Blog Archive

Tuesday, June 16, 2026

From the 9 Lemma to the Five Lemma

From the 3×3 Lemma to the Five Lemma

This note explains three basic results in homological algebra:

  1. two-out-of-three for exact complexes in a short exact sequence of complexes;

  2. the 3×3 lemma;

  3. the reduction of the five lemma to the short five lemma.

The guiding idea is:

Do not chase the whole diagram at once. Identify the structure first.

1. Two-out-of-three for exact complexes

Let

0ABC0

be a short exact sequence of complexes in an abelian category.

We claim that if two of

A,B,C

are exact, then the third is exact.

Proof

A short exact sequence of complexes induces a long exact sequence in homology:

Hn(A)Hn(B)Hn(C)δHn1(A)Hn1(B).

A complex X is exact if and only if

Hn(X)=0for all n.

Suppose first that A and B are exact. Then

Hn(A)=Hn(B)=Hn1(A)=0.

The exact segment

Hn(B)Hn(C)Hn1(A)

therefore becomes

0Hn(C)0.

Thus Hn(C)=0 for all n, so C is exact.

Suppose next that A and C are exact. The exact segment

Hn(A)Hn(B)Hn(C)

becomes

0Hn(B)0.

Thus Hn(B)=0 for all n, so B is exact.

Finally, suppose that B and C are exact. The exact segment

Hn+1(C)Hn(A)Hn(B)

becomes

0Hn(A)0.

Thus Hn(A)=0 for all n, so A is exact.

Therefore exactness satisfies two-out-of-three in a short exact sequence of complexes.


2. Subobjects and quotients of three-term complexes

We first isolate a small observation.

Lemma

Let

X2d2XX1d1XX0

be a three-term complex, so

d1Xd2X=0.

Suppose

S2d2SS1d1SS0i2i1i0X2d2XX1d1XX0

is a commutative diagram and i0,i1,i2 are monomorphisms. Then

S2d2SS1d1SS0

is also a complex.

Dually, suppose

X2d2XX1d1XX0q2q1q0Q2d2QQ1d1QQ0

is a commutative diagram and q0,q1,q2 are epimorphisms. Then

Q2d2QQ1d1QQ0

is also a complex.

Proof

For the subobject statement, compute:

i0d1Sd2S=d1Xd2Xi2=0.

Since i0 is monic,

d1Sd2S=0.

Hence S is a complex.

For the quotient statement, by duality or compute:

d1Qd2Qq2=q0d1Xd2X=0.

Since q2 is epic,

d1Qd2Q=0.

Hence Q is a complex.


3. The 3×3 lemma

Suppose we have a commutative diagram in an abelian category:

0AaBbC00AaBbC00AaBbC0

and suppose every column is exact.

We want to prove:

  1. if the bottom two rows are exact, then the top row is exact;

  2. if the top two rows are exact, then the bottom row is exact;

  3. if the top and bottom rows are exact and ba=0, then the middle row is exact.

Proof

Write the three rows as

R=(0AaBbC0),
R=(0AaBbC0),

and

R=(0AaBbC0).

A row is exact precisely when the corresponding finite complex is exact.

Since every column is exact, once the rows involved are complexes, the vertical maps give a short exact sequence of complexes:

0RRR0.
Case 1: the bottom two rows are exact

Then R and R are exact complexes. In particular, R is a complex.

Since every column is exact, the maps

AA,BB,CC

are monomorphisms. Hence R is a subobject of R in the category of three-term diagrams.

By the previous lemma, R is a complex.

Therefore we have a short exact sequence of complexes:

0RRR0.

Since R and R are exact, two-out-of-three implies that R is exact. Hence the top row is exact.

Case 2: the top two rows are exact

Then R and R are exact complexes. In particular, R is a complex.

Since every column is exact, the maps

AA,BB,CC

are epimorphisms. Hence R is a quotient object of R in the category of three-term diagrams.

By the previous lemma, R is a complex.

Therefore we have a short exact sequence of complexes:

0RRR0.

Since R and R are exact, two-out-of-three implies that R is exact. Hence the bottom row is exact.

Case 3: the top and bottom rows are exact, and ba=0

The top and bottom rows being exact means that R and R are exact complexes.

The additional assumption

AaBbC=0

is precisely the assertion that R is a complex.

Thus we again have a short exact sequence of complexes:

0RRR0.

Since R and R are exact, two-out-of-three implies that R is exact. Hence the middle row is exact.

This proves the 3×3 lemma.


4. The short five lemma
Lemma

Let A be an abelian category. Suppose we have a commutative diagram whose rows are short exact:

0AiBpC0fgh0AiBpC0.

If f and h are isomorphisms, then g is an isomorphism.

Proof using the arrow category

Regard the vertical maps

f:AA,g:BB,h:CC

as objects of the arrow category

Arr(A)=A[1].

The two short exact rows say exactly that these three arrow-objects form a short exact sequence in Arr(A):

0fgh0.

Apply the kernel functor

ker:Arr(A)A.

Since ker is left exact, we obtain an exact sequence:

0kerfkergkerh.

Because f and h are isomorphisms,

kerf=0,kerh=0.

Hence

0kerg0,

so

kerg=0.

Therefore g is monic.

Dually, apply the cokernel functor

coker:Arr(A)A.

Since coker is right exact, we obtain an exact sequence:

cokerfcokergcokerh0.

Again, because f and h are isomorphisms,

cokerf=0,cokerh=0.

Hence

0cokerg0,

so

cokerg=0.

Therefore g is epic.

In an abelian category, a morphism is an isomorphism if and only if its kernel and cokernel are both zero. Hence g is an isomorphism.

