Recently Ilya looked at the BEMC sampling fraction with the LOW_EM option enabled in starsim.  The nontrivial differences between the nominal GEANT tune and LOW_EM conflicted with previous studies that saw no difference when lowing the GEANT thresholds in the barrel.  What's going on?

Nominal GEANT tuning for the barrel is defined in pams/geometry/calbgeo/calbpars.g.  Absorbing material is defined with

• CUTGAM = 0.00008
• CUTELE = 0.001
• BCUTE = 0.0001
• CUTNEU = 0.001
• CUTHAD = 0.001
• CUTMUO = 0.001

while sensitive volumes have

• CUTGAM = 0.00001
• CUTELE = 0.00001
• CUTNEU = 0.001
• CUTHAD = 0.001
• CUTMUO = 0.001

Now the LOW_EM option, defined in pams/geometry/geometry/geometry.g, lowers a suite of GEANT thresholds in all volumes,

• CUTGAM = 0.00001
• CUTELE = 0.00001
• BCUTE = 0.00001
• BCUTM = 0.00001
• DCUTE = 0.00001
• DCUTM = 0.00001

Running with a fresh copy of the geometry from DEV with nominal cuts confirms the previous behavior roughly in agreement with the sampling fraction in the slow simulator,

Enabling the LOW_EM option, however, also confirms Ilya's study: the sampling fraction changes significantly.

The earlier study lowered only those thresholds defined in the barrel geometry.  In particular, only CUTGAM, CUTELE, and BCUTE in absorber volumes was tuned.  Indeed, lowering these values (what we'll call Tune Two) does not change the sampling fraction,

 The difference must be in one of the other cuts modified in LOW_EM.  Muonic effects are unlikely to be significant in electromagnetic showers, which brings the focus to the delta ray threshold DCUTE.  Lowering this threshold in both absorber and sensitive volumes (what we'll call Tune One) not only changes the sampling fraction, but it overshoots the LOW_EM values,

Upon further thought, lowering DCUTE alone is a bit naive.  In order to allow delta rays to really do their thing, CUTELE should also be lowered to allow the delta rays to fully propagate.  This tune, say Tune Three, does the trick.

The full results for all tunes across three energy ranges confirms the results,

Is there an advantage to the delta ray cuts alone verses LOW_EM?  There may be a computational advantage to the former, but initial tests are so far inconclusive.

UPDATE:   

The delta ray cuts seems to have a slight computation advantage over LOW_EM on average, but the difference is small enough that the added complications of changing the BEMC geometry may not be necessary.