Let be a positive measure on the set , and a measurable function, then we can define
For these functions are a norm. The only non-trivial condition to check is the triangle inequality: Minkowsky’s inequality.
Also if a.e., so it’s a semi-norm. Just work with the equivalence classes, where iff a.e.
In the case of finite measure space, there is a nice inclusion property: Proposition Suppose is a finite measure space and . Then
in particular .
Oss In a sense is the limit of when , for a finite measure space, since:
on the other hand, for any the set
has positive measure. This implies
taking the on gets that .
Proof Let , Apply Holder inequality to and . Since we want on the right the norm, use as indices and its conjugated:
Remark the result is true only for finite measure spaces. A cool counterexample, in which also the opposite inclusion when is also true is equipped with the counting measure. Indeed
Proposition Let . The space is complete: suppose that is a sequence such that , there exists such that
(The sequence is Cauchy). Then there exists an element such that
Proof Suppose now that . First we build a limit candidate, then we show that this limit belongs to . Since the sequence is Cauchy, we can fine a subsequence such that
to simplify the notation we identify the subsequence as the original sequence.
Clearly one can write with a telescopic series
we now define the candidate limit function as a series
and similalry
Note that . Applying the Minkowsky’s inequality and taking the limit we get
using Monotone Convergence Theorem
so that . This implies that is finite a.e., so that is absolutely convergent a.e., meaning that it’s also convergent a.e. and the definition of makes sense. Also and we can conclude that also .
By construction
we need to prove that this converges in also
taking the limit and using Dominated convergence theorem we get
The last step is to show that also the original sequence converges to .
since there exists an such that the first term on the left is less than . Since the original sequence was Cauchy there also exists an such that the second term on the right is less than . We can conclude that
The completeness of is straight forward, since the sequence is Cauchy there exists such that it holds almost everywhere that
For a fixed these sets of measure zero where the inequality does not holds can depend on and , consider the (contable) union
this set has also measure zero. Clearly for each we have pointwise convergence (since is complete), hence it makes sense to define as the pointwise limit of on this set. The convergence is also uniform on , so that
Dense subsets
Note
Let . Simple functions are dense in : for any there exists a sequence of simple functions such that:
Note
f = f_+ - f_-
, then we can assume w.l.o.g. thatf \geq 0
. By the definition of Lebesgue's integral, we can find a sequence of simple functions such thatWrite
So the case is proven. To prove the general case observe that:
Since pointwise, using Dominated convergence theorem:
Note
Let . Test functions are dense in : for any there exists a sequence such that:
Note
By definition