In this tutorial, we describe how to work with modules in Macaulay2.
First, let's define a ring.
i1 : R = QQ[a..f]; |
i2 : m = matrix{{a,b,d,e},{b,c,e,f}}
o2 = | a b d e |
| b c e f |
2 4
o2 : Matrix R <--- R
|
Use standard notation for cokernels, images and kernels (coker, cokernel, image, ker, kernel).
i3 : M = coker m
o3 = cokernel | a b d e |
| b c e f |
2
o3 : R-module, quotient of R
|
i4 : N = image m
o4 = image | a b d e |
| b c e f |
2
o4 : R-module, submodule of R
|
i5 : K = kernel m
-- ker (2) called with OptionTable: OptionTable{SubringLimit => infinity}
-- ker (2) returned CacheFunction: -*a cache function*-
-- ker (2) called with Matrix: | a b d e |
-- | b c e f |
-- ker (2) returned Module: image {1} | cd-be 0 e2-df ce-bf |
-- {1} | -bd+ae e2-df 0 -be+af |
-- {1} | b2-ac -ce+bf -be+af 0 |
-- {1} | 0 cd-be bd-ae b2-ac |
assert( ker(map(R^2,R^{{-1}, {-1}, {-1}, {-1}},{{a,b,d,e}, {b,c,e,f}})) === (image(map(R^{{-1}, {-1}, {-1}, {-1}},R^{{-3}, {-3}, {-3}, {-3}},{{c*d-b*e,0,e^2-d*f,c*e-b*f}, {-b*d+a*e,e^2-d*f,0,-b*e+a*f}, {b^2-a*c,-c*e+b*f,-b*e+a*f,0}, {0,c*d-b*e,b*d-a*e,b^2-a*c}}))))
o5 = image {1} | cd-be 0 e2-df ce-bf |
{1} | -bd+ae e2-df 0 -be+af |
{1} | b2-ac -ce+bf -be+af 0 |
{1} | 0 cd-be bd-ae b2-ac |
4
o5 : R-module, submodule of R
|
Given a module, one can find its presentation matrix.
i6 : presentation M -- this is just the original matrix
o6 = | a b d e |
| b c e f |
2 4
o6 : Matrix R <--- R
|
i7 : presentation N -- this one requires computation
o7 = {1} | cd-be 0 e2-df ce-bf |
{1} | -bd+ae e2-df 0 -be+af |
{1} | b2-ac -ce+bf -be+af 0 |
{1} | 0 cd-be bd-ae b2-ac |
4 4
o7 : Matrix R <--- R
|
To define a submodule $IN$ of a module $N$, where $I$ is an ideal, use
i8 : ideal(a,b)*N
o8 = image | a2 ab ad ae ab b2 bd be |
| ab ac ae af b2 bc be bf |
2
o8 : R-module, submodule of R
|
i9 : a*N + b*N
o9 = image | a2 ab ad ae ab b2 bd be |
| ab ac ae af b2 bc be bf |
2
o9 : R-module, submodule of R
|
In order to define a submodule of N generated by some elements of N, one way is the following.
i10 : N0 = image (a**N_{1}|N_{2}-N_{3})
o10 = image | ab d-e |
| ac e-f |
2
o10 : R-module, submodule of R
|
To understand what Macaulay2 is doing here, let's break this down. N_{i} defines a matrix $R^1 \to N$, which maps 1 to the i th generator of N. (See Section XX below for more information about module homomorphisms).
i11 : N_{1}
o11 = {1} | 0 |
{1} | 1 |
{1} | 0 |
{1} | 0 |
o11 : Matrix
|
One could use a*N_{1}, but it turns out that a ** N_{1} works better:
i12 : a ** N_{1}
o12 = {1} | 0 |
{1} | a |
{1} | 0 |
{1} | 0 |
o12 : Matrix
|
Next, remember that the vertical bar concatenates matrices.
