L | W | |
L | 2 1 | 0 0 |
W | 0 0 | 1 2 |
Tuesday, June 25, 2013
Maxmin Strategy in Battle of the Sexes
Kevin Leyton-Brown and Yoav Shoham Essentials of Game Theory, page 15:
MathJax TeX Test Page
Given the following payoff table for a typical Battle of the Sexes game,
calculate the maxmin strategies and values for each player.
Proof: Player 2 must minimizeu1(s)=2s1(L)s2(L)+s1(W)s2(W)=2s1(L)s2(L)+(1−s1(L))(1−s2(L))=s2(L)(3s1(L)−1)+1−s1(L)When s1(L)>1/3 and 3s1(L)−1 is positive, player 2 minimizes by playing s2(L)=0 and now the expression collapses to 1−s1(L) so that u1(s)<2/3. If s1(L)<1/3 and 3s1(L)−1 is negative, player 2 minimizes with s2(L)=1 and the expression collapses to 2s1(L) showing u1(s)<2/3. When s1(L)=1/3 then the expression invariably maximizes as u1(s)=2/3. This showss1(L)=1/3s1(W)=2/3is player 1's maxmin strategy with maxmin value 2/3. Symmetrically, similar holds for player 2. ◻
Thursday, June 20, 2013
Tensor Products and Extensions of Polynomial Ring Quotients (10.4.26)
Dummit and Foote Abstract Algebra, section 10.4, exercise 26:
MathJax TeX Test Page
Let S and R⊆S be commutative rings with 1R=1S, and let I⊆R[x1,...,xn] be an ideal. ProveS⊗R(R[x1,...,xn]/I)≅S[x1,...,xn]/IS[x1,...,xn]as S-algebras.
Proof: We must prove there is an isomorphism of rings and of S-modules between the two. Define a mappingφ : S \times (R[x_1,...,x_n]/I) → S[x_1,...,x_n]/IS[x_1,...,x_n](s,\overline{r(x)}) \mapsto \overline{\overline{sr(x)}}This map is well defined since(s,\overline{r(x)}) = (s,\overline{r(x)+i}) ⇒φ(s,\overline{r(x)}) = φ(s,\overline{r(x)+i}) - is⇒\overline{\overline{φ(s,\overline{r(x)})}} = \overline{\overline{φ(s,\overline{r(x)+i})}}This map is evidently R-balanced, and thus induces a homomorphism \Phi between the targets. For each s(x)= \overline{\overline{\sum s_ix^i}}∈S[x_1,...,x_n]/IS[x_1,...,x_n] we can choose a preimage \sum s_i \otimes \overline{x^i}∈S \otimes (R[x_1,...,x_n]/I) so that this homomorphism is surjective. Introduce the homomorphismS[x_1,...,x_n] → S \otimes_R (R[x_1,...,x_n]/I)\sum s_ix^i \mapsto \sum s_i \otimes \overline{x^i}which is seen to factor through IS[x_1,...,x_n] so that there is an inverse to \Phi making it injective, as well.
Now we may demonstrates\Phi(\sum s_i \otimes \overline{r_i(x)})=\sum \overline{\overline{ss_ir_i(x)}} = \Phi(s\sum s_i \otimes \overline{r_i(x)})so that \Phi is an S-module isomorphism. Finally, to show the ring isomorphism, we have\Phi(\sum s_i \otimes \overline{r_i(x)})\Phi(\sum s_j \otimes \overline{r_j(x)})=\sum \sum \overline{\overline{s_is_jr_i(x)r_j(x)}}=\Phi(\sum s_i \otimes \overline{r_i(x)} \cdot \sum s_j \otimes \overline{r_j(x)})~\square
Proof: We must prove there is an isomorphism of rings and of S-modules between the two. Define a mappingφ : S \times (R[x_1,...,x_n]/I) → S[x_1,...,x_n]/IS[x_1,...,x_n](s,\overline{r(x)}) \mapsto \overline{\overline{sr(x)}}This map is well defined since(s,\overline{r(x)}) = (s,\overline{r(x)+i}) ⇒φ(s,\overline{r(x)}) = φ(s,\overline{r(x)+i}) - is⇒\overline{\overline{φ(s,\overline{r(x)})}} = \overline{\overline{φ(s,\overline{r(x)+i})}}This map is evidently R-balanced, and thus induces a homomorphism \Phi between the targets. For each s(x)= \overline{\overline{\sum s_ix^i}}∈S[x_1,...,x_n]/IS[x_1,...,x_n] we can choose a preimage \sum s_i \otimes \overline{x^i}∈S \otimes (R[x_1,...,x_n]/I) so that this homomorphism is surjective. Introduce the homomorphismS[x_1,...,x_n] → S \otimes_R (R[x_1,...,x_n]/I)\sum s_ix^i \mapsto \sum s_i \otimes \overline{x^i}which is seen to factor through IS[x_1,...,x_n] so that there is an inverse to \Phi making it injective, as well.
Now we may demonstrates\Phi(\sum s_i \otimes \overline{r_i(x)})=\sum \overline{\overline{ss_ir_i(x)}} = \Phi(s\sum s_i \otimes \overline{r_i(x)})so that \Phi is an S-module isomorphism. Finally, to show the ring isomorphism, we have\Phi(\sum s_i \otimes \overline{r_i(x)})\Phi(\sum s_j \otimes \overline{r_j(x)})=\sum \sum \overline{\overline{s_is_jr_i(x)r_j(x)}}=\Phi(\sum s_i \otimes \overline{r_i(x)} \cdot \sum s_j \otimes \overline{r_j(x)})~\square
Wednesday, June 19, 2013
Nonsimple Tensors (10.4.20)
Dummit and Foote Abstract Algebra, section 10.4, exercise 20:
MathJax TeX Test Page
Let R=\mathbb{Z}[x] and let I=(2,x). Show that the element 2 \otimes 2 + x \otimes x in I \otimes_R I is not a simple tensor.
