4

I have these two definitions of span:

Span:

Suppose a vector space $(V,+,\cdot)$, and

$$S = \{u_1,\cdots,u_n\}$$

(and $S$ is a subset of $V$, not a subspace)

$$[S]=:\cap_{w\subset V, w\supseteq S} W$$

In other words, $[S]$ is, by definition, the intersection of all $W$, such that $W$ is a subspace of $V$ and $W$ contains $S$.

Span: $$S = \{u_1,\cdots,u_n\}$$ The span of $S$, denote by $[S]$ is, then: $$W = \{\alpha_1u_1+\cdots+\alpha_nu_n|\alpha_i\in\mathbb R\}$$

I need to prove that if $U = [S_1]$ is a subset of $V$ and $W = [S_2]$ is also a subset of $V$, then $$[S_1]+[S_2] = [S_1\cup S_2]$$

where, I think, $[S_1]+[S_2]$ means sum of subspaces

By my second definition, I did:

$$S_1 = {u_1,\cdots,u_n} \implies [S_1] = \{\lambda_1 u_1+\cdots+\lambda_n u_n|\lambda_i\in\mathbb R\}$$ $$S_2 = {w_1,\cdots,w_n} \implies [S_2] = \{\gamma_1 w_1+\cdots+\gamma_n w_n|\gamma_i\in\mathbb R\}$$ Then, the sum of two subspaces $U$ and $W$ is:

$$\{u+w|u\in U, w\in W\}$$ therefore

$$[S_1] + [S_2] = \{\lambda_1 u_1+\cdots+\lambda_n u_n + \gamma_1 w_1+\cdots+\gamma_n w_n|\lambda_i,\gamma_i \in \mathbb R\}\tag{1}$$

And note that: $$S_1\cup S_2 = \{u_1,\cdots,u_n,w_1,\cdots,w_m\}$$

therefore:

$$[S_1\cup S_2] = \{\beta_1 u_1+\cdots+\beta_n u_n+\beta_{n+1} w_1+\cdots+\beta_{n+m}w_m|\beta_i\in\mathbb R\}\tag{2}$$

So $(1)$ and $(2)$ are essentially the same thing. Is my reasoning correct? I did not prove that one set is a subset of the other, but I worked on the two sides of the equations and found what is essentially the same set. I think there's no problem with it, right?

Update: here's my idea for a better proof using the first definition:

$$S_1 = \{v_1,\cdots,v_n\}\\S_2 = \{u_1,\cdots,u_m\}\\ S_1 \cup S_2 = \{v_1,\cdots,v_n,u_1,\cdots,u_m\}\\ k\in \operatorname{span}(S_1\cup S_2)\implies \\k = a_1v_1+\cdots+a_nv_n + b_1u_1+\cdots+b_mu_m \\= (a_1v_1+\cdots+a_nv_n)+(b_1u_1+\cdots+b_mu_m) \in \operatorname{span}(S_1)+\operatorname{span}(S_2)$$ $$k\in \operatorname{span}(S_1)+\operatorname{span}(S_2)\implies\\k = (a_1v_1+\cdots+a_n v_n) + (b_1u_1 + \cdots+b_mu_m) = \\ a_1v_1+\cdots+a_nv_n + b_1u_1+\cdots+b_mu_m \in \operatorname{span}(S_1 \cup S_2)$$ $k$ was chosen arbitrarily, therefore everything above Works for all $k$. Then we have: $$\operatorname{span}(S_1) + \operatorname{span}(S_1) \subseteq \operatorname{span}(S_1\cup S_2)\\\operatorname{span}(S_1\cup S_2)\subseteq \operatorname{span}(S_1) + \operatorname{span}(S_1)\\ \implies\\\operatorname{span}(S_1) + \operatorname{span}(S_1) = \operatorname{span}(S_1\cup S_2)$$

The main question is: how to prove the same thing, but for infinite sets $S_1$ and $S_2$? This is a proof that might envolve the first definition, because it works for infinite sets. Of course it has something to do with proving $[S_1]+[S_2]\subseteq [S_1 \cup S_2]$ and $[S_1\cup S_2] \subseteq [S_1]+[S_2]$, but how to proceed?

What I tried:

I know by definition that

$$[S_1\cup S_2] = \cap_{w\subset V, w\supseteq S_1\cup S_2} W$$

Update: To illustrate, I've made this figure: enter image description here

that shows some possible $W_i$ of the intersection on the definition.

My idea is: pick any $x\in [S_1\cup S_2]$, therefore this $x$ is in the purple intersection of the figure. If it is in the intersection, it is in all subspaces $W_i$. But since $W_i$ is a subspace that contains $S_1\cup S_2$, one of these subspaces can be $[S_1]+[S_2]$ because:

$$[S_1]+[S_2] = \{u+v| u\in [S_1], v \in [S_2]\}$$ And, remember, this is a subspace, but must also contain $S_1\cup S_2$ by the definition. But it contains, because we could show that there exists $2$ subsets of $[S_1]+[S_2]$ in the form:

$$w_1 = \{u+0|u\in [S_1], 0 \in [S_2]\} = [S_1] \supset S_1 $$ $$w_2 = \{0+v|0\in [S_1], v \in [S_2]\} = [S_2] \supset S_2 $$

Since both $w_1\supset S_1$ and $w_2\supset S_2$ are in $[S_1]+[S_2]$, then $w_1\cup w_2 = [S_1]\cup [S_2] \supset S_1\cup S_2$

I showed that such subspace $[S_1]+[S_2]$ matches the definition, so all $x\in [S_1\cup S_2]\subset [S_1]+[S_2]$.

I think, then, one side is proved. I need to prove the converse.

Poperton
  • 6,594
  • I can help you with the book :) look at Friedberg's there they talk about infinite things :) hope this help – user162343 Dec 09 '14 at 15:52
  • For the first question, when you are proving two sets having same formula of definition, you are implicitly proving that one is subset of another and this holds in the reverse direction, hence are the same set. –  Mar 28 '19 at 12:26
  • For the second question, you can still use the same proof using definiton 2, since span of S is defined as every possible linear combination of vectors in the set. Note that linear combination is automatically meaning finite linear combination, so still hold in infinite set S –  Mar 28 '19 at 12:28
  • To use the first definition of span for infinite set S, you can prove the equivalence of two definition, which is easy. –  Mar 28 '19 at 12:29

0 Answers0