i13 : a ** N_{1} | N_{2}-N_{3}
o13 = {1} | 0 0 |
{1} | a 0 |
{1} | 0 1 |
{1} | 0 -1 |
o13 : Matrix
|
Now take the image of this matrix
i14 : N0 = image(a ** N_{1} | N_{2}-N_{3})
o14 = image | ab d-e |
| ac e-f |
2
o14 : R-module, submodule of R
|
The main advantage for using ** rather than * is that ** preservers homogeneity if possible.
i15 : isHomogeneous N0 o15 = true |
Quotients are defined using standard mathematical notation.
i16 : Nbar = N/N0
o16 = subquotient (| a b d e |, | ab d-e |)
| b c e f | | ac e-f |
2
o16 : R-module, subquotient of R
|
Notice that this returns a subquotient module. We treat these later. Ideals and modules are treated differently in Macaulay2 (and in commutative algebra in general). For example, asking for the dimension of an ideal I in a ring R gives the dimension of the quotient R/I, but the dimension of the module I gives a potentially very different answer. Use ideal and module to move between the two.
i17 : I = ideal(a^2, a*b, c^2)
2 2
o17 = ideal (a , a*b, c )
o17 : Ideal of R
|
i18 : J = module I
o18 = image | a2 ab c2 |
1
o18 : R-module, submodule of R
|
i19 : I == ideal J o19 = true |
i20 : codim I o20 = 2 |
i21 : codim J o21 = 0 |
Create a free resolution of an ideal (or module) using res.
i22 : C = res I
1 3 3 1
o22 = R <-- R <-- R <-- R <-- 0
0 1 2 3 4
o22 : ChainComplex
|
View the differential
i23 : C.dd
1 3
o23 = 0 : R <---------------- R : 1
| a2 ab c2 |
3 3
1 : R <---------------------- R : 2
{2} | -b -c2 0 |
{2} | a 0 -c2 |
{2} | 0 a2 ab |
3 1
2 : R <-------------- R : 3
{3} | c2 |
{4} | -b |
{4} | a |
1
3 : R <----- 0 : 4
0
o23 : ChainComplexMap
|
The (graded) betti numbers
i24 : betti C
0 1 2 3
o24 = total: 1 3 3 1
0: 1 . . .
1: . 3 1 .
2: . . 2 1
o24 : BettiTally
|
Use help (betti,GradedModule) for a detailed description of what this display means. Basically, it says that I has three generators of degree 2, one syzygy of degree 3, 2 syzygies of degree 4, and one second syzygy of degree 5. The free resolution of a module that is not a cokernel:
i25 : C = res Nbar
4 6 2
o25 = R <-- R <-- R <-- 0
0 1 2 3
o25 : ChainComplex
|
i26 : betti C
0 1 2
o26 = total: 4 6 2
0: . 1 .
1: 4 1 .
2: . 4 2
o26 : BettiTally
|
i27 : C.dd
4 6
o27 = 0 : R <----------------------------------------------------------- R : 1
{1} | 0 0 ce-bf -cd+be+ce-bf e2-df 0 |
{1} | 0 a -be bd-be 0 e2-df |
{1} | -1 0 b2-ac 0 bd-ae-be+af cd-be-ce+bf |
{1} | 1 0 0 0 0 0 |
6 2
1 : R <--------------------- R : 2
{1} | 0 0 |
{2} | e2-df 0 |
{3} | -d+e e-f |
{3} | -e f |
{3} | b -c |
{3} | -a b |
2
2 : R <----- 0 : 3
0
o27 : ChainComplexMap
|
Here is a problem to experiment with. What different betti diagrams are possible with an ideal generated by 3 homogeneous quadric polynomials, in a polynomial ring in any number of variables? Here is one to get you started.
i28 : R = QQ[a..h]; |
i29 : J = ideal(a*c+b*d,a*e+b*f,a*g+b*h) o29 = ideal (a*c + b*d, a*e + b*f, a*g + b*h) o29 : Ideal of R |
i30 : betti res J
0 1 2 3
o30 = total: 1 3 4 2
0: 1 . . .