Proof: Define a mapI \times I → R(i,j) \mapsto ijThis is clearly seen to be R-balanced, and so induces a homomorphism \Phi on I \otimes_R I. Assuming 2 \otimes 2 + x \otimes x = a \otimes b, then also\Phi(2 \otimes 2 + x \otimes x)=\Phi(2 \otimes 2) + \Phi(x \otimes x)=x^2+4=\Phi(a \otimes b)=abSince x^2+4 is a monomial quadratic with no roots, it does not factor in R, and thus either a or b is \pm 1 and now a \otimes b∉I \otimes I.~\square
Proof: Define a mapI \times I → R(i,j) \mapsto ijThis is clearly seen to be R-balanced, and so induces a homomorphism \Phi on I \otimes_R I. Assuming 2 \otimes 2 + x \otimes x = a \otimes b, then also\Phi(2 \otimes 2 + x \otimes x)=\Phi(2 \otimes 2) + \Phi(x \otimes x)=x^2+4=\Phi(a \otimes b)=abSince x^2+4 is a monomial quadratic with no roots, it does not factor in R, and thus either a or b is \pm 1 and now a \otimes b∉I \otimes I.~\square
Tuesday, June 18, 2013
Linear Associativity and Tensor Products (10.4.12)
Dummit and Foote Abstract Algebra, section 10.4, exercise 12:
MathJax TeX Test Page
Let V be a vector space over a field F and let v,v'∈V be nonzero. Prove v \otimes_F v' = v' \otimes_F v ⇔ v=av' for some a∈F.
Proof: (\Leftarrow)v' \otimes v = v' \otimes av' = av' \otimes v' = v \otimes v'(⇒) We may assume V has a basis \{e_i\}_{i∈I}.
Let v=\sum_{j∈J} f_je_j and v'=\sum_{j∈J} f_j'e_j for appropriate, finite J \subseteq I. Let W = \sum_{j∈J}Fe_j, so that W≅F^n for some finite n.
Now, v \otimes v'=v \otimes \sum_{j∈J} f_j'e_j = \sum_{j∈J} v \otimes f_j'e_j = \sum_{j∈J} f_j'v \otimes e_j, and similarly v' \otimes v = \sum_{j∈J} f_jv' \otimes e_j. We have v \otimes v',v' \otimes v ∈ W \otimes W≅F^n \otimes F^n and by 10.4.10(a) we have f_j'v=f_jv' for all j∈J and since v,v'≠0 we may say v=(f_j'^{-1}f_j)v'.
Proof: (\Leftarrow)v' \otimes v = v' \otimes av' = av' \otimes v' = v \otimes v'(⇒) We may assume V has a basis \{e_i\}_{i∈I}.
Let v=\sum_{j∈J} f_je_j and v'=\sum_{j∈J} f_j'e_j for appropriate, finite J \subseteq I. Let W = \sum_{j∈J}Fe_j, so that W≅F^n for some finite n.
Now, v \otimes v'=v \otimes \sum_{j∈J} f_j'e_j = \sum_{j∈J} v \otimes f_j'e_j = \sum_{j∈J} f_j'v \otimes e_j, and similarly v' \otimes v = \sum_{j∈J} f_jv' \otimes e_j. We have v \otimes v',v' \otimes v ∈ W \otimes W≅F^n \otimes F^n and by 10.4.10(a) we have f_j'v=f_jv' for all j∈J and since v,v'≠0 we may say v=(f_j'^{-1}f_j)v'.
Wednesday, June 12, 2013
Element Collapsing Between Tensor Products (10.4.2)
Dummit and Foote Abstract Algebra, section 10.4, exercise 2:
MathJax TeX Test Page
Show 2 \otimes 1=0∈\mathbb{Z} \otimes_\mathbb{Z} \mathbb{Z}/2\mathbb{Z} but 2 \otimes 1≠0∈\mathbb{2Z} \otimes_\mathbb{Z} \mathbb{Z}/2\mathbb{Z}.
Proof: For the first, notice 2 \otimes 1 = 1 \cdot 2 \otimes 1 = 1 \otimes 2 \cdot 1 = 1 \otimes 2 = 1 \otimes 0 = 0.
Now, if 2 \otimes 1 is zero in the latter tensor product, then by natural extension as a \mathbb{Z}-module, for an arbitrary element we have 2a \otimes b = (ab)(2 \otimes 1) = 0 so that \mathbb{2Z} \otimes_\mathbb{Z} \mathbb{Z}/2\mathbb{Z} is zero. However, we have\mathbb{2Z} \otimes_\mathbb{Z} \mathbb{Z}/2\mathbb{Z} ≅ \mathbb{Z}/2\mathbb{Z} \otimes_\mathbb{Z} \mathbb{2Z} ≅ \mathbb{2Z}/((\mathbb{2Z})(\mathbb{2Z})) ≅ \mathbb{2Z}/\mathbb{4Z} \not ≅ 0~\square
Proof: For the first, notice 2 \otimes 1 = 1 \cdot 2 \otimes 1 = 1 \otimes 2 \cdot 1 = 1 \otimes 2 = 1 \otimes 0 = 0.
Now, if 2 \otimes 1 is zero in the latter tensor product, then by natural extension as a \mathbb{Z}-module, for an arbitrary element we have 2a \otimes b = (ab)(2 \otimes 1) = 0 so that \mathbb{2Z} \otimes_\mathbb{Z} \mathbb{Z}/2\mathbb{Z} is zero. However, we have\mathbb{2Z} \otimes_\mathbb{Z} \mathbb{Z}/2\mathbb{Z} ≅ \mathbb{Z}/2\mathbb{Z} \otimes_\mathbb{Z} \mathbb{2Z} ≅ \mathbb{2Z}/((\mathbb{2Z})(\mathbb{2Z})) ≅ \mathbb{2Z}/\mathbb{4Z} \not ≅ 0~\square
Tuesday, June 11, 2013
Universal Multilinear Object (10.4 Corollary 16)
Dummit and Foote Abstract Algebra, section 10.4, corollary 16:
MathJax TeX Test Page
Show that if R is a commutative ring, A_i is an R-module, L is an abelian group, and φ : A_1 \times ... \times A_n → L is a multilinear map, then there exists a unique homomorphism of R-modules \Phi : A_1 \otimes ... \otimes A_n → L such that φ = \Phi \circ \iota, where \iota is the natural map from A_1 \times ... \times A_n to A_n \otimes ... \otimes A_n. Conversely, show that if there is a homomorphism of R-modules \Phi : A_1 \otimes ... \otimes A_n → L then φ = \Phi \circ \iota is a multilinear map.
