small coproducts, and direct summands. Hence, \overline{\langle H\rangle}^{[a,b]}[s] is the smallest strictly full subcategory containing (H[a,b])[s] which is closed under extensions, small coproducts, and direct summands. However, shifting by s yields

It follows that \overline{\langle H\rangle}^{[a,b]}[s] = \overline{\langle H\rangle}^{[a-s,b-s]}. Moreover, as H[a,b]\subseteq H[a^\prime,b^\prime], darkblue(?) implies \overline{\langle H\rangle}^{[a,b]} \subseteq \overline{\langle H\rangle}^{[a^\prime,b^\prime]}.

Third we prove darkblue(?). Choose a\leq k\leq b. Note that F is exact and F(H)\in \overline{\langle H^\prime\rangle}^{[c,d]}. Hence, by darkblue(?),

Set \mathcal{S} to be the full subcategory of E\in \mathcal{T} such that F(E)\in \overline{\langle H^\prime\rangle}^{[a+c,b+d]}. It follows that H[a,b]\subseteq \mathcal{S}. By darkblue(?), \overline{\langle H^\prime\rangle}^{[a+c,b+d]} is strictly full and closed under small coproducts, extensions, and direct summands. Since F preserves small coproducts, \mathcal{S} has the same properties. By minimality, darkblue(?) implies \overline{\langle H\rangle}^{[a,b]}\subseteq \mathcal{S}. Thus, darkblue(?) follows.

Next we prove darkblue(?). It is easy to see that \{E : \operatorname{Hom}(E,Y)=0\} is strictly full. Let \Lambda be a set. Choose X_\lambda \in \{E : \operatorname{Hom}(E,Y)=0\}. By the universal property of coproducts, there exists an isomorphism

See [Neeman:2001, Example 1.1.13] applied to \mathcal{T}^{op}. The last isomorphism follows from the fact that each X_\lambda \in \{E : \operatorname{Hom}(E,Y)=0\}. Hence, \{E : \operatorname{Hom}(E,Y)=0\} is closed under small coproducts. Let A,B\in \{E : \operatorname{Hom}(E,Y)=0\}. Consider an extension C given by a distinguished triangle

Apply \operatorname{Hom}(-,Y) to obtain a long exact sequence in Hom sets

Hence, \operatorname{Hom}(C,Y) \cong 0 if

This implies \{E : \operatorname{Hom}(E,Y)=0\} is closed under extensions. Moreover, if A\oplus B \in \{E : \operatorname{Hom}(E,Y)=0\}, then

Thus, \{E : \operatorname{Hom}(E,Y)=0\} is closed under direct summands. The hypothesis asserts that H[a,b]\subseteq \{E : \operatorname{Hom}(E,Y)=0\}. By darkblue(?), \overline{\langle H \rangle}^{[a,b]}\subseteq \{E : \operatorname{Hom}(E,Y)=0\}. A similar argument with H compact proves the second claim for darkblue(?). Indeed, the compactness of H gives the closure under coproducts, whereas the other conditions are analogous.

Lastly we prove darkblue(?). Let X\in \operatorname{smd}(\operatorname{Coprod}(H[a,b])). Choose X^\prime such that X\oplus X^\prime \in \operatorname{Coprod}(H[a,b]). The so-called Eilenberg swindle yields an isomorphism

Consequently, there exists a distinguished triangle

Rotating backwards yields

Since [a,b]\subseteq [a,b+1], \operatorname{Coprod}(H[a,b]) \subseteq \operatorname{Coprod}(H[a,b+1]). Observe that

Therefore,

This completes the proof.

By rotating backwards, there exists a distinguished triangle

Note that D[-1]\in \mathcal{T}^{\leq 1-m} \subseteq \mathcal{T}^{\leq 0}. Since aisles are closed under extensions, E\in \mathcal{T}^{\leq 0}.

This follows from [Neeman:2026, Corollary 3.2(1)] with weak approximation data for \mathcal{T}. To be self-contained, we give a proof. By Lemma 3.3, for any desired distinguished triangle

it follows that E_m\in \mathcal{T}^{\leq 0}. Hence, it suffices to produce the desired distinguished triangle. We argue by induction on m. The base case m=1 is Definition 2.1darkblue(?).