1: . 3 . .
2: . . 4 2
o30 : BettiTally
|
After that, try ideals generated by 4 quadrics.
Recall that the module N/N0 above displayed as something called a subquotient module. As Macaulay2 often returns such objects, it is useful to understand and be able to manipulate them. P The most common modules are quotients of free modules, or submodules of free modules. A useful generalization, which covers both of these types, are subquotients: submodules of quotients of free modules. P A subquotient module is determined by two matrices $f : R^m \to R^n$ and $g : R^p \to R^n$. The subquotient module with generators f, relations g is by definition the module M = (image f) + (image g) / (image g). Thus, if f is the identity map, M = coker g, and if g = 0, then M = image f.
i31 : use ring M o31 = QQ[a..f] o31 : PolynomialRing |
i32 : M
o32 = cokernel | a b d e |
| b c e f |
2
o32 : QQ[a..f]-module, quotient of (QQ[a..f])
|
i33 : N = a*M
o33 = subquotient (| a 0 |, | a b d e |)
| 0 a | | b c e f |
2
o33 : QQ[a..f]-module, subquotient of (QQ[a..f])
|
i34 : M/N
o34 = cokernel | a 0 a b d e |
| 0 a b c e f |
2
o34 : QQ[a..f]-module, quotient of (QQ[a..f])
|
The two matrices f and g mentioned above are recovered using the routines: generators, relations.
i35 : generators N
o35 = | a 0 |
| 0 a |
2 2
o35 : Matrix (QQ[a..f]) <--- (QQ[a..f])
|
i36 : relations N
o36 = | a b d e |
| b c e f |
2 4
o36 : Matrix (QQ[a..f]) <--- (QQ[a..f])
|
It is often necessary to find a presentation matrix for such modules.
i37 : presentation N
o37 = {1} | e d b a |
{1} | f e c b |
2 4
o37 : Matrix (QQ[a..f]) <--- (QQ[a..f])
|
Often the given representation of a module is not very efficient. Use trim to keep the module as a subquotient of the same ambient free module, but change the generators and relations to be minimal, or in the nonlocal or non-graded case, at least more efficient.
i38 : trim N
o38 = subquotient (| a 0 |, | e d b a |)
| 0 a | | f e c b |
2
o38 : QQ[a..f]-module, subquotient of (QQ[a..f])
|
Use minimalPresentation to also allow the ambient free module to be improved. This returns a quotient of a free module, but in the future might not do that.
i39 : minimalPresentation N
o39 = cokernel {1} | e d b a |
{1} | f e c b |
2
o39 : QQ[a..f]-module, quotient of (QQ[a..f])
|
prune is a synonym for minimalPresentation N
i40 : prune N
o40 = cokernel {1} | e d b a |
{1} | f e c b |
2
o40 : QQ[a..f]-module, quotient of (QQ[a..f])
|
Given a subquotient module N, there are several useful modules associated to N. The free module of which N is a subquotient is obtained using ambient.
i41 : ambient N
2
o41 = (QQ[a..f])
o41 : QQ[a..f]-module, free
|
This is the same as the target of either the generator or relation matrix.
i42 : ambient N == target generators N o42 = true |
i43 : ambient N == target relations N o43 = true |
N is a submodule of a quotient module $R^n/image(g)$. The routine super returns this quotient module
i44 : super N
o44 = cokernel | a b d e |
| b c e f |
2
o44 : QQ[a..f]-module, quotient of (QQ[a..f])
|
This is the same as
i45 : super N == coker relations N o45 = true |
The cover of N is basically the source of the matrix of generators.