Proof: Lemma 1: We have the natural map on simple tensors\iota ' : A_1 \otimes ... \otimes A_{n-1} \times A_n → (A_1 \otimes ... \otimes A_{n-1}) \otimes A_n ≅ A_1 \otimes ... \otimes A_n is well defined, and together with the evidently well-defined natural map\iota^* : A_1 \times ... \times A_n → A_1 \otimes ... \otimes A_{n-1} \times A_nwe have the relation \iota = \iota ' \circ \iota^*. Proof: The lemma will follow as soon as we show \iota ' is well defined. If (a_1 \otimes ... \otimes a_{n-1}, a_n)=(a_1' \otimes ... \otimes a_{n-1}', a_n') then by using these two given equivalencies in the coordinates we clearly have\iota '(a_1 \otimes ... \otimes a_{n-1}, a_n)=(a_1 \otimes ... \otimes a_{n-1}) \otimes a_n =(a_1' \otimes ... \otimes a_{n-1}') \otimes a_n' = \iota '(a_1' \otimes ... \otimes a_{n-1}', a_n')~\squareProceed by induction on n. For each a_n∈A_n defineφ_{a_n} : A_1 \times ... \times A_{n-1} → Lφ_{a_n}(a_1,...,a_{n-1})=φ(a_1,...,a_n)Note that φ_{a_n} is multilinearφ_{a_n}(a_1,...,ra_m+r'a_m',...,a_{n-1})=φ(a_1,...,ra_m+r'a_m',...,a_n)=rφ(a_1,...,a_m,...,a_n)+r'φ(a_1,...,a_m',...,a_n)=rφ_{a_n}(a_1,...,a_m,...,a_{n-1})+r`φ_{a_n}(a_1,...,a_m',...,a_{n-1})so that by induction there exists a unique homomorphism of R-modules\Phi_{a_n} : A_1 \otimes ... \otimes A_{n-1} → Lsuch that φ_{a_n}=\Phi_{a_n} \circ \iota^\Delta where \iota^\Delta is the natural map from A_1 \times ... \times A_{n-1} to A_1 \otimes ... \otimes A_{n-1}. Now defineσ' : A_1 \otimes ... \otimes A_{n-1} \times A_n → A_1 \otimes ... \otimes A_{n-1}σ'(a_1 \otimes ... \otimes a_{n-1}, a_n)=a_1 \otimes ... \otimes a_{n-1}ψ : A_1 \otimes ... \otimes A_{n-1} \times A_n → Lψ = \Phi_{a_n} \circ σ'(Where a_n in the definition of ψ depends on the input) Prove ψ is bilinear byψ(r \sum a_{i,1} \otimes ... \otimes a_{i,n-1} + r' \sum a_{i,1}' \otimes ... \otimes a_{i,n-1}',a_n)=\Phi_{a_n}(r \sum a_{i,1} \otimes ... \otimes a_{i,n-1} + r' \sum a_{i,1}' \otimes ... \otimes a_{i,n-1}')=r\Phi_{a_n}(\sum a_{i,1} \otimes ... \otimes a_{i,n-1})+r'\Phi_{a_n}(\sum a_{i,1}' \otimes ... \otimes a_{i,n-1}')=rψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)+r'ψ(\sum a_{i,1}' \otimes ... \otimes a_{i,n-1}',a_n)as well asψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},ra_n+r'a_n')=\Phi_{ra_n+r'a_n'}(\sum a_{i,1} \otimes ... \otimes a_{i,n-1})=\sum \Phi_{ra_n+r'a_n'}(a_{i,1} \otimes ... \otimes a_{i,n-1})=\sum (\Phi_{ra_n+r'a_n'} \circ \iota^\Delta)(a_{i,1},...,a_{i,n-1})=\sum φ_{ra_n+r'a_n'}(a_{i,1},...,a_{i,n-1})=\sum φ(a_{i,1},...,a_{i,n-1},ra_n+r'a_n')=\sum [rφ(a_{i,1},...,a_n)+r'φ(a_{i,1},...,a_n')]=\sum rφ(a_{i,1},...,a_n) + \sum r'φ(a_{i,1},...,a_n')=r\sum φ(a_{i,1},...,a_n) + r'\sum φ(a_{i,1},...,a_n')=rψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)+r'ψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n')so that there exists a unique homomorphism of R-modules\Phi : A_1 \otimes ... \otimes A_n → Lsuch that, in part, ψ = \Phi \circ \iota ' on simple tensors where \iota ' is described in the lemma. Now we can show\Phi \circ \iota = \Phi \circ \iota ' \circ \iota^* = ψ \circ \iota^* = \Phi_{a_n} \circ σ' \circ \iota^*Since clearly σ' \circ \iota^* = \iota^\Delta \circ σ whereσ : A_1 \times ... \times A_n → A_1 \times ... \times A_{n-1}σ(a_1,...,a_n)=(a_1,...,a_{n-1})we can continue\Phi_{a_n} \circ σ' \circ \iota^* = \Phi_{a_n} \circ \iota^\Delta \circ σ = φ_{a_n} \circ σ = φThis completed ⇒ argument in reverse also shows ψ = \Phi \circ \iota ' ⇔ φ = \Phi \circ \iota since \iota^* is surjective on simple tensors, which is seen to extend to \Leftarrow byψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)=\sum ψ(a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)=\sum (\Phi \circ \iota ')(a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)= \sum \Phi(a_{i,1} \otimes ... \otimes a_n)=\Phi(\sum a_{i,1} \otimes ... \otimes a_n) = \Phi((\sum a_{i,1} \otimes ... \otimes a_{i,n-1}) \otimes a_n)=(\Phi \circ \iota ')(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)This establishes the first part. For the converse, observe(\Phi \circ \iota)(a_1,...,ra_m+r'a_m',...,a_n)=\Phi(a_1 \otimes ... \otimes (ra_m + r'a_m') \otimes ... \otimes a_n)=r\Phi(a_1 \otimes ... \otimes a_m \otimes ... \otimes a_n) + r'\Phi(a_1 \otimes ... \otimes a_m' \otimes ... \otimes a_n) =r(\Phi \circ \iota)(a_1,...,a_m,...,a_n)+r'(\Phi \circ \iota)(a_1,...,a_m',...,a_n)~\square
