ConceptComplete

Total Derived Functor RF, LF

The total derived functor lifts a functor between abelian categories to a functor between their derived categories. Unlike the classical approach (which produces a sequence of functors RnFR^nF), the total derived functor RFRF is a single exact functor that encodes all derived functors simultaneously, together with the extension data between them.


Definitions

Definition7.12Total Right Derived Functor

Let F:A→BF : \mathcal{A} \to \mathcal{B} be a left exact functor between abelian categories, with A\mathcal{A} having enough injectives. The total right derived functor is the exact functor

RF:D+(A)→D+(B)RF : D^+(\mathcal{A}) \to D^+(\mathcal{B})

defined by RF(X∙)=F(I∙)RF(X^\bullet) = F(I^\bullet) where X∙→I∙X^\bullet \to I^\bullet is a quasi-isomorphism to a complex of injectives. The classical derived functors are recovered: RnF(A)=Hn(RF(A[0]))R^nF(A) = H^n(RF(A[0])).

Definition7.13Total Left Derived Functor

For a right exact functor G:A→BG : \mathcal{A} \to \mathcal{B} with A\mathcal{A} having enough projectives:

LG:D−(A)→D−(B)LG : D^-(\mathcal{A}) \to D^-(\mathcal{B})

is defined by LG(X∙)=G(P∙)LG(X^\bullet) = G(P^\bullet) where P∙→X∙P^\bullet \to X^\bullet is a quasi-isomorphism from a complex of projectives. Then LnG(A)=H−n(LG(A[0]))L_nG(A) = H^{-n}(LG(A[0])).

Definition7.14Adapted Class

A class of objects R⊆A\mathcal{R} \subseteq \mathcal{A} is adapted (or FF-acyclic) to a left exact functor FF if: (1) every object of A\mathcal{A} embeds into an object of R\mathcal{R}, and (2) FF sends acyclic complexes in K+(R)K^+(\mathcal{R}) to acyclic complexes. Then RFRF can be computed using R\mathcal{R}-resolutions instead of injective resolutions.


Examples

ExampleRF applied to a single object

For A∈AA \in \mathcal{A} viewed as A[0]∈D+(A)A[0] \in D^+(\mathcal{A}), RF(A[0])=F(I∙)RF(A[0]) = F(I^\bullet) where A→I∙A \to I^\bullet is an injective resolution. The cohomology Hn(RF(A[0]))=RnF(A)H^n(RF(A[0])) = R^nF(A) recovers the classical derived functors. But RF(A[0])RF(A[0]) is the entire complex, not just its cohomology.

ExampleRF remembers more than R^nF

The complex RF(A)RF(A) contains extension data lost by the individual RnF(A)R^nF(A). For example, the class of RF(A)RF(A) in D+(B)D^+(\mathcal{B}) determines not just the groups RnF(A)R^nF(A) but also the differentials in spectral sequences computing them. A complex is not determined by its cohomology groups alone.

ExampleTotal derived Hom and RHom

The total derived functor of HomR(A,−)\mathrm{Hom}_R(A, -) is RHomR(A,−):D+(R)→D+(Ab)R\mathrm{Hom}_R(A, -) : D^+(R) \to D^+(\mathbf{Ab}). For complexes A∙,B∙A^\bullet, B^\bullet, RHom(A∙,B∙)R\mathrm{Hom}(A^\bullet, B^\bullet) is a complex whose cohomology gives Extn(A∙,B∙)=Hn(RHom(A∙,B∙))\mathrm{Ext}^n(A^\bullet, B^\bullet) = H^n(R\mathrm{Hom}(A^\bullet, B^\bullet)).

ExampleDerived tensor product

The total left derived functor of −⊗R−- \otimes_R - is −⊗RL−:D−(R)×D−(R)→D−(R)- \otimes^L_R - : D^-(R) \times D^-(R) \to D^-(R). For modules, A⊗RLB=P∙⊗RBA \otimes^L_R B = P_\bullet \otimes_R B where P∙→AP_\bullet \to A is a projective resolution. Then Torn(A,B)=H−n(A⊗LB)\mathrm{Tor}_n(A, B) = H^{-n}(A \otimes^L B).

ExampleDerived global sections

RΓ(X,−):D+(Sh(X))→D+(Ab)R\Gamma(X, -) : D^+(\mathbf{Sh}(X)) \to D^+(\mathbf{Ab}) is the total derived functor of Γ(X,−)\Gamma(X, -). For a sheaf F\mathcal{F}, RΓ(X,F)R\Gamma(X, \mathcal{F}) is a complex whose cohomology is Hn(X,F)H^n(X, \mathcal{F}). One can compute using flasque, soft, or fine resolutions (not just injective).

ExampleDerived pushforward

For f:X→Yf : X \to Y, Rf∗:D+(Sh(X))→D+(Sh(Y))Rf_* : D^+(\mathbf{Sh}(X)) \to D^+(\mathbf{Sh}(Y)) is the total derived functor of f∗f_*. The stalk formula (Rnf∗F)y=Hn(f−1(y),F)(R^nf_*\mathcal{F})_y = H^n(f^{-1}(y), \mathcal{F}) comes from HnH^n of the total derived functor. The Leray spectral sequence Hp(Y,Rqf∗F)⇒Hp+q(X,F)H^p(Y, R^qf_*\mathcal{F}) \Rightarrow H^{p+q}(X, \mathcal{F}) arises from the composition RΓY∘Rf∗=RΓXR\Gamma_Y \circ Rf_* = R\Gamma_X.

