Is it trivial but advantageous to let emc2 output the averaged physical quantity from its own cluster expansion in a Monte Carlo simulation as another column?
If the CE is T-independent, that just the "energy" column (column 3 in emc2, if you specif -g2c).
If the CE is T-dependent, then you are right it does not output that, but that average would not have a physical meaning.
I meant to say physical quantities more than energy, like band gap, or something more, which can totally be T-independent.
oh I see. so using another CE. Since both emc2 and memc2 output the correlations, it should be possible to easily compute this with a simple awk or perl script, e.g.
cat mc.out | awk ‘{print $18empty_eci+$19point_eci+$20*1nn_pair_eci+…}’
I know it would be nice to have this within the code itself… but sorry I am time-constrained 
This is great! I hadn’t thought about any use of those columns, although it doesn’t allow for a different cluster set optimized for that very quantity, like https://scitation.aip.org/content/aip/jo … /1.3602149
. But I should be contented…
You could always build a set of cluster that is the union of the sets of clusters of each CE and set the appropriate eci to 0…
Do you mean a subset of the cluster set used in MC?
Yes, that’s another way to say it.