MHB Lower SemiContinuity Problem: Proofs & Step Funcs

  • Thread starter Thread starter joypav
  • Start date Start date
Click For Summary
The discussion centers on proving that a function f is lower semi-continuous at a point x_0 if and only if f(x_0) is less than or equal to the limit inferior of f as x approaches x_0. The proof involves demonstrating that if f is lower semi-continuous, then for any ε > 0, f(x) remains above f(x_0) - ε within a certain δ-neighborhood of x_0, leading to the conclusion that f(x_0) ≤ liminf f(x). Conversely, if f(x_0) ≤ liminf f(x), it can be shown that f must be lower semi-continuous at x_0. Additionally, the discussion explores the properties of step functions, establishing that a step function is lower semi-continuous if its value at the partition points is less than or equal to the minimum of the values in adjacent intervals. The conversation also clarifies definitions and addresses confusion regarding the formulation of limit inferior.
joypav
Messages
149
Reaction score
0
Problem:
(a)
Show that $f: E \rightarrow R$ is lower semi-continuous at $x_0 \in E$ if and only if
$f(x_0) \leq \liminf_{x \rightarrow x_0}f(x)$

Proof:
"$\rightarrow$"
Assume f is l.s.c. at $x_0 \in E$.
$\implies \forall \epsilon > 0, \exists \delta >0$ such that if $\left|x-x_0 \right|<\delta$, then $f(x_0)-\epsilon \leq f(x)$

Consider $\liminf_{x \rightarrow x_0}f(x) = \liminf_{n \rightarrow \infty}\left\{f(x): \left| x-x_0 \right|<1/n\right\}$.

Let $y \in R, y = \liminf_{x \rightarrow x_0}f(x) \in R$. Then $y = f(x')$ for some $x' \in E$.

Because f is l.s.c. at $x_0$,
$\forall \epsilon > 0, \exists \delta >0$ such that if $\left|x'-x_0 \right|<\delta$, then $f(x') \geq f(x_0)- \epsilon$
$\implies \liminf_{x \rightarrow x_0}f(x) \geq f(x_0) - \epsilon$

Let $\epsilon \rightarrow 0$
$\implies \liminf_{x \rightarrow x_0}f(x) \geq f(x_0)$.

"$\leftarrow$"
Assume $f(x_0) \leq \liminf_{x \rightarrow x_0}f(x)$.
I am thinking I should proceed with proving by contradiction? Meaning, assume there is an $x_0$ where f is not l.s.c. (b)
A real-valued function $\phi$ defined on an interval $[a, b]$ is called a step function if there is a partition $a = x_0 < x_1 <...< x_n = b$ such that for each i the function $\phi$ assumes only one value in the interval $(x_{i−1}, x_i)$. Show that a step function $\phi$ is lower semi-continuous if and only if $\phi(x_i)$ is less than or equal to the smaller of the two values assumed in $(x_{i−1}, x_i)$ and $(x_i, x_{i+1})$.

For this part I think I am just confused by the definition given for step function. However, I am still thinking about the problem. So I will add onto the post if I make some headway.
 
Physics news on Phys.org
Please state the definition you're using for l.s.c. functions. Also, are we assuming that $E$ is a subset of $\Bbb R$?
 
Euge said:
Please state the definition you're using for l.s.c. functions. Also, are we assuming that $E$ is a subset of $\Bbb R$?

Sorry! Yes, E is a subset of R.

The definition we were given is what I used in the proof:
A function f is l.s.c. on E if $\forall x_0 \in E$,
$\forall \epsilon > 0, \exists \delta >0$ such that if $\left|x-x_0 \right|<\delta$, then $f(x_0)-\epsilon \leq f(x)$
 
You do not seem to have the correct formulation of the inferior limit $\liminf\limits_{x\to x_0} f(x)$. If $x_0$ is a limit point of $E$, $\liminf\limits_{x\to x_0} f(x)$ is defined as $$\sup\limits_{\delta > 0}\, \inf\limits_{x\in E,\, 0 < |x - x_0|< \delta} f(x)$$ which is the same as $$\lim\limits_{\delta \to 0}\, \inf\limits_{x\in E,\, 0 < |x - x_0| < \delta} f(x)$$ Now suppose $f$ is l.s.c. at $x_0$. If $\epsilon > 0$, there exists a $\delta > 0$ such that for all $x$, $|x - x_0| < \delta$ implies $f(x_0) - \epsilon \le f(x)$. So for $0 < \eta < \delta$, $f(x_0) - \epsilon$ is a lower bound for the of values $f(x)$ where $x\in E$ and $0 <|x - x_0| < \eta$. Hence $f(x_0) - \epsilon \le \inf\{f(x):x\in E,\, 0 < |x - x_0| < \eta\}$ for all $0 < \eta < \delta$. Therefore $f(x_0) - \epsilon \le \liminf\limits_{x\to x_0} f(x)$. As $\epsilon$ was arbitrary, $f(x_0) \le \liminf\limits_{x\to x_0} f(x)$.