i46 : cover N
2
o46 = (QQ[a..f])
o46 : QQ[a..f]-module, free, degrees {2:1}
|
i47 : cover N == source generators N o47 = true |
A homomorphism $f : M \to N$ is represented as a matrix from the generators of M to the generators of N.
i48 : A = QQ[x,y]/(y^2-x^3) o48 = A o48 : QuotientRing |
i49 : M = module ideal(x,y)
o49 = image | x y |
1
o49 : A-module, submodule of A
|
One homomorphism $F : M \to A$ is $x \mapsto y, y \mapsto x^2$ (multiplication by y/x) We write this as:
i50 : F = map(A^1,M,matrix{{y,x^2}})
o50 = | y x2 |
o50 : Matrix
|
Notice that as is usual in Macaulay2, the target comes before the source.
i51 : source F == M o51 = true |
i52 : target F == A^1 o52 = true |
i53 : matrix F
o53 = | y x2 |
1 2
o53 : Matrix A <--- A
|
The image of F lies in the submodule M of $A^1$. To obtain the map $M \to M$, we use //. But first we need the inclusion map of M into $A^1$: Later we explain this, but for now, we just write down this map:
i54 : inducedMap(A^1,M) o54 = | x y | o54 : Matrix |
Now we use // to lift $F : M \to A$ along $M \to A^1$, to get $M \to M$:
i55 : G = F // inducedMap(A^1,M)
o55 = {1} | 0 x |
{1} | 1 0 |
o55 : Matrix
|
i56 : source G
o56 = image | x y |
1
o56 : A-module, submodule of A
|
i57 : target G
o57 = image | x y |
1
o57 : A-module, submodule of A
|
G is now a map from $M \to M$.
i58 : isWellDefined G o58 = true |
Let's start with a module M, and a submodule N.
i59 : R = QQ[x,y,z,w] o59 = R o59 : PolynomialRing |
i60 : M = ideal(x,y,z)/ideal(x^2,y^2,z*w)
o60 = subquotient (| x y z |, | x2 y2 zw |)
1
o60 : R-module, subquotient of R
|
i61 : N = z*M
o61 = subquotient (| xz yz z2 |, | x2 y2 zw |)
1
o61 : R-module, subquotient of R
|
i62 : M/N
o62 = subquotient (| x y z |, | xz yz z2 x2 y2 zw |)
1
o62 : R-module, subquotient of R
|
If two modules have the same ambient free module, then there is often a canonical map between them. Some modules having the same ambient free module:
i63 : M
o63 = subquotient (| x y z |, | x2 y2 zw |)
1
o63 : R-module, subquotient of R
|
i64 : ambient M
1
o64 = R
o64 : R-module, free
|
i65 : N = z*M
o65 = subquotient (| xz yz z2 |, | x2 y2 zw |)
1
o65 : R-module, subquotient of R
|
i66 : ambient(M/N)
1
o66 = R
o66 : R-module, free
|
i67 : super M
o67 = cokernel | x2 y2 zw |
1
o67 : R-module, quotient of R
|
i68 : super N
o68 = cokernel | x2 y2 zw |
1
o68 : R-module, quotient of R
|
i69 : image generators M
o69 = image | x y z |
1
o69 : R-module, submodule of R
|
If two modules M and N have the same ambient module $R^n$, then inducedMap(M,N) makes the canonical map $N \to M$ between them, if one exists. If a map doesn't exist, the returned map might not be a homomorphism.
