I am reading Tu's book titled "Introduction to Manifolds". In Chapter 2, paragraph 5.3, example 5.15 states that
…Since $\mathbb{R}^{m\times n}$ is isomorphic to $\mathbb{R}^{m n}$, we give it the topology of $\mathbb{R}^{m n}$…
I have a few questions about this statement:
-
If we have to say that $\mathbb{R}^{m\times n}$ is isomorphic to $\mathbb{R}^{m n}$, that means that these two spaces are different? How are they different? What is the first space and what is the second?
-
we give the topology of $\mathbb{R}^{m n}$ to what? To $\mathbb{R}^{m\times n}$? And if so, why are we allowed to give the topology of some space to another, given that they are isomorphic?
-
How is all that useful for what is next, i.e. defining the general linear group of $n\times n$ matrices with real entries?
Any help will be appreciated…
Best Answer
Q1:
Both $\mathbb{R}^{m\times n}$ and $\mathbb{R}^{m n}$ are vector spaces over the ground field $\mathbb R$ with dimensio $m \cdot n$. However, the elements of $\mathbb{R}^{m\times n}$ are $(m \times n)$-matrices $A = (a_{ij})$ and the elements of $\mathbb{R}^{m n}$ are tupels $(x_1,\ldots, x_{mn})$. The two vector spaces are isomorphic, but not identical. An isomorphism is given by writing down the rows of a matrix one after the other, that is $$\phi : \mathbb{R}^{m\times n} \to \mathbb{R}^{m n}, \phi((a_{ij}) = (a_{11},\ldots,a_{1n},a_{21}, \ldots, a_{2n}, \ldots, a_{m1},\ldots,a_{mn}) .$$
Note that there are many other isomorphisms. For example we can write down the columns of a matrix one after the other.
Euclidean spaces $\mathbb R^k$ have a standard topology induced by the Euclidean norm $\lVert (x_1,\ldots,x_k) \rVert = \sqrt{\sum_{i=1}^k x_i^2}$.
We can now give $\mathbb{R}^{m\times n}$ the unique topology making $\phi$ a homeomorphism. But doesn't this topology depend on the choice of a linear isomorphism $\phi$ between $\mathbb{R}^{m\times n}$ and $\mathbb{R}^{m n}$? It is an easy exercise to show that this is not the case: For all isomorphisms $\phi : \mathbb{R}^{m\times n} \to \mathbb{R}^{m n}$ we get the same topology. By the way, this topology is induced on the Euclidean matrix norm $\lVert (a_{ij}) \rVert = \sqrt{\sum_{i,j} a_{ij}^2}$.
Q2:
This has been already answered. Generally speaking, if we have a bijection $\beta : S \to X$ from a set $S$ to a topological space $X$, we can give $S$ the unique topology making $\beta$ a homeomorphism. Explicitly, if $\mathcal T_X$ is the topology of $X$, then we give $S$ the topology $\mathcal T_S = \beta^{-1}(\mathcal T_X) = \{ \beta^{-1}(U) \mid U \in \mathcal T_X \}$.
Q3:
The general linear group $\operatorname{GL}(n,\mathbb R)$ is a certain subset of $\mathbb{R}^{n\times n}$. Since we endowed $\mathbb{R}^{n\times n}$ with a topology, $\operatorname{GL}(n,\mathbb R)$ becomes a topological space (with the subspace topology inherited from $\mathbb{R}^{n\times n}$). In fact it is an open subspace of $\mathbb{R}^{n\times n}$ and therefore a manifold.