Actually I have another proof, see 10.1

https://www.researchgate.net/publication/405327103_Morphism_Tests_and_the_Logic_of_Separation_Property_Axiomatization_by_Uniqueness_of_Hom-Restriction


5. A functorial comparison lemma

The reduction of the five lemma to the short five lemma rests on two small comparison facts.

Lemma: functorial comparison for cokernels and kernels

Let A be an abelian category.

Cokernel comparison

Suppose we have a commutative square

A1aA2euB1bB2

where e is epic and u is an isomorphism. Then the morphism induced by the cokernel functor

coker(a)coker(b)

is an isomorphism.

Kernel comparison

Dually, suppose we have a commutative square

A4aA5umB4bB5

where u is an isomorphism and m is monic. Then the morphism induced by the kernel functor

ker(a)ker(b)

is an isomorphism.

Proof

We prove the cokernel statement first.

The square says

ua=be.

Since u is an isomorphism, the cokernel of ua is canonically identified with the cokernel of a:

coker(a)coker(ua).

Concretely, if

qa:A2coker(a)

is the cokernel of a, then

qau1:B2coker(a)

is the cokernel of ua.

Since e is epic, be and b have the same cokernel. Indeed, a morphism

t:B2Z

satisfies

tbe=0

if and only if

tb=0,

because e is epic.

Thus

coker(be)coker(b).

Using ua=be, we obtain

coker(a)coker(ua)=coker(be)coker(b).

This is exactly the morphism induced by applying

coker:Arr(A)A

to the original square.

The kernel statement is dual.

The square says

ma=bu.

Since u is an isomorphism, the kernel of bu is canonically identified with the kernel of b:

ker(bu)ker(b).

Since m is monic, ma and a have the same kernel. Indeed, a morphism

s:TA4

satisfies

mas=0

if and only if

as=0.

Therefore

ker(a)ker(ma).

Using ma=bu, we obtain

ker(a)ker(ma)=ker(bu)ker(b).

This is exactly the morphism induced by applying

ker:Arr(A)A

to the original square.


6. Reducing the five lemma to the short five lemma

We now prove the strong form of the five lemma.

Suppose we have a commutative diagram with exact rows:

A1a1A2a2A3a3A4a4A5f1f2f3f4f5B1b1B2b2B3b3B4b4B5.

Assume

f1 is epic,f2 and f4 are isomorphisms,f5 is monic.

We prove that

f3:A3B3

is an isomorphism.

Step 1: cut A3 into a short exact sequence

Define

CA:=coker(a1),KA:=ker(a4).

Let

qA:A2CA

be the cokernel of a1, and let

kA:KAA4

be the kernel of a4.

Since the top row is exact at A2,

ker(a2)=im(a1).

Thus a2 kills a1, so a2 factors uniquely through qA:

iA:CAA3,iAqA=a2.

By the first isomorphism theorem,

CA=coker(a1)A2/im(a1)=A2/ker(a2)im(a2).

Hence iA is monic and

im(iA)=im(a2).

On the other hand, exactness at A4 gives

im(a3)=ker(a4)=KA.

Thus a3 factors uniquely through kA:

pA:A3KA,kApA=a3.

Since

im(a3)=KA,

the morphism pA is epic.

Because kA is monic,

ker(pA)=ker(a3).

Exactness at A3 gives

ker(a3)=im(a2)=im(iA).

Therefore we obtain a short exact sequence:

0CAiAA3pAKA0.

Equivalently,

0coker(a1)A3ker(a4)0.

Applying the same construction to the bottom row gives

CB:=coker(b1),KB:=ker(b4),

and a short exact sequence

0CBiBB3pBKB0.
Step 2: the original diagram induces a diagram of short exact sequences

The original five-term diagram induces

0CAiAA3pAKA0γf3δ0CBiBB3pBKB0.

Here

γ:CA=coker(a1)CB=coker(b1)

is obtained by applying the cokernel functor to the left square:

A1a1A2f1f2B1b1B2.

Similarly,

δ:KA=ker(a4)KB=ker(b4)

is obtained by applying the kernel functor to the right square:

A4a4A5f4f5B4b4B5.

The two squares in the short exact sequence diagram commute by functoriality and by the commutativity of the original diagram.

Step 3: identify the two outer vertical maps

By the functorial comparison lemma applied to the left square,

γ:coker(a1)coker(b1)

is an isomorphism, because

f1 is epicandf2 is an isomorphism.

By the dual part of the same lemma applied to the right square,

δ:ker(a4)ker(b4)

is an isomorphism, because

f4 is an isomorphismandf5 is monic.

Thus we have a commutative diagram of short exact sequences:

0coker(a1)A3ker(a4)0↓∼f3↓∼0coker(b1)B3ker(b4)0.

By the short five lemma, the middle vertical map

f3:A3B3

is an isomorphism.

This proves the five lemma.


7. Summary

The five lemma is not primarily a mysterious diagram chase. Its structure is:

A3 is an extension of ker(a4) by coker(a1).

Similarly,

B3 is an extension of ker(b4) by coker(b1).

The left-side hypotheses

f1 epic,f2 iso

ensure

coker(a1)coker(b1).

The right-side hypotheses

f4 iso,f5 monic

ensure

ker(a4)ker(b4).

So the middle objects A3 and B3 are extensions with isomorphic left and right parts. The short five lemma then says that the middle map between these extensions is also an isomorphism.

In one sentence:

five lemma = cut the middle term into a short exact sequence, then apply the short five lemma.

No comments:

Post a Comment

Popular Posts