Derived functors formalize the idea that a non-exact functor can be “corrected” by replacing objects with resolutions and then taking homology/cohomology.

Throughout, let A,B\mathcal A,\mathcal B be and F:A→BF:\mathcal A\to\mathcal B an additive functor.

Right derived functors (from injective resolutions)

Assume FF is and A\mathcal A has enough injectives (so exist).

For A∈AA\in\mathcal A, choose an injective resolution 0→A→I∙0\to A\to I^\bullet. Apply FF termwise to get a cochain complex F(I∙)F(I^\bullet) in B\mathcal B. The right derived functors of FF are

RnF(A)  :=  Hn(F(I∙))(n≥0),R^nF(A)\;:=\;H^n\bigl(F(I^\bullet)\bigr)\qquad (n\ge 0),

where HnH^n is .

Key facts (standard in homological algebra):

  • R0F≅FR^0F \cong F.
  • RnF(A)R^nF(A) is independent of the chosen injective resolution up to canonical isomorphism.
  • A short exact sequence in A\mathcal A induces a in the RnFR^nF.
Left derived functors (from projective resolutions)

Assume FF is right exact and A\mathcal A has enough projectives (so exist).

For A∈AA\in\mathcal A, choose a projective resolution P∙→AP_\bullet\to A. Apply FF termwise to get a chain complex F(P∙)F(P_\bullet) in B\mathcal B. The left derived functors are

LnF(A)  :=  Hn(F(P∙))(n≥0),L_nF(A)\;:=\;H_n\bigl(F(P_\bullet)\bigr)\qquad (n\ge 0),

where HnH_n is .

Again:

  • L0F≅FL_0F \cong F.
  • LnF(A)L_nF(A) is well-defined up to canonical isomorphism.
  • Short exact sequences yield long exact sequences in the LnFL_nF.
Fundamental examples: Ext and Tor

In A=R-Mod\mathcal A = R\text{-Mod}:

  • The functor Hom⁡R(M,−)\operatorname{Hom}_R(M,-) is , and its right derived functors are RnHom⁡R(M,−)≅ExtRn(M,−)R^n\operatorname{Hom}_R(M,-)\cong \mathrm{Ext}_R^n(M,-) (see ).
  • The functor −⊗RN-\otimes_R N is , and its left derived functors are Ln(−⊗RN)≅TornR(−,N)L_n(-\otimes_R N)\cong \mathrm{Tor}_n^R(-,N) (see ).

See also .

Examples (explicit computations)

Example 1: Computing Tor1Z(Z/n,Z/m)\mathrm{Tor}_1^{\mathbb Z}(\mathbb Z/n,\mathbb Z/m) as a left derived functor

Let F(−)=−⊗ZZ/mZF(-)= -\otimes_{\mathbb Z}\mathbb Z/m\mathbb Z. Take the projective resolution

0→Z→×nZ→Z/nZ→0.0\to \mathbb Z \xrightarrow{\times n} \mathbb Z \to \mathbb Z/n\mathbb Z \to 0.

Applying FF gives

0→Z/m→×nZ/m→0,0\to \mathbb Z/m \xrightarrow{\times n} \mathbb Z/m \to 0,

so

L1F(Z/n)≅H1≅ker⁡(×n:Z/m→Z/m)≅Z/gcd⁡(m,n)Z.L_1F(\mathbb Z/n)\cong H_1\cong \ker(\times n:\mathbb Z/m\to\mathbb Z/m)\cong \mathbb Z/\gcd(m,n)\mathbb Z.

Thus Tor1Z(Z/n,Z/m)≅Z/gcd⁡(m,n)Z\mathrm{Tor}_1^{\mathbb Z}(\mathbb Z/n,\mathbb Z/m)\cong \mathbb Z/\gcd(m,n)\mathbb Z (see ).

Example 2: Computing ExtZ1(Z/n,Z)\mathrm{Ext}^1_{\mathbb Z}(\mathbb Z/n,\mathbb Z) as a right derived functor

Let F(−)=Hom⁡Z(Z/n,−)F(-)=\operatorname{Hom}_{\mathbb Z}(\mathbb Z/n,-), which is left exact. Use the injective resolution

0→Z→Q→Q/Z→0.0\to \mathbb Z \to \mathbb Q \to \mathbb Q/\mathbb Z \to 0.

Applying FF gives

0→Hom⁡(Z/n,Q)→Hom⁡(Z/n,Q/Z)→0,0\to \operatorname{Hom}(\mathbb Z/n,\mathbb Q)\to \operatorname{Hom}(\mathbb Z/n,\mathbb Q/\mathbb Z)\to 0,

with Hom⁡(Z/n,Q)=0\operatorname{Hom}(\mathbb Z/n,\mathbb Q)=0 and Hom⁡(Z/n,Q/Z)≅Z/n\operatorname{Hom}(\mathbb Z/n,\mathbb Q/\mathbb Z)\cong \mathbb Z/n. Hence

R1F(Z)≅H1≅Z/nZ,R^1F(\mathbb Z)\cong H^1 \cong \mathbb Z/n\mathbb Z,

i.e. ExtZ1(Z/n,Z)≅Z/nZ\mathrm{Ext}^1_{\mathbb Z}(\mathbb Z/n,\mathbb Z)\cong \mathbb Z/n\mathbb Z (see ).

Example 3: Exact functors have no higher derived functors

If FF is an (e.g. in module categories, localization is exact on suitable classes of modules), then applying FF to any resolution preserves exactness. Consequently,

RnF=0 (n>0)andLnF=0 (n>0),R^nF=0 \ (n>0)\quad\text{and}\quad L_nF=0 \ (n>0),

whenever the relevant derived functors are defined (right/left).