1-2. In SiLENSe simulations, a ladder-like structure is formed in the MQW
active region with approximately equal voltage drop between the adjacent
QWs. That is why the current drops down with the QW number. Note that the
highest barrier is between the n-GaN and first IngaN QW. Then a ladder-like
structure goes down (see a schemeattached).

In our view, the problem is related to the fact that barrier thickness is
compared with the free path (at least for electrons). Therefore, then an
electron overcomes the first barrier, it can not only be captured in the
first QW, but also can fly to the next QWs. On the contrary, the
drift-diffusion model implied that electrons immediately lose their kinetic
energy and are captured by the first QW. To reach next QW, they need again
to overcome the barrier, which is much higher than the thermal energy. So a
high potential drop is needed to provide significant electron flux through
the MQW region.

On the contrary, in experiments there is no significant dependence on the
turn-on voltage on the QW number. Besides, the light output is slightly
increased with the QW number.

In my view, experiments support the idea that as far as the electrons are
injected in the MQW region, they do not need a big potential drop to travel
from one QW to another. The most of recombination occurs in the QW closest
to the p-region, because holes has much lower mobility that the electrons,
and do not have time to spread to the next QWs before recombination. SiLENSe
simulations also predict the same distribution of the recombination. While
the light output increase is completely due to IQE rise, because the crystal
quality of the first 1-3 QWs is lower than for the next ones. In my opinion,
this is the reason why most of people use MQW structures. Please find
attached a recent paper by Lumileds people. Their conclusions supports the
above consideration. Besides, people from MOVPE group in Ioffe Institute
also feel that the light is actually comes from the top QW only, while
botton (former) QWs to be grown just to improve the crystal quality.
However, the only way to reflect such a situation in SiLENSe simulations is
manual specification of different non-radiative recombination parameters in
different QWs.

To conclude, I think that the problem with MQW structures is a general
drawback of the drift-diffusion approach, and, therefore, it exists in all
other simulators based on drift-diffusion model. We are thinking what may be
done to overcome it.