Assume that the desired distinguished triangle exists for m, i.e. a distinguished triangle

with E_m\in \overline{\langle H\rangle}^{[1-m-A,A]} \cap \mathcal{T}^{\leq 0} and D_m\in \mathcal{T}^{\leq -m}. Since D_m[-m]\in \mathcal{T}^{\leq 0}, Definition 2.1darkblue(?) yields a distinguished triangle

where E^\prime \in \overline{\langle H\rangle}^{[-A,A]} and D^\prime \in \mathcal{T}^{\leq -1}.

After shifting by [m], we have a distinguished triangle

where E^\prime[m]\in \overline{\langle H\rangle}^{[-A-m,A-m]} and D^\prime[m]\in \mathcal{T}^{\leq -1-m}. Set D_{m+1} := D^\prime[m]. Complete the composite g\circ f \colon F\to D^\prime[m] to a distinguished triangle,

Set E_{m+1}:=R[-1]. The cocones of f, g, and g\circ f are respectively E_m, E^\prime[m], and E_{m+1}. By Remark 3.2darkblue(?), this yields a distinguished triangle

By Lemma 3.1darkblue(?)darkblue(?),

This completes the proof.

This is [Neeman:2026, Remark 1.24]; cf. [Burke/Neeman/Pauwels:2023, Lemma 3.5(1)]. On one hand,

On the other hand, the strictly full subcategory

contains H[n] for n\geq n_0. Moreover, by Lemma 3.1darkblue(?), this subcategory above is closed under small coproducts and extensions. Consequently, the subcategory above contains \mathcal{T}^{\leq -n_0}. Thus, the desired claim follows.

If n=0, then this is the definition of the glued aisle. Since i^\ast and j^\ast commute with shifts,

Moreover, the functors i^\ast, j^\ast, i_\ast, j_! are right t-exact for the n=0 case. After shifting, the desired containments follow.

darkblue(?) and darkblue(?) are essentially from [Burke/Neeman/Pauwels:2023, Lemma 3.11]. The observation below is [Burke/Neeman/Pauwels:2023, Lemma 3.5(2)] (see also [GuisadoVillaalgordo/Lank:2026, Lemma A.1]). We recall the argument for convenience.

We prove darkblue(?). Recall that j_! preserves compact objects. Hence, j_! G_U\in \mathcal{D}^c. We observe that for all F,F^\prime\in \mathcal{D}^c, there exists an integer N such that \operatorname{Hom}(F,F^\prime[n])=0 for all n\geq N. First, consider the full subcategory \mathcal{H} of F\in \mathcal{D}^c where such an N exists with F^\prime = G^\prime. This subcategory is thick (e.g. use the long exact sequences of \operatorname{Hom}(-,G^\prime[n])). By hypothesis, G^\prime \in \mathcal{H}. As G^\prime is a classical generator of \mathcal{D}^c (see ?), \mathcal{H} = \mathcal{D}^c. Second, fix F\in \mathcal{D}^c. Consider the full subcategory \mathcal{H}^\prime of F^\prime\in \mathcal{D}^c for which such an N exists. Again, \mathcal{H}^\prime is thick, which can be shown using the long exact sequences of \operatorname{Hom}(F,-). Also, by the first step above, G^\prime \in \mathcal{H}^\prime. Hence, \mathcal{H}^\prime = \mathcal{D}^c, which proves the observation.

Applying the observation to F = j_! G_U and F^\prime = G^\prime, and using adjunction, we have for n\gg 0,

Since \mathcal{D}_U is weakly approximable and \mathcal{D}_U^{\leq 0} is in the preferred equivalence class, [Burke/Neeman/Pauwels:2023, Corollary 3.10(1)] implies that j^\ast G^\prime\in \mathcal{D}_U^{\leq m} for some m\geq 0.

Note that j_! G_U is compact. Hence, the hypotheses of [GuisadoVillaalgordo/Lank:2026, Lemma A.1] hold: G^\prime and G_U are respectively compact generators of \mathcal{D} and \mathcal{D}_U, \mathcal{D}_U^{\leq 0} is the aisle compactly generated by G_U, and j^\ast G^\prime\in \mathcal{D}_U^{\leq m} by darkblue(?). Consequently, darkblue(?) and darkblue(?) follow from [GuisadoVillaalgordo/Lank:2026, Lemma A.1].

To prove darkblue(?), apply the observation with F=F^\prime=G.