Conversely, suppose $f(x_0) \le \liminf\limits_{x\to x_0} f(x)$. Set $L = \liminf\limits_{x\to x_0} f(x)$. Fix $\epsilon > 0$, and choose a positive number $\delta$ such that for all $x\in E$, $0 < |x - x_0| < \delta$ implies $f(x) > L - \epsilon$. By assumption $f(x_0) \ge L$, so $f(x) > f(x_0) - \epsilon$ whenever $|x - x_0| < \delta$. Hence, $f$ is l.s.c. at $x_0$.
 
Thank you!
Here is what I have for part b.

"$\leftarrow$"
Assume $\phi(x_i)$ is less than or equal to the smaller of the two values assumed in $(x_{i-1},x_i),(x_i,x_{i+1})$.

Case I: $x_0 \in (x_{i-1},x_i)$, some i
Then, choose $\delta = \frac{x_i-x_0}{2}$
If $\left| x-x_0 \right| < \delta = \frac{x_i-x_0}{2} \implies x \in (x_{i-1},x_i) \implies \phi(x)=\phi(x_0)$

Then, $\forall \epsilon > 0, \phi(x_0) - \epsilon \leq \phi(x_0) \implies \forall \epsilon > 0$ and for $\delta = \frac{x_i-x_0}{2}, \phi(x_0) - \epsilon \leq \phi(x)$
$\implies \phi$ is lower semi-continuousCase II: $x_0$ is an endpoint of an interval, WLOG say $x_0 = x_i$, some i
for $\forall x \in (x_{i-1}, x_i), \phi(x) = a$
for $\forall x \in (x_i, x_{i+1}), \phi(x) = b$
WLOG, assume a<b. Then, by assumption, $\phi(x_i) \leq a$.

Then, for $\delta = x_i - x_{i-1}$,
$\forall \epsilon>0$, if $\left| x-x_0 \right| = \left| x-x_i \right| < \delta \implies x \in (x_{i-1}, x_i)$ or $x \in (x_i, x_{i+1})$

if $x \in (x_{i-1}, x_i)$, then
$\phi(x_0) - \epsilon = \phi(x_i) - \epsilon \leq a - \epsilon < a = \phi(x) \implies \phi(x_0) - \epsilon \leq \phi(x)$

if $x \in (x_i, x_{i+1})$, then
$\phi(x_0) - \epsilon = \phi(x_i) - \epsilon \leq a - \epsilon < a < b = \phi(x) \implies \phi(x_0) - \epsilon \leq \phi(x)$

$\implies \phi$ is lower semi-continuous"$\rightarrow$"
Assume $\phi$ is lower semi-continuous
$\implies$ for every $x_0 \in [a, b], \forall \epsilon>0, \exists \delta>0, \left| x-x_0 \right|<\delta \implies \phi(x_0) - \epsilon \leq \phi(x)$
By way of contradiction, assume $\phi(x_i)$ is greater than the smaller of the two values assumed in $(x_{i-1},x_i),(x_i,x_{i+1})$.

for $\forall x \in (x_{i-1}, x_i), \phi(x) = a$
for $\forall x \in (x_i, x_{i+1}), \phi(x) = b$
WLOG, assume a<b. Then, by assumption, $\phi(x_i) > a$.

Choose $x_0=x_i$.
Then, if $x<x_0$,
$\phi(x_0) - \epsilon = \phi(x_i) - \epsilon > a - \epsilon = \phi(x) - \epsilon$
$\implies \phi(x_0) - \epsilon > \phi(x)$, a contradiction

$\implies \phi(x_i) \leq a$
 
We all know the definition of n-dimensional topological manifold uses open sets and homeomorphisms onto the image as open set in ##\mathbb R^n##. It should be possible to reformulate the definition of n-dimensional topological manifold using closed sets on the manifold's topology and on ##\mathbb R^n## ? I'm positive for this. Perhaps the definition of smooth manifold would be problematic, though.

Similar threads

  • · Replies 11 ·
Replies
11
Views
2K
Replies
1
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 14 ·
Replies
14
Views
3K
  • · Replies 38 ·
2
Replies
38
Views
5K
Replies
32
Views
3K
Replies
7
Views
2K
  • · Replies 0 ·
Replies
0
Views
2K
  • · Replies 4 ·
Replies
4
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K