Proof: Lemma 1: We have the natural map on simple tensors\iota ' : A_1 \otimes ... \otimes A_{n-1} \times A_n → (A_1 \otimes ... \otimes A_{n-1}) \otimes A_n ≅ A_1 \otimes ... \otimes A_n is well defined, and together with the evidently well-defined natural map\iota^* : A_1 \times ... \times A_n → A_1 \otimes ... \otimes A_{n-1} \times A_nwe have the relation \iota = \iota ' \circ \iota^*. Proof: The lemma will follow as soon as we show \iota ' is well defined. If (a_1 \otimes ... \otimes a_{n-1}, a_n)=(a_1' \otimes ... \otimes a_{n-1}', a_n') then by using these two given equivalencies in the coordinates we clearly have\iota '(a_1 \otimes ... \otimes a_{n-1}, a_n)=(a_1 \otimes ... \otimes a_{n-1}) \otimes a_n =(a_1' \otimes ... \otimes a_{n-1}') \otimes a_n' = \iota '(a_1' \otimes ... \otimes a_{n-1}', a_n')~\squareProceed by induction on n. For each a_n∈A_n defineφ_{a_n} : A_1 \times ... \times A_{n-1} → Lφ_{a_n}(a_1,...,a_{n-1})=φ(a_1,...,a_n)Note that φ_{a_n} is multilinearφ_{a_n}(a_1,...,ra_m+r'a_m',...,a_{n-1})=φ(a_1,...,ra_m+r'a_m',...,a_n)=rφ(a_1,...,a_m,...,a_n)+r'φ(a_1,...,a_m',...,a_n)=rφ_{a_n}(a_1,...,a_m,...,a_{n-1})+r`φ_{a_n}(a_1,...,a_m',...,a_{n-1})so that by induction there exists a unique homomorphism of R-modules\Phi_{a_n} : A_1 \otimes ... \otimes A_{n-1} → Lsuch that φ_{a_n}=\Phi_{a_n} \circ \iota^\Delta where \iota^\Delta is the natural map from A_1 \times ... \times A_{n-1} to A_1 \otimes ... \otimes A_{n-1}. Now defineσ' : A_1 \otimes ... \otimes A_{n-1} \times A_n → A_1 \otimes ... \otimes A_{n-1}σ'(a_1 \otimes ... \otimes a_{n-1}, a_n)=a_1 \otimes ... \otimes a_{n-1}ψ : A_1 \otimes ... \otimes A_{n-1} \times A_n → Lψ = \Phi_{a_n} \circ σ'(Where a_n in the definition of ψ depends on the input) Prove ψ is bilinear byψ(r \sum a_{i,1} \otimes ... \otimes a_{i,n-1} + r' \sum a_{i,1}' \otimes ... \otimes a_{i,n-1}',a_n)=\Phi_{a_n}(r \sum a_{i,1} \otimes ... \otimes a_{i,n-1} + r' \sum a_{i,1}' \otimes ... \otimes a_{i,n-1}')=r\Phi_{a_n}(\sum a_{i,1} \otimes ... \otimes a_{i,n-1})+r'\Phi_{a_n}(\sum a_{i,1}' \otimes ... \otimes a_{i,n-1}')=rψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)+r'ψ(\sum a_{i,1}' \otimes ... \otimes a_{i,n-1}',a_n)as well asψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},ra_n+r'a_n')=\Phi_{ra_n+r'a_n'}(\sum a_{i,1} \otimes ... \otimes a_{i,n-1})=\sum \Phi_{ra_n+r'a_n'}(a_{i,1} \otimes ... \otimes a_{i,n-1})=\sum (\Phi_{ra_n+r'a_n'} \circ \iota^\Delta)(a_{i,1},...,a_{i,n-1})=\sum φ_{ra_n+r'a_n'}(a_{i,1},...,a_{i,n-1})=\sum φ(a_{i,1},...,a_{i,n-1},ra_n+r'a_n')=\sum [rφ(a_{i,1},...,a_n)+r'φ(a_{i,1},...,a_n')]=\sum rφ(a_{i,1},...,a_n) + \sum r'φ(a_{i,1},...,a_n')=r\sum φ(a_{i,1},...,a_n) + r'\sum φ(a_{i,1},...,a_n')=rψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)+r'ψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n')so that there exists a unique homomorphism of R-modules\Phi : A_1 \otimes ... \otimes A_n → Lsuch that, in part, ψ = \Phi \circ \iota ' on simple tensors where \iota ' is described in the lemma. Now we can show\Phi \circ \iota = \Phi \circ \iota ' \circ \iota^* = ψ \circ \iota^* = \Phi_{a_n} \circ σ' \circ \iota^*Since clearly σ' \circ \iota^* = \iota^\Delta \circ σ whereσ : A_1 \times ... \times A_n → A_1 \times ... \times A_{n-1}σ(a_1,...,a_n)=(a_1,...,a_{n-1})we can continue\Phi_{a_n} \circ σ' \circ \iota^* = \Phi_{a_n} \circ \iota^\Delta \circ σ = φ_{a_n} \circ σ = φThis completed ⇒ argument in reverse also shows ψ = \Phi \circ \iota ' ⇔ φ = \Phi \circ \iota since \iota^* is surjective on simple tensors, which is seen to extend to \Leftarrow byψ(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)=\sum ψ(a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)=\sum (\Phi \circ \iota ')(a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)= \sum \Phi(a_{i,1} \otimes ... \otimes a_n)=\Phi(\sum a_{i,1} \otimes ... \otimes a_n) = \Phi((\sum a_{i,1} \otimes ... \otimes a_{i,n-1}) \otimes a_n)=(\Phi \circ \iota ')(\sum a_{i,1} \otimes ... \otimes a_{i,n-1},a_n)This establishes the first part. For the converse, observe(\Phi \circ \iota)(a_1,...,ra_m+r'a_m',...,a_n)=\Phi(a_1 \otimes ... \otimes (ra_m + r'a_m') \otimes ... \otimes a_n)=r\Phi(a_1 \otimes ... \otimes a_m \otimes ... \otimes a_n) + r'\Phi(a_1 \otimes ... \otimes a_m' \otimes ... \otimes a_n) =r(\Phi \circ \iota)(a_1,...,a_m,...,a_n)+r'(\Phi \circ \iota)(a_1,...,a_m',...,a_n)~\square
Thursday, June 6, 2013
Direct Products of Free Modules Need Not be Free (10.3.24)
Dummit and Foote Abstract Algebra, section 10.3, exercise 24:
For each i \in \mathbb{Z}^+ let M_i=\mathbb{Z} be free and let M=\prod_{i \in \mathbb{Z}^+}M_i. Let N be the restricted direct product of M considered as a submodule of M by componentwise multiplication by \mathbb{Z}. Assume M is free with basis B.
