Transmission spectrum with trench

[perfetti]

Dear Everyone,
        I am sorry to ask so many questions, but I couldn't figure it out by myself, and usually I ask after I tried a lot.Still thanks for your time!
        I got an transmission spectrum like following. It has a lot of trenches in it. I checked the electrode length, and it's OK. And I changed the exchange correlation from PZ,LDA to GGA,PBE, but it has no big effect.
        1)I am not sure if I got the right transmission spectrum?
        2)What's the difference between PZ,LDA and GGA,PBE? Will it affect a lot on the result? For my case, PBE,GGA method is 10 times slower. But I see all others they are using this method.
        
        

[zh]

You calculated transmission spectrum looks reasonable, because it exhibits the usual features for a semiconducting system. Around the Fermi level, there is a gap. 

Both LDA and GGA would underestimate the band gap of a semiconductor. They are at the same level. This is why you observe that no significant difference obtained by these two exchange correlation functionals. 

[perfetti]

Thanks Mr. Zh.

I was wondering about the trench in the right part cause it's not nicely looking.

What should I use if LDA and GGA would underestimate the bandgap of a semiconductor?

You calculated transmission spectrum looks reasonable, because it exhibits the usual features for a semiconducting system. Around the Fermi level, there is a gap. 

Both LDA and GGA would underestimate the band gap of a semiconductor. They are at the same level. This is why you observe that no significant difference obtained by these two exchange correlation functionals. 

[nori]

I was wondering about the trench in the right part cause it's not nicely looking.

If the central region of your structure is the same as the electrodes and gamma point approximation is used, the transmission coefficient at each energies is coincident with the number of Bloch state at each energies.
So the trench can be possible.

What should I use if LDA and GGA would underestimate the bandgap of a semiconductor?

DFT+U method and meta-GGA are sorts of a prescription.