3

I'm curious if people study analysis while using fields that are not $\mathbb{R}$. I remember seeing a post about doing analysis on $\mathbb{Q}$, but $\mathbb{Q}$ is not complete! Mostly I'm interested in applying functional analysis with differing underlying fields. The fields I have in mind are like field extensions of a given polynomial, or a given set of polynomials. I'm unsure if one can construct a complete field in this way (without constructing the reals). This seems to dip into algebra, analysis and topology so for me, it's probably a bit to advanced to actually make any progress with alone.

Has anyone done analysis in this sort of way? Are there any applications whatsoever (am I being incredibly silly considering this)?

DanZimm
  • 5,861
  • 2
  • 23
  • 39

2 Answers2

6

If you want to do analysis over some field $\mathbb{K}$ you need some kind of absolute value on it (sometimes called a valuation). There are only two types of valuations: archimedean and non-archimedean. And this claim is not a tautology, see lemma 1.2 in Local fields. J. W. S. Cassels. Simply speaking archimedean valuations are good and non-archimedean are weird.

By Ostrowski's theorem, there are only two archimedean valued fields: $\mathbb{R}$ and $\mathbb{C}$. So this part of analysis is very well understood. The remaining part is usually called non-archimedean. It is not so well studied but, there is a lot of literature on this question already. The main example of a complete non-archimedean field is $\mathbb{Q}_p$. It is so important that non-archimedean analysis is sometimes called the $p$-adic analysis.

Long story short, a lot of things in non-archimedean analysis is significantly trivializes or not true in general. Examples of drawbacks:

  • non-archimedean fields have no natural $<$ relation. As a consequence, you cannot build a good theory of integration.

  • non-archimedean fields have no natural involution (like $\mathbb{C}$), so you cannot make a bunch of things in complex analysis. For the same reason, you cannot build a decent theory of involutive algebras and $C^*$-algebras.

  • non-archimedean fields can be non-separable and not even locally compact.

  • Liouville's theorem is true only for algebraically closed non-archimedean fields.

  • A non-archimedean field must be spherically complete in order to have Hahn-Banach theorem at hand. So for non-spherically complete fields, you cannot do a lot of things in functional analysis.

  • Non-archimedean fields do not always possess a Haar measure on non-discrete locally compact groups, so in most interesting cases you cannot do abstract harmonic analysis.

  • Non-archimedean fields have no analog of Gelfan-Mazur's theorem, so a big part of the spectral theory is unavailable.

  • Non-archimedean local fields are totally disconnected, so continuity theory evolves in a different manner.

  • Non-archimedean fields are so restrictive, that most non-archimedean Banach spaces are linearly topologically isomorphic to $c_0(S)$ for some index set $S$. Therefore, locally convex spaces are of much more interest than Banach spaces.

Of course, these drawbacks make non-archimedean analysis interesting in some sense, because a researcher has the challenge to extend the well-known theories to the cases where all standard tools do not work.

If you are mainly interested in functional analysis I recommend taking a look at this discussion. For the first reading I recommend Non-archimedean functional analysis A. C. M. van Rooij and for the more modern treatment Nonarchimedean functional analysis P. Schneider and after that Locally convex spaces over non-archimedean valued fields. C. Perez-Garcia, W. H. Schikhof.

If you don't have enough time to read all that books take a look at this presentation Introduction to non-archimedean functional analysis. W. Sliwa

Norbert
  • 58,398
  • A couple points to clarify what you said: First, p-adic fields do possess a locally compact topology induced by the p-adic evaluation, so we do have a Haar measure and thus a measure integration theory, though we don't have the nice sets on which to integrate that we are used to, and thus links between differentiation and integration break down.

    Second, the claim that totally disconnected spaces have totally uninteresting continuity properties is a little strange. For example, continuity on the rationals is obviously weaker and harder to work with than on the reals, but it is not vacuous .

    – jxnh May 23 '14 at 06:54
  • @JHance ok I changed that phrase – Norbert May 23 '14 at 07:33
1

You might look up p-adic analysis.

Robert Israel
  • 470,583