(a) Prove N is countable.
(b) Prove there exists countable B_1 \subseteq B such that N \subseteq N_1 = RB_1. Prove N_1 is countable.
(c) Prove \overline{M}=M/N_1 is free. Deduce that for any nonzero \overline{x} \in \overline{M} there are only a finite number of integers k such that \overline{x}=k\overline{m} for some \overline{m} \in \overline{M}.
(d) Let S=\{(b_1,...)~|~b_i = \pm i!\}, and prove S is uncountable. Deduce there exists s\in S with s \not \in N_1.
(e) Prove there exists s \in S with s \not \in N_1 such that for every positive integer k there exists \overline{m} \in \overline{M} such that \overline{s}=k\overline{m}, a contradiction by (c).
Proof: (a) Recall that countable unions of countable sets are countable. We haveN=\bigcup_{i∈\mathbb{Z}^+}N_i'where N_i'=\{x~|~x∈N \land \text{last}(x)=i\} for \text{last}(x) being the position of the last nonzero coordinate of x, is a countable union, so that it suffices to show N_i' is countable. We haveN_i'=\bigcup_{j∈\mathbb{Z}^++\{0\}}N_{i,j}''where N_{i,j}''=\{x~|~x∈N_i'' \land \sum_{y∈x} |y| = j\} is also a countable union, so we must prove N_{i,j}'' is countable; in fact, since for every y∈x∈N_{i,j}'' we must have -j≤y≤j implying |N_{i,j}''| < (2j+1)^i is finite, and as such is countable.
In general, this implies that any restricted direct product of copies of \mathbb{Z} will be countable.
(b) LetB_1=\bigcup_{x∈N}V_xwhere V_x is the set of elements from B involved in the nonzero terms of the unique sum representation of x. Since each V_x is finite and N is countable, B_1 is countable and clearly N \subseteq N_1 = RB_1. Since N_1 has B_1=\{a_1,...\} as a free basis (uniqueness of sums by the freeness of M), we have N_1 ≅ Ra_1 \oplus ... by n=z_1a_1+... \mapsto (z_1,...) and in this fashion N_1 is the restricted direct product of a number of copies of \mathbb{Z} and by the note at the end of (a) N_1 is thus countable.
(c) We claim \overline{B \setminus B_1} is a free basis for \overline{M}. We have \overline{B} generates \overline{M}, and the terms involving elements from B_1 may be removed as they are within N_1, so that \overline{B \setminus B_1} generates \overline{M}. Further, if \overline{z_1c_1+...+z_kc_k}=\overline{z_1'c_1+...+z_k'c_k} as sums with c_i∈B \setminus B_1, then for some n∈N-1 we have z_1c_1+...+z_kc_k+n=z_1c_1+...+z_kc_k+z_1''a_1+...+z_n''a_n=z_1'c_1+...+z_k'c_k and now z_1'c_1+...+z_k'c_k-z_1c_1-...-z_kc_k=z_1''a_1+...+z_n''a_n. Since the terms a_i and c_i don't coincide, by the freeness of M we can conclude z_1c_1+...+z_kc_k=z_1'c_1+...+z_k'c_k and again by freeness that z_i=z_i'.
Since such k manifest as the common divisors of the coefficients in \mathbb{Z} of the elements from \overline{B \setminus B_1}, and there are only a finite number of divisors for any integer, we have there are only a finite number of k per \overline{x}.
(d) Since there are two choices per coordinate of an element of b_i, this is equivalent to proving \prod^\infty \{0,1\} is uncountable, which is standard by Cantor's diagonal argument. Since S is now uncountable, S \subseteq N_1 would allow a natural surjection \mathbb{Z} → S by the countability of N_1; therefore, such s∈S and s∉N_1 exists.
(e) Let s=(b_1,...) be as in (d) and choose arbitrary k∈\mathbb{Z}. We have\overline{(b_1,...)}=\overline{(0,...,b_k,...)+(b_1,...,b_{k-1},0,...)}=\overline{(0,...,b_k,...)}=k\overline{(0,...,b_k',...)}as k \mid b_i=\pm i! for all i \geq k.~\square
(a) Prove N is countable.
(b) Prove there exists countable B_1 \subseteq B such that N \subseteq N_1 = RB_1. Prove N_1 is countable.
(c) Prove \overline{M}=M/N_1 is free. Deduce that for any nonzero \overline{x} \in \overline{M} there are only a finite number of integers k such that \overline{x}=k\overline{m} for some \overline{m} \in \overline{M}.
(d) Let S=\{(b_1,...)~|~b_i = \pm i!\}, and prove S is uncountable. Deduce there exists s\in S with s \not \in N_1.
(e) Prove there exists s \in S with s \not \in N_1 such that for every positive integer k there exists \overline{m} \in \overline{M} such that \overline{s}=k\overline{m}, a contradiction by (c).
Proof: (a) Recall that countable unions of countable sets are countable. We haveN=\bigcup_{i∈\mathbb{Z}^+}N_i'where N_i'=\{x~|~x∈N \land \text{last}(x)=i\} for \text{last}(x) being the position of the last nonzero coordinate of x, is a countable union, so that it suffices to show N_i' is countable. We haveN_i'=\bigcup_{j∈\mathbb{Z}^++\{0\}}N_{i,j}''where N_{i,j}''=\{x~|~x∈N_i'' \land \sum_{y∈x} |y| = j\} is also a countable union, so we must prove N_{i,j}'' is countable; in fact, since for every y∈x∈N_{i,j}'' we must have -j≤y≤j implying |N_{i,j}''| < (2j+1)^i is finite, and as such is countable.