The following proof is a variant of [Burke/Neeman/Pauwels:2023, Lemma 5.3]. Specifically, the lift is constructed directly from the adjunct g rather than from a roof [Burke/Neeman/Pauwels:2023, Lemma 5.2], and it lies in \mathcal{D}^{\leq K} for the same K that bounds Y.

Since i^\ast is exact, i^\ast(G[n]) can be identified with (i^\ast G)[n]. Hence, f^\prime yields a morphism i^\ast(G[n])\to i^\ast Y. Denote by \eta the unit and \overline{\varepsilon} the counit of i^\ast\dashv i_\ast.

Set M := j^\ast Y. By Lemma 4.2, M\in \mathcal{D}_U^{\leq K}, and so M[K]\in \mathcal{D}_U^{\leq 0}. Apply Lemma 3.4 to the weak approximation data (G_U,\mathcal{D}_U^{\leq 0},A) of Setup 4.4darkblue(?), M[K], and 3A+1 = K+2A in place of m. This gives a distinguished triangle

Here W^\prime\in \overline{\langle G_U\rangle}^{[-4A,A]}\cap \mathcal{D}_U^{\leq 0} and V^\prime\in \mathcal{D}_U^{\leq -3A-1}. Define W := W^\prime[-K] and V := V^\prime[-K]. Shifting by [-K] yields a distinguished triangle

By Lemma 3.1darkblue(?),

Moreover, W\in \mathcal{D}_U^{\leq 0}[-K] = \mathcal{D}_U^{\leq K} and V\in \mathcal{D}_U^{\leq -3A-1}[-K] = \mathcal{D}_U^{\leq -2A}.

Now let g\colon G[n]\to i_\ast i^\ast Y be the adjunct of f^\prime. That is, g = i_\ast(f^\prime)\circ \eta_{G[n]}, and so f^\prime = \overline{\varepsilon}_{i^\ast Y}\circ i^\ast(g). Set

to be the morphism in darkblue(4.1), and \xi := \varphi\circ g. By Lemma 4.2, j_! V\in \mathcal{D}^{\leq -2A}. As n\leq A,

Hence, Setup 4.4darkblue(?) gives

Applying j_! to darkblue(?) gives a distinguished triangle

By [Neeman:2001, Lemma 1.1.10], \operatorname{Hom}(G[n-1],-) is a homological functor on \mathcal{D}. This yields an exact sequence

By darkblue(?),

Hence, the left-hand map is surjective. Thus, there exists \xi^{\prime\prime}\colon G[n-1]\to j_! W such that \xi[-1] = j_!(w)\circ \xi^{\prime\prime}. Setting \xi^\prime := \xi^{\prime\prime}[1]\colon G[n]\to j_! W[1], we get \xi = j_!(w)[1]\circ \xi^\prime.

Next we construct the lift. Completing \xi^\prime and rotating backwards twice yields a distinguished triangle

Consider the diagram

with distinguished rows. Indeed, the top row is darkblue(?), whereas the bottom row is the first triangle of darkblue(4.1) for E = Y (i.e. use that M = j^\ast Y). The right-hand square commutes because

By TR3, applied after rotating both rows backwards, there exists q\colon Q\to Y making the diagram commute. In particular, \eta_Y\circ q = g\circ \beta.

Set \theta := i^\ast\beta\colon i^\ast Q\to i^\ast(G[n]) = (i^\ast G)[n]. Recall that i^\ast is exact and i^\ast j_! = 0. Applying i^\ast to darkblue(?) gives a distinguished triangle

and so \theta is an isomorphism. Moreover, applying i^\ast to \eta_Y\circ q = g\circ \beta and using the triangle identity \overline{\varepsilon}_{i^\ast Y}\circ i^\ast(\eta_Y) = 1_{i^\ast Y}, it follows that

