Functional Analysis 21 | Isomorphisms [dark version]

Isomorphisms are the yardstick for when two mathematical spaces are “the same” in a structural sense—no information is lost, and the structure is preserved in both directions. The core idea starts with homomorphisms, which are structure-preserving maps: for vector spaces, that means preserving addition and scalar multiplication; for metric spaces, it means distances can only shrink (never grow). An isomorphism tightens this by requiring bijectivity and the same structure preservation for the inverse map as well, so the spaces become interchangeable for the purposes of the theory.

For vector spaces, a homomorphism is a linear map: scaling commutes with the map, expressed as F(λx)=λF(x), and addition commutes as well, so mapping after adding equals adding after mapping. For metric spaces, the structure is the distance function. If a map keeps distances exactly the same, it preserves all metric information—but that’s a very strong requirement. A weaker, more flexible homomorphism condition is that distances may decrease but not increase, ensuring the map never stretches the space.

An isomorphism is then described as a special homomorphism that is bijective, with the inverse also preserving the relevant structure. In the vector-space setting, this becomes a bijective linear map; since the inverse of a linear bijection is automatically linear, the “both directions” requirement is automatic. In the metric setting, the map must preserve the metric in both directions, which effectively means the two metric spaces look identical up to relabeling: points correspond through the map, and all distance relations match.

The transcript then highlights a common source of confusion in normed (Banach) spaces, where extra structure is present. Because a Banach space is simultaneously a vector space and a normed space, “isomorphism” could be misread as merely a linear bijection. But the norm must also be preserved to match the metric-style notion of structure preservation. That leads to the term “isometric isomorphism,” meaning a bijective linear map that satisfies the norm equality (so distances induced by the norm stay the same). In other words, the correct Banach-space isomorphism respects the full normed structure, not just the algebraic one.

Finally, concrete examples are used on ℓ^p spaces. A right-shift operator on sequences indexed by N (one-sided sequences) is linear and preserves the ℓ^p norm, but it fails to be an isomorphism because it is not surjective: the image always begins with a zero, so not every sequence can be reached. When the index set is changed to Z (two-sided sequences), the right shift becomes bijective while still preserving the ℓ^p norm. In that infinite-dimensional Banach-space context, the map becomes an isometric isomorphism. The transcript closes by noting that the need for these isometric isomorphisms in Banach spaces will be revisited in the next installment.