In general, this implies that any restricted direct product of copies of \mathbb{Z} will be countable.
(b) LetB_1=\bigcup_{x∈N}V_xwhere V_x is the set of elements from B involved in the nonzero terms of the unique sum representation of x. Since each V_x is finite and N is countable, B_1 is countable and clearly N \subseteq N_1 = RB_1. Since N_1 has B_1=\{a_1,...\} as a free basis (uniqueness of sums by the freeness of M), we have N_1 ≅ Ra_1 \oplus ... by n=z_1a_1+... \mapsto (z_1,...) and in this fashion N_1 is the restricted direct product of a number of copies of \mathbb{Z} and by the note at the end of (a) N_1 is thus countable.
(c) We claim \overline{B \setminus B_1} is a free basis for \overline{M}. We have \overline{B} generates \overline{M}, and the terms involving elements from B_1 may be removed as they are within N_1, so that \overline{B \setminus B_1} generates \overline{M}. Further, if \overline{z_1c_1+...+z_kc_k}=\overline{z_1'c_1+...+z_k'c_k} as sums with c_i∈B \setminus B_1, then for some n∈N-1 we have z_1c_1+...+z_kc_k+n=z_1c_1+...+z_kc_k+z_1''a_1+...+z_n''a_n=z_1'c_1+...+z_k'c_k and now z_1'c_1+...+z_k'c_k-z_1c_1-...-z_kc_k=z_1''a_1+...+z_n''a_n. Since the terms a_i and c_i don't coincide, by the freeness of M we can conclude z_1c_1+...+z_kc_k=z_1'c_1+...+z_k'c_k and again by freeness that z_i=z_i'.
Since such k manifest as the common divisors of the coefficients in \mathbb{Z} of the elements from \overline{B \setminus B_1}, and there are only a finite number of divisors for any integer, we have there are only a finite number of k per \overline{x}.
(d) Since there are two choices per coordinate of an element of b_i, this is equivalent to proving \prod^\infty \{0,1\} is uncountable, which is standard by Cantor's diagonal argument. Since S is now uncountable, S \subseteq N_1 would allow a natural surjection \mathbb{Z} → S by the countability of N_1; therefore, such s∈S and s∉N_1 exists.
(e) Let s=(b_1,...) be as in (d) and choose arbitrary k∈\mathbb{Z}. We have\overline{(b_1,...)}=\overline{(0,...,b_k,...)+(b_1,...,b_{k-1},0,...)}=\overline{(0,...,b_k,...)}=k\overline{(0,...,b_k',...)}as k \mid b_i=\pm i! for all i \geq k.~\square
Wednesday, June 5, 2013
Generalization of the Primary Decomposition Theorem (10.3.22)
Dummit and Foote Abstract Algebra, section 10.3, exercise 22:
MathJax TeX Test Page
Let M=Tor(M) be an R-module for R a PID and let p∈R be prime. The p-primary component of M is the set of all elements of M that are annihilated by some positive power of p.
(a) Prove the p-primary component of M is a submodule of M.
(b) Prove this definition of p-primary components of M is consistent with the one given in 10.3.18 when Ann(M)≠0.
(c) Letting M_p be the p-primary component of M, proveM=\oplus_{p\text{prime}}M_pProof: (a) Let x,y∈M_p with p^ax=p^by=0. We have p^{a+b}(x+ry)=p^b(p^ax)+p^ar(p^by)=0 so that x+ry∈M_p.
(b) For Ann(M)=(a) for a=p_1^{\alpha_1}...p_k^{\alpha_k} with the previous definition of M^*_{p_i}=\{m~|~p_i^{\alpha_i}m=0\} we clearly have M^*_{p_i} \subseteq M_{p_i}. As well, for any m∈M_{p_i} we have am=0, so that for Ann(m)=(d) we have d \mid a. As well, by m∈M_{p_i} we have d \mid p_i^a for some positive a. This implies d \mid (a,p_i^a)=p_i^b for b ≤ \alpha_i so that now p_i^{\alpha_i}m=p^c(dm)=0 and M_{p_i} = M^*_{p_i}.
(c) Choose arbitrary x∈M. Letting X=Rx we have (a) ⊆ Ann(x) = (b) so that b | a = p_1^{\alpha_1}...p_k^{\alpha_k}. Say b = q_1^{β_1}...q_k^{β_k} where β_i ≤ α_i. By 10.3.18 we have X=X_{q_1} \oplus ... \oplus X_{q_k} ⊆ M_{p_1} ⊕ ... ⊕ M_{p_k} so that x is writable as a sum while in M_{p_1} + ... + M_{p_k} and now M=\sum_{p prime}M_p.
Proceed in the fashion of 10.3.22(ii) ⇒ (i). Let m∈M_{p_1} \cap (M_{p_2} + ... + M_{p_k}) for any selection such that p_j are distinct primes. Write m=n_1=n_2+...+n_k in accordance with the first and second halves, with p_i^{\alpha_i}n_i=0 for any i. For Ann(m)=(d) have d \mid p_1^{\alpha_1} for some positive a, and by observing the result of (\prod_{2 ≤ i ≤ k}p_k^{\alpha_k})m, we notice d \mid \prod_{2 ≤ i ≤ k}p_k^{\alpha_k} and now d=1 and 1m=m=0.~\square
(a) Prove the p-primary component of M is a submodule of M.
(b) Prove this definition of p-primary components of M is consistent with the one given in 10.3.18 when Ann(M)≠0.
(c) Letting M_p be the p-primary component of M, proveM=\oplus_{p\text{prime}}M_pProof: (a) Let x,y∈M_p with p^ax=p^by=0. We have p^{a+b}(x+ry)=p^b(p^ax)+p^ar(p^by)=0 so that x+ry∈M_p.
