touschek_scatter calculation appears not to take closed orbit into account

[GPenn]

I am having trouble using the touschek_scatter on some lattices with large errors.
When I calculate the momentum aperture, it is significantly worse than the nominal
machine, but looks like what I expect. There is also good agreement with calculations done with AT.

However, when I do the lifetime calculations by Monte Carlo method using touschek_scatter,
the lifetime is significantly worse than the formula based on the momentum aperture.
What is especially noticeable is that a lot of the TSCATTER elements report losing all
(or almost all) of the scattered test particles. I routinely use
Momentum_Aperture_scale = 0.75
so by the definition of the MA, one would expect most of the particles between
the lower bound and the actual MA to survive.

The only explanation I can think of is that the touschek_scatter routine is not
shifting the scattered particles to be centered about the closed orbit.
The particle data saved using "bunch = %s-%03ld.bun" is centered about 0 in all
transverse coordinates. Of course, the closed orbit parameters could be subtracted out
when saving to the file.

I have done some test runs by comparing a lattice with small errors that works fine
with the same lattice that has an orbit bump added to it, which shifts the closed
orbit vertically within that region. In the region of the orbit bump, the same behavior
appears where almost every scattered particle is lost, while outside of the bump
at most about 2/3 of the test particles are lost (including the weighting factor,
only about half of the actual scattering rate turns into losses).

Thus, it seems that neither the x,y coordinates nor the transverse angles of the
closed orbit are being used to orient the scattered particles.
The MA calculation, on the other hand, does show some degradation due to the
orbit bump but there are no extreme effects even when the closed orbit
offsets are comparable to the dynamic aperture at that location.
The reduced MA values are used as inputs to the touschek_scatter calculation anyway.

I'm sure it is atypical to have closed orbits so far off-axis, so this would not
be noticeable in most cases.

It could be that the part of the code that backtracks the particles to s=0
is causing the problem, but in either case the touschek_scatter routine would
have to be doing something differently from the momentum_aperture calculation.

I am using elegant 2021.4.0, on a Mac and also on one of the LBNL clusters.

Files are attached; probably just the PDFs and the files related to 'test_MC.ele' are going to be useful.

Thanks,
Gregg

[michael_borland]

Gregg,

I think you are right about this. It should be relatively simple to solve, so I'll try to get it into the next version.

Thanks--Michael