To add some context, let $(\Omega,\mathcal{F},\mu)$ be a measure space. The statement of the Extended Dominated Convergence Theorem(EDCT) is:
$\lbrace f_{n}\rbrace_{n\geq 1}$ be a sequence of functions from $L^{1}(\Omega,\mathcal{F},\mu)$ such that $f_{n}\to f$ a.e.($\mu$) and there exists a sequence $\lbrace g_{n}\rbrace_{n\geq 1}\in L^{1}(\Omega,\mathcal{F},\mu)$ such that $g_{n}\to g$ a.e.($\mu$), $|f_{n}|\leq g_{n}$ a.e.($\mu$) for all $n\geq 1$ and $\int g_{n}d\mu\to\int gd\mu$. Then, (i) $f\in L^{1}(\Omega,\mathcal{F},\mu)$ (ii) $f_{n}\xrightarrow{L^{1}} f$ (iii) $\int f_{n}d\mu \to \int fd\mu$
I'm stuck while trying to find a counterexample to the above theorem such that, $\lbrace f_{n}\rbrace_{n\geq 1}, f\in L^{1}(\Omega,\mathcal{F},\mu), f_{n}\to f$ a.e.($\mu$) and $\int f_{n}d\mu\to\int f d\mu$ but $||f_{n}-f||_{L^{1}(\Omega,\mathcal{F},\mu)}\not\to 0$.
If the sequence of functions is non-negative, we can see $|f_{n}-f|\leq f_{n}+f=g_{n}$. Also, $\int g_{n}d\mu=\int f_{n}d\mu+\int fd\mu\to2\int fd\mu$. So applying the EDCT we cannot find a counter-example if the the sequence of functions is non-negative.
So is there any such example that shows that a sequence $g_{n}$ that dominates $f_{n}$ is necessary for convergence in $L^{1}$-norm. Thanks in advance for any help.
Counter-example: Consider the measure space $(\mathbb{R},\mathcal{B}(\mathbb{R}),m)$, where $m$ is the usual Lebesgue measure. Let $f_{n}:\mathbb{R}\rightarrow\mathbb{R}$ be defined by $f_{n}(x)=n^{2}1_{(\frac{1}{2n},\frac{1}{n})}(x)-n^{2}1_{(-\frac{1}{n},-\frac{1}{2n})}(x)$. Define $f:\mathbb{R}\rightarrow\mathbb{R}$ by $f(x)=0$. Clearly $f_{n}\rightarrow f$ pointwisely. Moreover $\int f_{n}dm=\int dm=0$. However $||f_{n}-f||_{1}\not\rightarrow0$.
/////////////////////////////////////////////////////////
However, we have:
Let $(X,\mathcal{F},\mu)$ be a measure space. Let $(f_{n})$ be a sequence of real-valued integrable functions defined on $X$. Let $f:X\rightarrow\mathbb{R}$ be an integrable function. Suppose that $f_{n}\rightarrow f$ a.e., then $\lim_{n}\left(||f_{n}||-||f||-||f_{n}-f||\right)=0$, where $||f||=\int|f|d\mu$ etc...
We go to prove the above assertion. Observe that $|f_{n}-f|\leq|f_{n}|+|f|$, so $|f_{n}|-|f|-|f_{n}-f|\geq-2|f|.$ On the other hand, $|f_{n}|\leq|f_{n}-f|+|f|$ and hence $|f_{n}|-|f|-|f_{n}-f|\leq0\leq2|f|$. Combining, we have $\Biggl||f_{n}|-|f|-|f_{n}-f|\Biggr|\leq2|f|$. Clearly, $|f_{n}|-|f|-|f_{n}-f|\rightarrow0$ a.e.. By Dominated Convergence Theorem, we have that $\int\left(|f_{n}|-|f|-|f_{n}-f|\right)d\mu\rightarrow 0$. That is, $||f_{n}||-||f||-||f_{n}-f||\rightarrow0$.
Remark:
In particular, if $||f_n||\rightarrow ||f||$, then $||f_n-f||\rightarrow 0$
Counter-example does not exist if $f_n$ and $f$ are non-negative.