Topic: `"Field Extensions and"` `sl(2 , RR)`.

Introduction: `"Why Extensions Matter for Lie Algebras"`.
In our "Vector Space" track , we established that a field extension `bbb"E"` over `bbb"F"` is just a vector space. In Lie Algebra theory , we often "change the scalars" of our algebra. If we have a Lie algebra defined over `RR` (like our `3D` model) , we can extend the scalars to `CC`.

This is crucial because , while a polynomial like `x^2 + 1 = 0` has no roots in `RR` , it has roots in the algebraic extemsion `CC`. In the context of our basis `{e , f , h}` , this determines whether we can find specific eigenvectors , which are unique representations (canonical forms) for our operators.

Theory: `"The Degree of the Adjoint Extension"`.
First we will give a short outline of the theory. Beneath there is a more detailed explanation of this theory.
Recall our basis `{e , f , h}` for `sl(2 , RR)`. Every element `x` in the Lie algebra defines a linear transformation called the adjoint map , denoted `ad_x (y) = [x , y]`.
If we look at an operator `ad_h` , its characteristic polynomial is `P(lamda) = lambda^3 - 4 lamda`. The roots (eigenvalues) are `{0 , 2 , -2}`. Since these roots already live in `RR` , the "extension" required to decompose this operator is of degree 1. So the smallest field over which the operstor `ad_h` splits into eigenvalues is just `RR` itself.
However , for other elements in algebra , the characteristic polynomial might not split in `RR`. If we encounter a polynomial of degree n that is irreducible over `bbb"F"` , we must move to a finite extension `bbb"E"` , where `[bbb"E" : bbb"F"] = n` to fully "see" the structure of the algebra.

Further details of the theory :
`Setup: sl(2 , RR)`. Take for example the standard basis of `sl(2 , RR)`

`h = ([1 , 0] , [0 , -1]) , e = ([0 , 1] , [0 , 0]) , f = ([0 , 0] , [1 , 0]) `

This is the conventional order of the basis elements, but there are other conventions in use , depending on the purpose. And the trace will always be 0.
They satisfy the commutative relations `[h , e] = 2e , [h , f] = -2f , [e , f] = h`.
The adjoint operator `ad_h` is the linear map , `ad_h : sl(2 , RR) rarr sl(2 , RR) , \ ad_h(x) = [h , x] `.
Compute its action on the basis:
`ad_h(e) = [h , e] = 2e`.
`ad_h(f) = [h , f] = -2f`.
`ad_h(h) = [h , h] = 0`.
So relative to the ordered basis `(e , h , f)` we have:
`ad_h(e) = 2e + 0h + 0f`.
`ad_h(h) = 0e + 0h + 0f`.
`ad_h(f) = 0e + 0h -2f`.
Thus the matrix of `ad_h` is

`[ad_h]_{(e , h , f)} = ([2 , 0 , 0] , [0 , 0 , 0] , [0 , 0 , -2])`.

`"Characteristic polynomial and its computation"`.
For a linear operator `T` on an `n` - dimensional vector space over a field `bbb"F"` , the characteristic polynomial is `P_T(lamda) = det(lamda I - T)` , a degree - n polynomial with coefficients in `bbb"F"`.
Here `T = ad_h` , the space has dimension 3 , and in our chosen basis

`lamda I - [ad_h] = ([lamda - 2 , 0 , 0] , [0 , lamda , 0] , [0 , 0 , lamda + 2])`.

Compute the determinant : `P(lamda) = det (lamda I - [ad_h]) = (lamda -2) * lamda * (lamda + 2) = lamda (lamda^2 - 4) = lamda^3 - 4 lamda`.
So in this context , `P(lamda) = lamda^3 - 4 lamda` is the characteristic polynomial of the adjoint operator `ad_h` acting on `sl(2 , RR)`.

`"Eigenvalues and their relation to" \ P(lamda)`.
An eigenvalue `lamda` of a linear operator `T` is a scalar for which there exists a nonzero vector with `T(v) = lamda v`. Equivalently , `lamda` is an eigenvalue if and only if `det(lamda I - T) = 0` , i.e. `lamda` is the root of the characteristic polynomial. We factor:
`P(lamda) = lamda^3 - 4 lamda = lamda(lamda^2 - 4) = lamda (lamda - 2)(lamda + 2)`. Thus the eigenvalues are: `lamda = 0 \ , \ lamda = 2 \ , \ lamda = -2`. You can see this directly from the matrix as well: `[ad_h]` is diagonal with diagonal entries `2 , 0 , -2` , and those diagonal entries are exactly its eigenvalues.
Geometrical:
- The line spanned by `e` is an eigenspace with eigenvalue 2 , since ad_h (e) = 2e.
- The line spanned by `h` is an eigenspace with eigenvalue 0 , since `ad_h (h) = 0`.
- The line spanned by `f` is an eigenspace with eigenvalue -2 , since `ad_h (f) = -2f`.

`"Why the field extension has degree 1"`.
Given a linear operator `T` defined over a field `F` , one often asks over which field the characteristic polynomial splits completely into linear factors , so that you can decompose the space into eigenspaces. In our case here the roots of the characteristic polynomial are all lying in `RR`.

