Generating Sets and Span: Let `bbb"V"` be a vector space over a field `bbb"F"` , and let `S = {bb{a}_i : i in I}` be a non-empty collection of vectors in `bbb"V"`.
-Part a: Using the previous exercise (intersections and subspaces) , define the subspace of `bbb"V"` generated by S.
-Part b: Prove that the vectors in the subspace generated by S are precisely the finite linear combinations of vectors in S.

Purpose: This is a fundamental brigde in linear algebra. It connects the "cutting-down" definition of a subspace (using intersections) with the "building-up" construction (using linear combinations).

Reformulate the Given Exercise: The objective is to formally define the span of a set S as the "smallest" subspace containing S (the intersection of all subspaces) and then prove that this abstractly defined set is identical to the set of all possible finite linear combinations of elements from S.

Give a General Strategy for a Solution:
For part a: We will use the result that the intersection of subspaces is a subspace to define the "subspace generated by S".
For part b: This is a double containment proof (`A subseteq B` and `B subseteq A`).
1. Let `bbb"W"` be the intersection of all subspaces containing S.
2. Let L be the set of all finite linear combinations of vectors in S.
3. Show `bb{L subseteq bbb"W"}` by showing that eny subspace containing S must , by closure , contain all linear combinations of S.
4. Show `bb{bbb"W" subseteq L}` by proving that L itself is a subspace containing S. Since `bbb"W"` is the intersection of all such subspaces , it must be contained within L.

Write Down the Specific Theory Used:
- Intersection Property: The intersection of any collection of subspaces is a subspace.
- Definition of Linear Combination: A vector `bb{v}` is a linear combination of vectors `bb{a}_1 , ... , bb{a}_n in S` if `bb{v} = sum_(j = 1)^n c_j bb{a}_j` for some scalars `c_j in bbb"F"`.
Closure Axioms: A subspace is closed under addition and scalar multiplication. By induction , it is closed under any finite number of these operations (i. e. finite linear combinations).

Additional Explanations:
The structure of `bbb"V"`:
- A large set `bbb"V"`: the entire vector space.
- Several overlapping smaller sets: `bbb"W"_1 , bbb"W"_2 , bbb"W"_3 \ ... :` various subspaces of `bbb"V"` that contain the set S.
- A small cluster S: the given set of vectors lying within all the subspaces.
- The intersection of all `bbb"W"_i subseteq bbb"V"` : this is the smallest subspace containing S , i.e. the subspace generated by S.
Or more formally: `langle S rangle = bigcap {bbb"W" subseteq bbb"V" : S subseteq bbb"W"}`.

A chain of inclusions (shrinking intersections):
The chain line renders as: `bbb"V" supseteq bbb"W"_1 supseteq bbb"W"_2 supseteq \ ... \ supseteq bbb"W"_n supseteq langle S rangle supseteq S` , showing each larger subspace down towards `langle S rangle` , where S is just a subset of the smallest subspace `langle S rangle` , meaning S is a set of vectors and not a subspace. From this "cutting down" process , we can also get a "building up" picture:

The Cutting Down Chain: `bbb"V" supseteq bbb"W"_1 supseteq bbb"W"_2 supseteq \ ... \ supseteq bbb"W"_n supseteq langle S rangle supseteq S`.
The Building Up Chain: `S subseteq S_1 subseteq S_2 subseteq \ ... \ subseteq langle S rangle`.

Interpretation of Cutting Down Chain:
- Every subspace `bbb"W"_i` containing S is "larger" than the `span(S) = langle S rangle`.
- The span `langle S rangle` is the core intersection zone where all those larger subspaces overlap - basically , the minimal "footprint" needed to contain S and be a subspace.
- The Cuttin Down Chain expresses the successive restriction of all subspaces of `bbb"V"` that contain S, narrowing down by intersection until only the smallest such subspace remains — the generated subspace `langle S rangle`.

Interpretation of Building Up Chain:
- The Building Up Chain is the constructive process of successively enlarging S by adding all its finite linear combinations — first single scalar multiples, then sums of two, three, and so on — until closure under addition and scalar multiplication is complete, yielding the full subspace `langle S rangle`.

This is why defining the span as an intersection or as all linear combinations are equivalent - both capture precisely those minimal elements common to every admissable subspace containing S.

Is S the smallest set ? Yes - initially , S is just a set of vectors , not necessarily a subspace. When we "expand" S into its span , we're enclosing it inside the smallest subspace that contains it. When you go from S to span(S) , you are turning a mere collection of vectors into a structured set - a vector space on its own (a subspace of `bbb"V"`). So S itself is not the smallest subspace , but its span is the smallest subspace that contains S.

"Expanding" S: How does it work ? You take all finite linear combinations of elements of S: `span(S) = {sum_(k = 1)^n lamda_k bb{a}_k \ : n in NN , lamda_k in bbb"F" , bb{a}_k in S}`. You are allowing three operations to occur freely:
1. Scaling vectors in S by any scalar from the field `bbb"F"`.
2. Adding the results.
3. Repeating finitely many times.
And because any subspace must be closed under addition and scalar multiplication , this collection is naturally the smallest subspace you can have that contain S.