Lastly, we check that Q lies in \mathcal{P}. By darkblue(?), Q\in j_! W\star G[n]. Since \mathcal{P} is closed under extensions, it suffices to show that j_! W, G[n]\in \mathcal{P}. The functor j_! is exact, preserves small coproducts, and satisfies j_! G_U\in \overline{\langle G\rangle}^{[0,0]} by Setup 4.4. By Lemma 3.1darkblue(?), j_! W\in \overline{\langle G\rangle}^{[1-3A,2A+1]}. As A\geq 1, we have 2A+1\leq 3A = B, so Lemma 3.1darkblue(?) gives j_! W\in \overline{\langle G\rangle}^{[-B,B]}. Moreover, by Lemma 4.2, W\in \mathcal{D}_U^{\leq K} implies j_! W\in \mathcal{D}^{\leq K}. Since -A-1\leq n\leq A, one has -A\leq -n\leq A+1, and so G[n] = G[-(-n)]\in G[-A,A+1]. Applying Lemma 3.1darkblue(?) with A+1\leq 3A, G[n]\in \overline{\langle G\rangle}^{[-A,A+1]}\subseteq \overline{\langle G\rangle}^{[-B,B]}. As G\in \mathcal{D}^{\leq 0} by Setup 4.4, G[n]\in \mathcal{D}^{\leq -n}\subseteq \mathcal{D}^{\leq A+1} = \mathcal{D}^{\leq K}. Thus, Q\in \mathcal{P}, and so (q,\theta) is a \mathcal{P}-lift of f^\prime.

First we show that \mathcal{C} is strictly full. Let \sigma\colon E^\prime\to E^{\prime\prime} be an isomorphism with E^{\prime\prime}\in \mathcal{C}. Fix f^\prime\colon E^\prime\to i^\ast Y with Y\in \mathcal{D}^{\leq K}. Since E^{\prime\prime}\in \mathcal{C}, f^\prime\circ \sigma^{-1} has a \mathcal{P}-lift (q,\theta). Then \sigma^{-1}\circ \theta is an isomorphism, and i^\ast q = f^\prime\circ (\sigma^{-1}\circ \theta). Thus, (q,\sigma^{-1}\circ \theta) is a \mathcal{P}-lift of f^\prime. Hence, E^\prime\in \mathcal{C}.

By Lemma 4.6, (i^\ast G)[n] \in \mathcal{C} for -A-1\leq n\leq A.

Next we show that \mathcal{C} is closed under small coproducts. Choose a set \Lambda. Pick some E^\prime_\lambda\in \mathcal{C} for each \lambda\in \Lambda. Put E^\prime := \bigoplus_{\lambda\in \Lambda} E^\prime_\lambda. Denote by \iota_\lambda\colon E^\prime_\lambda\to E^\prime the inclusions. Fix f^\prime\colon E^\prime\to i^\ast Y such that Y\in \mathcal{D}^{\leq K}. For each \lambda \in \Lambda, there exists a \mathcal{P}-lift (q_\lambda\colon E_\lambda\to Y,\theta_\lambda) of f^\prime\circ \iota_\lambda. Define E := \bigoplus_{\lambda\in \Lambda} E_\lambda. Set \kappa_\lambda\colon E_\lambda\to E to be the inclusions for each \lambda \in \Lambda. Since \mathcal{P} is closed under small coproducts, E\in \mathcal{P}. Denote by q\colon E\to Y the morphism with q\circ \kappa_\lambda = q_\lambda for each \lambda \in \Lambda. As i^\ast preserves small coproducts, i^\ast E is the coproduct of the i^\ast E_\lambda with inclusions i^\ast(\kappa_\lambda) for \lambda \in \Lambda. Hence, there exists \theta\colon i^\ast E\to E^\prime with \theta\circ i^\ast(\kappa_\lambda) = \iota_\lambda\circ \theta_\lambda for each \lambda \in \Lambda. Note that \theta is an isomorphism because it is a coproduct of the isomorphisms \theta_\lambda. For each \lambda \in \Lambda,

Hence, by the universal property of coproducts, i^\ast q = f^\prime\circ \theta. Thus, (q,\theta) is a \mathcal{P}-lift of f^\prime.

Now we prove that \mathcal{C} is closed under extensions. Consider a distinguished triangle

where E^\prime_1,E^\prime_2\in \mathcal{C}. Fix f^\prime\colon E^\prime\to i^\ast Y with Y\in \mathcal{D}^{\leq K}. We will lift f^\prime in two stages: first on E^\prime_1, and then on E^\prime_2, into the cone of the first lift.

We start with E^\prime_1. As E^\prime_1\in \mathcal{C}, f^\prime_1 := f^\prime\circ u^\prime has a \mathcal{P}-lift (q_1\colon E_1\to Y,\theta_1). Complete to a distinguished triangle

Since E_1\in \mathcal{P} \subseteq \mathcal{D}^{\leq K}, E_1[1]\in \mathcal{D}^{\leq K-1}\subseteq \mathcal{D}^{\leq K}. As aisles are closed under extensions, Y^\prime\in Y\star E_1[1]\subseteq \mathcal{D}^{\leq K}.