ExampleComposition of derived functors

For composable left exact functors F:A→BF : \mathcal{A} \to \mathcal{B} and G:B→CG : \mathcal{B} \to \mathcal{C}, there is a natural morphism R(G∘F)→RG∘RFR(G \circ F) \to RG \circ RF. The Grothendieck spectral sequence arises from the failure of this to be an isomorphism at the cohomological level: RpG(RqF(A))⇒Rp+q(G∘F)(A)R^pG(R^qF(A)) \Rightarrow R^{p+q}(G \circ F)(A).

ExampleAcyclic resolutions

Flasque sheaves are Γ\Gamma-acyclic, so RΓ(X,F)R\Gamma(X, \mathcal{F}) can be computed using a flasque resolution. Similarly, soft sheaves on a paracompact space and fine sheaves on a smooth manifold are Γ\Gamma-acyclic. The de Rham complex ΩX∙\Omega^\bullet_X is a fine resolution of RX\mathbb{R}_X, so RΓ(X,RX)≃Γ(X,ΩX∙)R\Gamma(X, \mathbb{R}_X) \simeq \Gamma(X, \Omega^\bullet_X).

ExampleDerived pullback

For f:X→Yf : X \to Y, the inverse image f∗:Sh(Y)→Sh(X)f^* : \mathbf{Sh}(Y) \to \mathbf{Sh}(X) is exact, so Lf∗=f∗Lf^* = f^* (no higher derived functors). But for OX\mathcal{O}_X-modules, f∗=f−1(−)⊗f−1OYOXf^* = f^{-1}(-) \otimes_{f^{-1}\mathcal{O}_Y} \mathcal{O}_X involves a tensor product, giving Lf∗=f−1(−)⊗f−1OYLOXLf^* = f^{-1}(-) \otimes^L_{f^{-1}\mathcal{O}_Y} \mathcal{O}_X with higher Tor terms.

ExampleAdjunction at the derived level

If F⊣GF \dashv G is an adjunction with FF right exact and GG left exact, then under suitable hypotheses LF⊣RGLF \dashv RG is an adjunction at the derived level: HomD(B)(LF(X),Y)≅HomD(A)(X,RG(Y))\mathrm{Hom}_{D(\mathcal{B})}(LF(X), Y) \cong \mathrm{Hom}_{D(\mathcal{A})}(X, RG(Y)). The derived tensor-Hom adjunction A⊗L−⊣RHom(A,−)A \otimes^L - \dashv R\mathrm{Hom}(A, -) is the prototypical example.

ExampleDualizing complex

On a scheme XX, the dualizing complex ωX∙∈Dcb(X)\omega_X^\bullet \in D^b_c(X) represents the functor RHom(−,ωX∙)R\mathrm{Hom}(-, \omega_X^\bullet), which gives Grothendieck duality: Rf∗RHom(F,ωX∙)≅RHom(Rf!F,ωY∙)Rf_* R\mathrm{Hom}(\mathcal{F}, \omega_X^\bullet) \cong R\mathrm{Hom}(Rf_!\mathcal{F}, \omega_Y^\bullet). For a smooth variety, ωX∙=ωX[dim⁡X]\omega_X^\bullet = \omega_X[\dim X].

ExampleDerived categories of coherent sheaves

The functor RHom(E,−):Db(Coh(X))→Db(Coh(X))R\mathrm{Hom}(\mathcal{E}, -) : D^b(\mathrm{Coh}(X)) \to D^b(\mathrm{Coh}(X)) for a vector bundle E\mathcal{E} is computed using a locally free resolution of E\mathcal{E}. The Fourier-Mukai transform ΦP=Rp2∗(p1∗(−)⊗LP)\Phi_{\mathcal{P}} = Rp_{2*}(p_1^*(-) \otimes^L \mathcal{P}) is a composition of total derived functors.

ExampleProjection formula

For f:X→Yf : X \to Y proper and F∈Db(X)\mathcal{F} \in D^b(X), G∈Db(Y)\mathcal{G} \in D^b(Y): the projection formula states Rf∗(F⊗Lf∗G)≅Rf∗F⊗LGRf_*(\mathcal{F} \otimes^L f^*\mathcal{G}) \cong Rf_*\mathcal{F} \otimes^L \mathcal{G}. This is an identity of total derived functors, not just of individual cohomology groups.


Theorem7.5Existence and uniqueness of total derived functors

Let F:A→BF : \mathcal{A} \to \mathcal{B} be a left exact functor with A\mathcal{A} having enough injectives. Then RF:D+(A)→D+(B)RF : D^+(\mathcal{A}) \to D^+(\mathcal{B}) exists and is unique (up to canonical isomorphism) as the right derived functor of FF in the sense of derived categories. It satisfies the universal property: RFRF is initial among exact functors G:D+(A)→D+(B)G : D^+(\mathcal{A}) \to D^+(\mathcal{B}) equipped with a natural transformation Q∘F→G∘QQ \circ F \to G \circ Q (where QQ is the localization functor).

RemarkWhy total derived functors matter

The passage from classical derived functors RnFR^nF to total derived functors RFRF is analogous to the passage from cohomology groups HnH^n to the derived category. Individual groups lose information about extensions and differentials; the total object preserves them. This is why derived categories and total derived functors are essential for the Grothendieck duality, base change theorems, and the formalism of six functors.