(b) For Ann(M)=(a) for a=p_1^{\alpha_1}...p_k^{\alpha_k} with the previous definition of M^*_{p_i}=\{m~|~p_i^{\alpha_i}m=0\} we clearly have M^*_{p_i} \subseteq M_{p_i}. As well, for any m∈M_{p_i} we have am=0, so that for Ann(m)=(d) we have d \mid a. As well, by m∈M_{p_i} we have d \mid p_i^a for some positive a. This implies d \mid (a,p_i^a)=p_i^b for b ≤ \alpha_i so that now p_i^{\alpha_i}m=p^c(dm)=0 and M_{p_i} = M^*_{p_i}.
(c) Choose arbitrary x∈M. Letting X=Rx we have (a) ⊆ Ann(x) = (b) so that b | a = p_1^{\alpha_1}...p_k^{\alpha_k}. Say b = q_1^{β_1}...q_k^{β_k} where β_i ≤ α_i. By 10.3.18 we have X=X_{q_1} \oplus ... \oplus X_{q_k} ⊆ M_{p_1} ⊕ ... ⊕ M_{p_k} so that x is writable as a sum while in M_{p_1} + ... + M_{p_k} and now M=\sum_{p prime}M_p.
Proceed in the fashion of 10.3.22(ii) ⇒ (i). Let m∈M_{p_1} \cap (M_{p_2} + ... + M_{p_k}) for any selection such that p_j are distinct primes. Write m=n_1=n_2+...+n_k in accordance with the first and second halves, with p_i^{\alpha_i}n_i=0 for any i. For Ann(m)=(d) have d \mid p_1^{\alpha_1} for some positive a, and by observing the result of (\prod_{2 ≤ i ≤ k}p_k^{\alpha_k})m, we notice d \mid \prod_{2 ≤ i ≤ k}p_k^{\alpha_k} and now d=1 and 1m=m=0.~\square
Direct Sum Equivalences (10.3.21)
Dummit and Foote Abstract Algebra, section 10.3, exercise 21:
MathJax TeX Test Page
Let I be an indexing set for submodules N_i of M. Prove\sum_{i∈I}N_i ≅ \oplus_{i∈I}N_i ⇔(\{i_1,...,i_k\} \subseteq I ⇒ N_{i_1} ∩ (N_{i_2}+...+N_{i_k}) = 0) ⇔(\{i_1,...,i_k\} \subseteq I ⇒ N_{i_1}+...+N_{i_k} = N_{i_1} \oplus ... \oplus N_{i_k}) ⇔(x∈\sum_{i∈I}N_i ⇒ ∃! \prod a_i ∈ \prod N_i~s.t.~(|\{y~|~0≠y∈\prod a_i\}|< \infty \land x=\sum y))Proof: (i) ⇒ (ii) Let n∈N_{i_1} ∩ (N_{i_2}+...+N_{i_k})). By the first half, we can simply write n as a sum, and by the second half we can write a sum only using elements of N_{i_2},...,N_{i_k}, which would necessarily be distinct sums when n≠0.
(ii) ⇒ (iii) This is simply proposition 5 applied to N_{i_1}+...+N_{i_k}.
(iii) ⇒ (iv) Since x∈\sum_{i∈I}N_i, we must necessarily have it written as a finite sum of nonzero elements, say x=\sum_{k∈K} b_k for some finite K \subseteq I where b_k≠0 for all k. Assume further that x=\sum_{j∈J} c_j for some finite J \subseteq I where b_j≠0 for all j, is another sum. We thus have x∈\sum_{l∈J \cup K} N_l = \oplus_{l∈J \cup K} N_l is writable as two sums, so that these two sums are in fact the same.
(iv) ⇒ (i) Define φ : \sum_{i∈I}N_i → \oplus_{i∈I}N_i by x = \sum n_i \mapsto \prod n_i where \sum n_i is the unique sum representation of x. We easily see this is a well-defined isomorphism of R-modules.~\square
(ii) ⇒ (iii) This is simply proposition 5 applied to N_{i_1}+...+N_{i_k}.
(iii) ⇒ (iv) Since x∈\sum_{i∈I}N_i, we must necessarily have it written as a finite sum of nonzero elements, say x=\sum_{k∈K} b_k for some finite K \subseteq I where b_k≠0 for all k. Assume further that x=\sum_{j∈J} c_j for some finite J \subseteq I where b_j≠0 for all j, is another sum. We thus have x∈\sum_{l∈J \cup K} N_l = \oplus_{l∈J \cup K} N_l is writable as two sums, so that these two sums are in fact the same.
(iv) ⇒ (i) Define φ : \sum_{i∈I}N_i → \oplus_{i∈I}N_i by x = \sum n_i \mapsto \prod n_i where \sum n_i is the unique sum representation of x. We easily see this is a well-defined isomorphism of R-modules.~\square
Primary Decomposition Theorem (10.3.18)
Dummit and Foote Abstract Algebra, section 10.3, exercise 18:
MathJax TeX Test Page
Let M be an R-module for R a PID, and let Ann(M)=(a)≠0. Let a=p_1^{\alpha_1}...p_k^{\alpha_k} be its decomposition into prime factors. Let M_i=Ann(p_i^{\alpha_i}), i.e. the p_i-primary component of M. ProveM=M_1 \oplus~...~\oplus M_k
Proof: Let n_i=a/p_i^{\alpha_i}. We can see that no prime factor divides all of the n_i, so that their greatest common divisor is 1 and there exist r_i such that 1=r_1n_1+...+r_kn_k. For any m∈M we have m=r_1n_1m+...+r_kn_km, and since p_i^{\alpha_i}(r_in_im)=r_i(am)=0, we have r_in_im∈M_i and now M=M_1+...+M_k. To observe uniqueness of sums, letm∈M_j∩(M_1+...+M_{j-1}+M_{j+1}+...+M_k)with Ann(m)=(d). We have d \mid p_j^{\alpha_j} from the first half as well as d \mid n_j from the second so that d=1 and now 1m=m=0.~\square
Tuesday, June 4, 2013
R-Module Homomorphisms from Free Modules (10.3.13)
Dummit and Foote Abstract Algebra, section 10.3, exercise 13:
MathJax TeX Test Page
Let R be a commutative ring. Prove Hom_R(R^n,R)≅R^n as R-modules.