Applying i^\ast to this distinguished triangle gives the top row of the diagram

Since i^\ast q_1 = f^\prime_1\circ \theta_1, the left square commutes. By construction, the other two squares commute. As all vertical maps are isomorphisms, the bottom row is isomorphic to a distinguished triangle, and hence is distinguished.

Next we find the morphism that f^\prime induces on E^\prime_2. Again, we complete to a distinguished triangle

We apply Remark 3.2darkblue(?) to E^\prime_1\xrightarrow{u^\prime} E^\prime\xrightarrow{f^\prime} i^\ast Y. For this, we use the given distinguished triangle on u^\prime, the chosen distinguished triangle on f^\prime, and the bottom row of the diagram above as the distinguished triangle on f^\prime\circ u^\prime = f^\prime_1. This gives a distinguished triangle

such that the following diagram commutes:

Now we lift f^\prime_2. Since E^\prime_2\in \mathcal{C} and Y^\prime\in \mathcal{D}^{\leq K}, f^\prime_2 has a \mathcal{P}-lift (q_2\colon E_2\to Y^\prime,\theta_2). Choose a distinguished triangle

Next we combine the two lifts. Complete to a distinguished triangle

The cocones of r, s_2, and s_2\circ r are respectively E_1, E_2, and E. Hence, Remark 3.2darkblue(?) gives a distinguished triangle E_1\to E\to E_2\to E_1[1]. Since \mathcal{P} is closed under extensions, E\in E_1\star E_2\subseteq \mathcal{P}.

It remains to compare i^\ast q with f^\prime. We do this in two steps. First, applying i^\ast to the distinguished triangle on q_2 gives the top row of the diagram

Note that the bottom row is the distinguished triangle on f^\prime_2 found above. As i^\ast q_2 = f^\prime_2\circ \theta_2, the left square commutes. By TR3, there exists a morphism \chi\colon i^\ast Y^{\prime\prime}\to R^\prime making the diagram commute. Hence, \chi\circ i^\ast s_2 = h. Since \theta_2 and the identity are isomorphisms, so is \chi. Combine with h\circ i^\ast r = p^\prime to obtain

Applying i^\ast to the distinguished triangle on q gives

The bottom row is the distinguished triangle on f^\prime, whereas the middle square commutes by the diagram above. By TR3, applied after rotating both rows, there exists a morphism \theta\colon i^\ast E\to E^\prime making the diagram commute. Hence, f^\prime\circ \theta = i^\ast q. As before, \theta is an isomorphism, because 1 and \chi are. Since E\in \mathcal{P}, (q,\theta) is a \mathcal{P}-lift of f^\prime. Hence, E^\prime\in \mathcal{C}.

Finally, we prove the two inclusions. The objects of (i^\ast G)[-A,A+1] are (i^\ast G)[-k] with -A\leq k\leq A+1. For n = -k, these are the objects (i^\ast G)[n] with -A-1\leq n\leq A. Hence, by Lemma 4.6, these lie in \mathcal{C}. Moreover, by definition, \operatorname{Coprod}((i^\ast G)[-A,A+1]) is the smallest strictly full subcategory containing these objects that is closed under small coproducts and extensions. Since \mathcal{C} has all these properties, \operatorname{Coprod}((i^\ast G)[-A,A+1])\subseteq \mathcal{C}. The first inclusion is Lemma 3.1darkblue(?).

By [StacksProject, Tag 0AED], G^\prime is perfect. Hence, G^\prime \in D^-_{\operatorname{qc},S}(X). By [GuisadoVillaalgordo/Lank:2026, Lemma 3.40], it follows that \operatorname{Hom}(G^\prime[-t],G^\prime)=0 for t\gg 0. Thus, \operatorname{Hom}(G^\prime,G^\prime[n])=0 for n\gg 0. (See also the proof of [GuisadoVillaalgordo/Lank:2026, Lemma A.8]).