Relationship Between "Linear Combination" and "Span": The two terms are closely related , but not identical in meaning - rather , "span" is built from linear combinations.
- A linear combination is a single element of the form `lamda_1 bb{a}_1 + \ ... \ + lamda_n bb{a}_n`.
- The span of S is the set of all possible linear combinations of elements of S. So in other words: linear combinations are the ingredients and the span is the dish made from all these ingredients.

Formally:                  `tt "span(s) = {all finite linear combinations of elements of S"`}.

Coefficients from the Field `bbb"F"`: All coefficients you use when forming linear combinations must come from the underlying field `bbb"F"` that defines the vector space. That is what ties the vector space structure together. For example:
- In `RR^n` , the coefficients are real numbers.
- In `CC^n` , the coefficients must be complex.
- If `bbb"V"` were a vector space over `QQ` , only rationals are allowed.

Example: The set of vectors S is not necessarily closed under operations.
For example {(1 , 0) , (0 , 1)} in `RR^2`. The linear combination `lamda_1 bb{a}_1 + lamda_2 bb{a}_2` is a single vector , as in the example 3(1 , 0 ) + 2(0 , 1) = (3 , 2).
The set of all finite linear combinations of vectors in S is the span(S). The span(S) is a set and subspace of for example `RR^2`. It can also be the whole space `bbb"V" = RR^2` when S spans all directions. Look at `lamda_1 (1 , 0) + lamda_2 (0 , 1) = (lamda_1 , lamda_2)` , with `lamda_1 , lamda_2 in RR^2`. The set of all such vectors is `RR^2` itself. Hence , `span (S) = RR^2 `.

So the distinction is:
- Linear combination = one vector constructed from S.
- Span = the set of all such vectors , which naturally form a subspace.

Why only finite combination ?
Because vector spaces are defined with only two binary operations. To add infinitely many vectors meaningfully you'd need additional structure , namely:
- a topology , to make sense of limits (like in Banach or Hilbert spaces) ,
- and a definition of convergence for infinite sums.
That kind of structure doesn't exist in a general algebraic vector space unless explically added.

The conceptional development: The subspace generated by S is called `bb{langle S rangle}`. At first "generated" is not given (as for groups) , but has to be defined in a sensible way in the vector space setting. Guessing that the smallest intersection `bbb"W"` containing S might be the subspace generated by S , and equal to the finite linear combinations of vectors in S. The set of all these combinations is defined to be the `span(bb{S})`.

1. We start with a vector space `bbb"V"` and a subset `S subseteq bbb"V"`.
2. We realize that in analogy with group theory one should be able to talk about the subspace generated by S.
- In groups , "generated" means "the smallest subgroup containing S".
- So we follow the same idea:
Define the subspace generated by S as the smallest subspace containing S.
3. How do we express "smallest"? By intersection: The family of all subspaces `bbb"W"` of a vector space `bbb"V"` containing a given set S forms a chain where the smallest is the `span(S) = langle S rangle`. Because intersections of subspaces are again subspaces , this is well-defined.
4. Now , once we have this definition , we want an explicit characterization of what vectors lie inside. The natural conjecture: "Maybe every vector in that intersection can be written as a finite linear combination of element of S".
5. That becomes the theorem to prove.

Definition: Let `bbb"V"` be a vector space over a field `bbb"F"` and `S subseteq bbb"V"`.
Define: `langle S rangle = bigcap {bbb"W" subseteq bbb"V" : S subseteq bbb"W"}`. Here the intersection runs over every subspace `bbb"W"` of `bbb"V"` that contains S. This object is called the subspace generated by S.

The Proof: Let `L = {sum_(j = 1)^n c_j bb{a}_{i_j} : n in NN , c_j in F , bb{a}_{i_j} in S}` be the set of all finite linear combinations of vectors in S. We want to show `langle S rangle = L`.
1. Proof that `L subseteq langle S rangle`: Let `bb[v} in L`. Then `bb{v}` is a finite linear combination of vectors in S. Let `bbb"W"` be any subspace that contains S. Since `bbb"W"` is a subspace , it is closed under addition and scalar multiplication. Therefore , `bbb"W"` must contain every finite linear combination of its elements. Since `S subseteq bbb"W"` , then `bb{v} in langle S rangle`.
2. Proof that `langle S rangle subseteq L`: To show this , we simply show that L is one of the subspaces we are intersecting.
- Contains S: Any `bb{a}_i in S` can be written as `1 * bb{a}_i` , which is a finite linear combination. Thus `S subseteq L`.
- Is a subspace: The zero vector is a "trivial" linear combination with all `c_j = 0`.
Sum: The sum of two finite linear combinations is another finite linear combination.
Since L is a subspace containing S , and `langle S rangle` is the intersection of all such subspaces , then `langle S rangle` must be a subset of L , because it can't be bigger than any single member of the intersection. QED

Conclusion: Since `L subseteq langle S rangle` and `langle S rangle subseteq L` , we have `langle S rangle = L`. END