Proof: Define ψ : Hom_R(R^n,R) → R^n by φ \mapsto (φ(e_1),...,φ(e_n)). We have ψ(φ_1+rφ_2) = ((φ_1+rφ_2)(e_1),...,(φ_1+rφ_2)(e_n)) =(φ_1(e_1),...,φ_1(e_n)) + r(φ_2(e_1),...,φ_2(e_n)) = ψ(φ_1)+rψ(φ_2)so ψ is a homomorphism of R-modules. If ψ(φ)=(0,...,0), then for arbitrary r^*=(r_1,...,r_n)=r_1e_1+...+r_ne_n∈R^n we have φ(r^*)=r_1φ(e_1)+...+r_nφ(e_n)=0 so that φ=0 and now ψ is injective. Again for arbitrary r^*∈R^n, define φ∈Hom_R(R^n,R) by φ(e_i)=r_i and extending linearly, so that ψ(φ)=r^* and now ψ is surjective and an isomorphism of R-modules.~\square
Proof: Define ψ : Hom_R(R^n,R) → R^n by φ \mapsto (φ(e_1),...,φ(e_n)). We have ψ(φ_1+rφ_2) = ((φ_1+rφ_2)(e_1),...,(φ_1+rφ_2)(e_n)) =(φ_1(e_1),...,φ_1(e_n)) + r(φ_2(e_1),...,φ_2(e_n)) = ψ(φ_1)+rψ(φ_2)so ψ is a homomorphism of R-modules. If ψ(φ)=(0,...,0), then for arbitrary r^*=(r_1,...,r_n)=r_1e_1+...+r_ne_n∈R^n we have φ(r^*)=r_1φ(e_1)+...+r_nφ(e_n)=0 so that φ=0 and now ψ is injective. Again for arbitrary r^*∈R^n, define φ∈Hom_R(R^n,R) by φ(e_i)=r_i and extending linearly, so that ψ(φ)=r^* and now ψ is surjective and an isomorphism of R-modules.~\square
Quotients and Finite Generation of Modules (10.3.7)
Dummit and Foote Abstract Algebra, section 10.3, exercise 7:
MathJax TeX Test Page
If N is a submodule of the R-module M, prove that if M/N and N are finitely generated, then so is M.
Proof: Let A=\{a_1,...,a_n\} generate N, and let B=\{b_1,...,b_m\} be such that \overline{B} generates M/N. We claim A \cup B generates M. Note that for an arbitrary element m∈M we have \overline{m}=\overline{r_1b_1+...+r_mb_m} for some r_i∈R so that m=r_1b_1+...+r_mb_m+n for some n∈N, so that m=r_1b_1+...+r_mb_m+s_1a_1+...+s_na_n for some s_i∈R, and now M is generated by A \cup B.~\square
Proof: Let A=\{a_1,...,a_n\} generate N, and let B=\{b_1,...,b_m\} be such that \overline{B} generates M/N. We claim A \cup B generates M. Note that for an arbitrary element m∈M we have \overline{m}=\overline{r_1b_1+...+r_mb_m} for some r_i∈R so that m=r_1b_1+...+r_mb_m+n for some n∈N, so that m=r_1b_1+...+r_mb_m+s_1a_1+...+s_na_n for some s_i∈R, and now M is generated by A \cup B.~\square
Saturday, June 1, 2013
R-Module Homomorphisms from R (10.2.9-10)
Dummit and Foote Abstract Algebra, section 10.2, exercises 9-10:
MathJax TeX Test Page
Let R be a commutative ring with 1.
9. Prove Hom_R(R,M)≅M as R-modules.
10. Prove End(R) ≅ R as rings.
Proof: (9) Define a mapping ψ : Hom_R(R,M) → M by φ \mapsto φ(1), and prove it is an R-module isomorphism. We have ψ(φ_1+rφ_2)=φ_1(1)+rφ_2(1)=ψ(φ_1)+rψ(φ_2) for R-module homomorphicity, and if ψ(φ)=φ(1)=0 then φ(r)=rφ(1)=0 so that φ=0 and ψ is thus injective. We show ψ is surjective by constructing the mapping φ defined by φ(1)=m and φ(r)=r \cdot m and showing φ∈Hom_R(R,M): We have this because φ(r_1+r^* \cdot r_2)=(r_1+r^* \cdot r_2)m=r_1 \cdot m +r^* \cdot (r_2 \cdot m)=φ(r_1)+r^* \cdot φ(r_2).
(10) By the above they are already isomorphic as abelian groups, so it suffices to show ψ in this case fulfills ψ(φ_1 \circ φ_2)=(φ_1 \circ φ_2)(1)=φ_1(φ_2(1))=φ_1(φ_2(1) \cdot 1))=φ_1(1)φ_2(1)=ψ(φ_1)ψ(φ_2)~\square
9. Prove Hom_R(R,M)≅M as R-modules.
10. Prove End(R) ≅ R as rings.
Proof: (9) Define a mapping ψ : Hom_R(R,M) → M by φ \mapsto φ(1), and prove it is an R-module isomorphism. We have ψ(φ_1+rφ_2)=φ_1(1)+rφ_2(1)=ψ(φ_1)+rψ(φ_2) for R-module homomorphicity, and if ψ(φ)=φ(1)=0 then φ(r)=rφ(1)=0 so that φ=0 and ψ is thus injective. We show ψ is surjective by constructing the mapping φ defined by φ(1)=m and φ(r)=r \cdot m and showing φ∈Hom_R(R,M): We have this because φ(r_1+r^* \cdot r_2)=(r_1+r^* \cdot r_2)m=r_1 \cdot m +r^* \cdot (r_2 \cdot m)=φ(r_1)+r^* \cdot φ(r_2).
(10) By the above they are already isomorphic as abelian groups, so it suffices to show ψ in this case fulfills ψ(φ_1 \circ φ_2)=(φ_1 \circ φ_2)(1)=φ_1(φ_2(1))=φ_1(φ_2(1) \cdot 1))=φ_1(1)φ_2(1)=ψ(φ_1)ψ(φ_2)~\square
Subscribe to:
Posts (Atom)