Professional Documents
Culture Documents
10.4.11. Let {e1 , e2 } be a basis in V = R2 . Show that the element e1 ⊗ e2 + e2 ⊗ e1 is not (cannot be written
as) as a simple tensor in V ⊗R V .
Solution. Every simple
tensor in V ⊗ V has form u ⊗ v for some u = a1 e1 + a2 e2 ∈ V and v = b1 e1 + b2 e2 ∈ V ,
and in the basis e1 ⊗ e1 , e1 ⊗ e2 , e2 ⊗ e1 , e2 ⊗ e2 is written as
a 1 b1 e 1 ⊗ e 1 + a 1 b2 e 1 ⊗ e 2 + a 2 b1 e 2 ⊗ e 1 + a 2 b2 e 2 ⊗ e 2 .
Thus, the matrix of the coordinates of u ⊗ v in this basis is aa12 bb11 aa12 bb22 . This matrix is degenerate (has zero
determinant), whereas the matrix (01 10) of the coordinates of e1 ⊗ e2 + e2 ⊗ e1 is nondegenerate. Hence, this
tensor is not simple.
Now let w be a simple tensor from R/I ⊗ R/J, w = (r mod I) ⊗ (r′ mod J). Then
ψ(ϕ(w)) = ψ(rr′ mod(I+J)) = (1 mod I)⊗(rr′ mod J) = r(1 mod I)⊗(r′ mod J) = (r mod I)⊗(r′ mod J) = w.
Another solution. We have (R/I) ⊗ (R/J) ∼= (R/J)/(I(R/J)). The preimage in R of the submodule I(R/J)
of R/J is I + J, so by the 3rd isomorphism theorem, (R/J)/(I(R/J)) ∼
= R/(I + J).
1
10.4.21. Let I and J be ideals in R.
(a) Show that there exists a surjective homomorphism ϕ: I ⊗R J −→ IJ with ϕ(i ⊗ j) = ij for all i ∈ I,
j ∈ J.
Solution. The mapping (i, j) 7→ ij is bilinear on I ×P
J, thus extends to a homomorphism ϕ: I ⊗ J −→ IJ.
n
To show
Pn that ϕ is surjective, let a ∈ IJ; then a = k=1 ik jk for some i1 , . . . , in ∈ I, j1 , . . . , jn ∈ J. Put
u = k=1 ik ⊗ jk ∈ I ⊗ J, then a = ϕ(u).
(b) Give an example to show that ϕ may not be injective.
Solution. Take R = Z4 and I = J = (2) = {0, 2}; then IJ = 0. I claim that I ⊗ J 6= 0, so ϕ cannot be
injective. Indeed, define a mapping I × J −→ Z2 by (0, 0), (0, 2), (2, 0) 7→ 0 and (2, 2) 7→ 1; it is easy to check
that this mapping is bilinear and so, defines a homomorphism onto Z2 .
Or: I and J are, actually, Z4 /(2) = Z2 -modules, both isomorphic to Z2 , and I ⊗Z4 J = I ⊗Z2 J ∼ = Z2 .
Another solution. Take R = Z[x, y] and I = (x, y). For tensor w = x ⊗ y − y ⊗ x we have ϕ(w) = 0; I claim
that w 6= 0 in I ⊗R I. Indeed, the bilinear mapping β: I × I −→ Z defined by β(a1 x + a2 y + a3 xy + · · · , b1 x +
b2 y + b3 xy + · · ·) = a1 b2 induces a homomorphism ψ: I ⊗ I −→ Z, and we have ψ(w) = 1 6= 0, so w 6= 0.
(α1 ⊗ xn1 )(α2 ⊗ xn2 ) = (α1 α2 ) ⊗ xn1 +n2 7→ α1 α2 xn1 +n2 = (α1 xn1 )(α2 xn2 ),
so it is an ring isomorphism.
(ϕ1 + ψ1 ) ⊗1 ϕ2 (u1 ⊗ u2 ) = (ϕ1 + ψ1 )(u1 ) ⊗ ϕ2 (u2 ) = ϕ1 (u1 ) ⊗ ϕ2 (u2 ) + ψ1 (u1 ) ⊗ ϕ2 (u2 )
= (ϕ1 ⊗1 ϕ2 )(u1 ⊗ u2 ) + (ψ1 ⊗1 ϕ2 )(u1 ⊗ u2 ),
Bonus:
2
B1. Let A be a unital R-algebra and M be an A-module. By restricting scalars, consider M as an R-module
(with au = (a1A )u). Prove that the R-module homomorphism ϕ: M −→ A ⊗R M defined by ϕ(u) = 1A ⊗ u
is injective and that 1A ⊗ M = ϕ(M ) is a direct summand in A ⊗R M .
Solution. To prove both assertions it suffices to find an R-module homomorphism π: A ⊗R M −→ M such
that π ◦ϕ = IdM : if such π exists, then ϕ is injective, and π is a (left) splitting homomorphism of the short
exact sequence
ϕ
0 −→ M −→ A ⊗R M −→ A ⊗R M /ϕ(M ) −→ 0.
But this is easy: simply put π(α ⊗ u) = αu, α ∈ A, u ∈ M . (Well, to prove that this π is well defined, we
need to start with a bilinear mapping A × M −→ M , etc.)