i70 : inducedMap(M,M) == id_M o70 = true |
i71 : inducedMap(super M,M) == map(super id_M) -- the map $(P+Q)/Q \to R^n/Q$, where $M=(P+Q)/Q$. o71 = true |
i72 : inducedMap(super M,ambient M) -- the quotient map $R^n \to R^n/Q$ o72 = | 1 | o72 : Matrix |
i73 : inducedMap(M,N) -- the inclusion map
o73 = {1} | z 0 0 |
{1} | 0 z 0 |
{1} | 0 0 z |
o73 : Matrix
|
The projection map $M \to M/N$
i74 : inducedMap(M/N,M) -- the projection map
o74 = {1} | 1 0 0 |
{1} | 0 1 0 |
{1} | 0 0 1 |
o74 : Matrix
|
The projection map $N \to M/N$, which is the zero map
i75 : inducedMap(M/N,N) -- the zero map o75 = 0 o75 : Matrix |
Not all such maps can be defined. The functions 'inducedMap' normally checks that the result is a well-defined homomorphism. The option 'Verify' controls that behavior.
i76 : inducedMap(M,M/N,Verify => false)
o76 = {1} | 1 0 0 |
{1} | 0 1 0 |
{1} | 0 0 1 |
o76 : Matrix
|
i77 : inducedMap(M/N,x*M)
o77 = {1} | 0 y 0 |
{1} | 0 0 0 |
{1} | 0 0 0 |
o77 : Matrix
|
i78 : inducedMap(M/N,M) * inducedMap(M,x*M) == inducedMap(M/N,x*M) o78 = true |
Before doing interesting homomorphisms, let's see how to write down some canonical homomorphisms associated to M. exercises: 1. isomorphism theorems. Given submodules M and N of a module P, (a) find $(M+N)/M$ (b) find $N/(M \cap N)$ (c) find in Macaulay2, an isomorphism between them.
2. Given a homomorphism $M \to A$. Suppose that the image lies in M (M is a submodule of $A^1$). Find the map $M \to M$.
i79 : A = QQ[x,y,Degrees=>{2,3}]/(y^2-x^3)
o79 = A
o79 : QuotientRing
|
i80 : M = module ideal(x,y)
o80 = image | x y |
1
o80 : A-module, submodule of A
|
i81 : H = Hom(M,M)
-- ker (3) called with OptionTable: OptionTable{SubringLimit => infinity}
-- ker (3) returned CacheFunction: -*a cache function*-
-- ker (3) called with Matrix: {-3} | 0 0 -x 0 |
-- {-2} | x y 0 -x |
-- {-4} | 0 0 0 0 |
-- {-3} | y x2 -x -y |
-- ker (3) returned Module: image {-2} | x y x2 0 |
-- {-3} | y x2 xy 0 |
assert( ker(map(image(map(A^{{5}, {6}},A^{{3}, {2}, {4}, {3}},{{x,y,0,0}, {0,0,x,y}})),image(map(A^{{2}, {3}},A^{{0}, {-1}, {1}, {0}},{{x,y,0,0}, {0,0,x,y}})),{{0,0,-x,0}, {x,y,0,-x}, {0,0,0,0}, {y,x^2,-x,-y}})) === (image(map(A^{{2}, {3}},A^{{0}, {-1}, {-2}, {-3}},{{x,y,x^2,0}, {y,x^2,x*y,0}}))))
o81 = image {-2} | x y |
{-3} | y x2 |
2
o81 : A-module, submodule of A
|
The elements of H correspond to homomorphisms $M \to A$. The homomorphism associated to elements of H may be obtained using the routine homomorphism.