`P(lamda)` already splits completely into linear factors over `RR` (the extension degree is always relative to a base field). There is no need to pass to a larger field like `CC` to obtain eigenvalues or eigenspaces. The splitting field of `P(lamda)` over `RR` is just `RR` itself , which is an extension of `"degree" \ 1`.

Formally , if `bbb"E"` is the smallest field containing `RR` and all roots of `P` , then here `bbb"E" = RR`. The extension degree `[bbb"E" : RR]` equals `1` , so we say that the extension required to decompose this this operator is of `"degree" \ 1`. This means generally that `bbb"E = F"` because `bbb"E"` is considered a vector space over `bbb"F"` , and when `bbb"E = F"` , the set `{1}` form a basis for `bbb"E"` over `bbb"F"`. Since the basis has exactly one element , the dimension , known as the degree of the extension `[bbb"E : F" ]` , is `1`. If `bbb"E = F"` , every element `alpha in bbb"E"` can be written as `alpha = a * 1` , where `a in bbb"F"`. Thus `{1}` spans `bbb"E"` and is linearly independent , making it a basis. This extension is also called a trivial extension (or improper) when `bbb"E = F"`.

Exercise: `"Extending the Scalar Field"`.
Consider a hypothetical element `w` in a Lie algebra over `QQ` whose adjoint map `ad_h` has the minimal polynomial `P(x) = x^2 - 2`.
1. Is `ad_w` diagonalizable over the field `QQ` `?`
2. If we move to the algebraic extension `bbb"E" = QQ(sqrt 2)` , what is the degree of this extension `[bbb"E" : QQ(sqrt 2)]` , what is the degree of this extension `[bbb"E" : QQ]` `?`
3. How many basis elements are needed to represent an element of `bbb"E" = QQ(sqrt 2)` over `QQ` `?`

Solution:
1. No. The roots of `x^2` are `+- sqrt 2` , which are not in `QQ`.
2. The `"degree"` is `2` , because the minimal polynomial `x^2 -2` is of degree `2`.
3. As a vector space , `bbb"E"` over `QQ` requires `2` basis elements `{1 , sqrt 2}`. By extending the field to include these roots , we can now find the "eigenspaces" for `ad_w` that were previously hidden. QED

Alternative Solution:
`"Reformulation of exercise"`: Consider a hypothetical element `w` in the Lie algebra over `QQ`. Its adjoint map `ad_w` (the linear map `x rarr [w , x]`) has minimal polynomial `m(x) = x^2 - 2 in QQ[x]`. Is the operator diagonalizable over the field `QQ` `?` Justify your answer carefully.

`"Step 1 : Recall the criterion using the minimal polynomial"`.
For a linear operstor `T` on a finite dimensional vector space over a field `bbb"F"` , there is a standard criterion : `T` is diagonalizable over `bbb"F"` if and only if its minimal polynomial splits into distinct linear factors over `bbb"F"` , i.e. `m_T(x) = prod_i (x - lamda_i)` , where each `lamda_i in bbb"F"` and no factor is repeated. Equivalently: The minimal polynomial must have only simple roots , and all of them must lie in the base field `bbb"F"`.

`"Step 2 : Analyse the given minimal polynomial over" \ QQ`.
Here the minimal polynomial is `m(x) = x^2 - 2`.
1. Factorization over `QQ` : Over `QQ` , the polynomial `x^2 - 2` is irreducible (by the rational root test) , it has no rational roots , so there is no linear factor in `QQ[x]`. Therefore , over `QQ` , it does not split into linear factors , it remains a quadratic with no roots in `QQ`.
2. Factorization over `RR` (just a comparision): Over `RR` , we have `x^2 - 2 = (x - sqrt 2)(x + sqrt 2)` , so it splits into distinct linear factors in `RR[x]`. But `sqrt 2 ne QQ` , which is the key issue for diagonalizability over `QQ`.

`"Step 3 : Conclusion about diagonalizability over" \ QQ`.
Apply the criterion: Minimal polynomial of `ad_w` is `x^2 - 2`. Over `QQ` , it does not split into linear factors (it has no roots in `QQ`). Therefore `ad_w` is not diagonalizable over `QQ`.

So the detailed answer: Over `QQ` , the operator `ad_w` cannot be represented by a diagonal matrix (with entries in `QQ`) in any basis , because its minimal polynomial does not have all its roots in `QQ`.

`"Note"`: A diagonal matrix is a square matrix where all the entries off the main diagonal (from top-left to bottom-right) are zero.

`"Step 4 : What happens over an extension field ? "`
As an illustration , consider the field extension `QQ(sqrt 2)`. Over `E = QQ(sqrt 2)` , the same polynomial factors as `x^2 - 2 =(x - sqrt2)(x + sqrt 2)` , which is a product of distinc linear factors in `bbb"E" [x]`. Thus over `E = QQ(sqrt 2)` , the minimal polynomial of `ad_w` , splits completely into simple linear factors. By the diagonalizability criterion , `ad_w` is diagonalizable over `QQ(sqrt 2)`. So:
- Not diagonalizable over `QQ`.
- Diagonalizable over the quadratic extension `QQ(sqrt 2)` , whose degree over `QQ` is `2`. This matches the general philosophy: extending the field to include the eigen values (roots of the minimal polynomial) can turn a non-diagonalizable operator over the smaller field into a diagonalizable one over the larger field.