ΘρϵηΠατπ

Uniform Convergence Equivalence In Terms of Lim Sup
Suppose that S⊆ℝn and a sequence of functions (fk):ℕ1→C(S,ℝm), then fk→f if and only if eventualy we have fk−f∈Cb(S,Rm) for all k sufficiently large and limk→∞⁡‖fk−f‖∞=0
Dini's
If f and fn are continuous functions on [a,b] such that fn≥fn+1 for all n∈ℕ1 and if fn→pw⁡f then fn→unif⁡f

We first prove that for any x0∈[a,b] and ϵ∈ℝ+ there exists some integer N∈ℕ1 and r∈ℝ+ such that |x−x0|≤r⟹|gN(x)|≤ϵ where gn(x)=f(x)−fn(x)

This follows from a few facts, we first start with continuity of f,fn, specifically at x0 and therefore using ϵ we obtain some δ1∈ℝ+ such that f(B(x0,δ1))⊆B(gn(x0),ϵ), similarly we have a δ2∈ℝ+ such that f(B(x0,δ2))⊆B(f(x0),ϵ), then also we know that since fn→pw⁡f then since it's true at x0 and using ϵ we obtain some K∈ℕ1 such that for all n∈ℕK we have |fn(x0)|−f(x0)≤ϵ

Now we combine using the triangle inequality, so let ϵ∈ℝ+ , then let r=min⁡(δ1,δ2) so that we obtain the inequalities mentioned in the previous paragraph, and we get some K∈ℕ1 as mentioned in the previous paragrah, then |gK(x)|=|f(x)−fK(x)|≤|f(x)−f(x0)|+|f(x0)−fK(x0)|+|fK(x0)−fK(x)|≤3ϵ

With that out of the way suppose for the sake of contradiction that fn doesn't convege uniformly to f, since f−fn is eventually bounded then this is equivalent to limn→∞⁡‖f−fn‖∞=d∈ℝ+

Since gn is continuous as the sum of continuous functions and has compact domain then by EVT, we know that it attains its max and min, and therefore there exists a xk∈[a,b] such that ‖f−fk‖∞=gk(xk), so we have limk→∞⁡gk(xk)=d, and note that this means that for any ϵ′∈ℝ+ there is some J∈ℕ1 such that for all j∈ℕJ we have |gj(xj)−d|<ϵ′ which implies that d−ϵ′≤gj(xj).

Since xk is in a compact set it has a convergent subsequence xnk→x∈[a,b]. Now recall the fact we proved initially, by using d−ϵ we know that there is some N∈ℕ1 such that gN(x)<d−ϵ, now at the same time the inequality discussed in the previous paragraph also holds on the subsequence (this is because if a sequence goes somewhere, then any subsequence must also go there), namely we have some J∈ℕ1 such that for all j∈ℕJ d−ϵ≤gnj(xnj) but then there is some i such that ni>max⁡(N,J) and therefore gi<gN and so chaining inequalities we have d−ϵ≤gni(xxi)≤gN(x)≤d−ϵ which is a contradiction, and therefore fn→unif⁡f