Fix a closed subset S\subseteq |X| with quasi-compact complement. Set T := |X|\setminus |U|. By [GuisadoVillaalgordo/Lank:2026, Lemma A.4], there exists a recollement

where i_\ast is the natural inclusion. We apply Theorem 1.1 to this recollement. In the notation of Setup 4.1, the roles are as follows:

• \mathcal{D}_Z = D_{\operatorname{qc},|U|\cap S}(U), where \mathbf{L}j^\ast, \mathbf{R}j_\ast, j^!_S (of the diagram above) are respectively identified with i^\ast, i_\ast, i^!

• \mathcal{D} = D_{\operatorname{qc},S}(X)

• \mathcal{D}_U = D_{\operatorname{qc},S\cap T}(X), where i_\ast, i^!, i^\ast are respectively identified with j_!, j^\ast, j_\ast (in Setup 4.1).

First, |U|\setminus (|U|\cap S) = |U|\cap (|X|\setminus S). Since |U|\cap (|X|\setminus S) is the intersection of two quasi-compact opens of the quasi-separated space X, it is quasi-compact. As U satisfies W.A., \mathcal{D}_Z is weakly approximable.

Next, apply [GuisadoVillaalgordo/Lank:2026, Lemma A.6] to T\cap S. This gives a triangulated equivalence

Note that |Y|\setminus f^{-1}(T\cap S) = f^{-1}(|U|\cup (|X|\setminus S)). Since Y is quasi-compact and X is quasi-separated, f is quasi-compact [StacksProject, Tag 073B]. As |U|\cup (|X|\setminus S) is a quasi-compact open, f^{-1}(|U|\cup (|X|\setminus S)) is quasi-compact. Moreover, as Y satisfies W.A., D_{\operatorname{qc},f^{-1}(T\cap S)}(Y) is weakly approximable. By [Canonaco/Neeman/Stellari:2024, Remark 3.4(i)], weak approximability is invariant under triangulated equivalence. Thus, \mathcal{D}_U is weakly approximable.

Observe that |X|\setminus (S\cap T) = (|X|\setminus S)\cup |U|, which is quasi-compact. Moreover, the three triangulated categories in the recollement are singly compactly generated [StacksProject, Tags 0AED, 09M8, & 0AEC]. If G^\prime is a compact generator of D_{\operatorname{qc},S}(X), then Lemma 5.4 says that \operatorname{Hom}(G^\prime,G^\prime[n])=0 for n\gg 0. Consequently, Theorem 1.1 asserts that D_{\operatorname{qc},S}(X) is weakly approximable. Therefore, as S was arbitrary, X satisfies W.A.

By [GuisadoVillaalgordo/Lank:2026, Lemma A.9], the standard t-structure on \mathcal{T} lies in the preferred equivalence class. Hence, we compute \mathcal{T}^-_c and \mathcal{T}^b_c with respect to it.

Pick E\in\mathcal{T}^-_c and n>0. Choose a distinguished triangle

with D\in\mathcal{T}^c and F\in\mathcal{T}^{\leq -n}. By [StacksProject, Tag 0AED], D is perfect, and hence, pseudocoherent [StacksProject, Tag 08G8]. Since \mathcal{H}^j(F)=0 for j\geq 1-n, F is (1-n)-pseudocoherent. Hence, E is (1-n)-pseudocoherent [StacksProject, Tag 08FV]. As n was arbitrary, E is pseudocoherent.

Conversely, let E\in\mathcal{T} be pseudocoherent. Fix n>0. The open immersion X\setminus S\to X is quasi-compact [StacksProject, Tag 073B]. By [StacksProject, Tags 08HP & 08HJ], there exist a perfect complex P supported on S and P\to E inducing isomorphisms on \mathcal{H}^i for i>1-n and a surjection on \mathcal{H}^{1-n}. Note that the cone of P\to E lies in \mathcal{T}^{\leq -n}. By [StacksProject, Tag 0AED], P\in\mathcal{T}^c. Therefore, E\in\mathcal{T}^-_c, which gives the first claim.

We show the second claim. By [StacksProject, Tag 08IK], the pseudocoherent objects of D_{\operatorname{qc}}(X) are exactly those of D^-_{\operatorname{coh}}(X). Hence, \mathcal{T}^-_c = D^-_{\operatorname{coh}}(X) \cap \mathcal{T} = D^-_{\operatorname{coh},S}(X), and \mathcal{T}^b_c = \mathcal{T}^b \cap\mathcal{T}^-_c =D^b_{\operatorname{coh},S}(X). This completes the proof.