i82 : F = homomorphism(H_{0})
o82 = {2} | 1 0 |
{3} | 0 1 |
o82 : Matrix
|
i83 : G = homomorphism(H_{1})
o83 = {2} | 0 x |
{3} | 1 0 |
o83 : Matrix
|
i84 : source F == M o84 = true |
i85 : target F == M o85 = true |
i86 : ker F
-- ker (4) called with OptionTable: OptionTable{SubringLimit => infinity}
-- ker (4) returned CacheFunction: -*a cache function*-
-- ker (4) called with Matrix: {2} | 1 0 |
-- {3} | 0 1 |
-- ker (4) returned Module: image 0
assert( ker(map(image(map(A^1,A^{{-2}, {-3}},{{x,y}})),image(map(A^1,A^{{-2}, {-3}},{{x,y}})),{{1,0}, {0,1}})) === (image(map(A^1,A^{{-5}, {-6}},0))))
o86 = image 0
1
o86 : A-module, submodule of A
|
i87 : coker F
o87 = subquotient (| x y |, | x y |)
1
o87 : A-module, subquotient of A
|
i88 : m = matrix{{x,y},{y,x}}
o88 = | x y |
| y x |
2 2
o88 : Matrix A <--- A
|
i89 : Hom(m,A^2)
-- ker (5) called with OptionTable: OptionTable{SubringLimit => infinity}
-- ker (5) returned CacheFunction: -*a cache function*-
-- ker (5) called with Matrix: 0
-- 4
-- ker (5) returned Module: A
assert( ker(map(A^0,A^{{3}, {3}, {3}, {3}},0)) === (A^{{3}, {3}, {3}, {3}}))
-- ker (6) called with OptionTable: OptionTable{SubringLimit => infinity}
-- ker (6) returned CacheFunction: -*a cache function*-
-- ker (6) called with Matrix: 0
-- 4
-- ker (6) returned Module: A
assert( ker(map(A^0,A^4,0)) === (A^4))
o89 = {-3} | x 0 y 0 |
{-3} | 0 x 0 y |
{-3} | y 0 x 0 |
{-3} | 0 y 0 x |
4 4
o89 : Matrix A <--- A
|
i90 : Hom(A^2,m)
-- ker (7) called with OptionTable: OptionTable{SubringLimit => infinity}
-- ker (7) returned CacheFunction: -*a cache function*-
-- ker (7) called with Matrix: 0
-- 4
-- ker (7) returned Module: A
assert( ker(map(A^0,A^4,0)) === (A^4))
-- ker (8) called with OptionTable: OptionTable{SubringLimit => infinity}
-- ker (8) returned CacheFunction: -*a cache function*-
-- ker (8) called with Matrix: 0
-- 4
-- ker (8) returned Module: A
assert( ker(map(A^0,A^{{-3}, {-3}, {-3}, {-3}},0)) === (A^{{-3}, {-3}, {-3}, {-3}}))
o90 = | x y 0 0 |
| y x 0 0 |
| 0 0 x y |
| 0 0 y x |
4 4
o90 : Matrix A <--- A
|
In Macaulay2, ** denotes the tensor product operator.
i91 : m ** m
o91 = | x2 xy xy y2 |
| xy x2 y2 xy |
| xy y2 x2 xy |
| y2 xy xy x2 |
4 4
o91 : Matrix A <--- A
|
i92 : (coker m) ** (coker m)
o92 = cokernel | x y 0 0 x y 0 0 |
| y x 0 0 0 0 x y |
| 0 0 x y y x 0 0 |
| 0 0 y x 0 0 y x |
4
o92 : A-module, quotient of A
|
Notice that tensor products of matrices and of modules are very different.
i93 : M = coker m
o93 = cokernel | x y |
| y x |
2
o93 : A-module, quotient of A
|
i94 : M2 = prune(M ** M)
o94 = cokernel | 0 -x y x x 0 |
| -x 0 x y 0 x |
| x 0 0 0 y 0 |
| 0 x 0 0 0 y |
4
o94 : A-module, quotient of A
|
i95 : A = QQ[a,b,c] o95 = A o95 : PolynomialRing |
i96 : A ** A o96 = QQ[2:a..c] o96 : PolynomialRing |
Oops! Macaulay2 doesn't know what a should be!
i97 : B = oo o97 = B o97 : PolynomialRing |
i98 : a == B_3 o98 = true |
i99 : a == B_0 o99 = false |
To remedy this, one can give the variables as an option to tensor.
i100 : tensor(A,A,Variables=>{a,b,c,d,e,f})
o100 = QQ[a..f]
o100 : PolynomialRing
|