Highlights of What’s New in Version 2021.4

Here is a summary of what’s changed since release 2021.3.

A memory leak was fixed in the FMULT element.
As pointed out by P. Anisimov, there were problems with the IBSCATTER element when more than one was inserted in the beamline. This was fixed. In addition, the code now detects negative growth rates (typically in the transverse planes) and correctly refuses to operate in “random” mode, since that mode is only able to increase the emittances.
As pointed out by G. Penn (LBNL), the KSEXT element gave incorrect results for synchrotron radiation if K1 or J1 was significant. This was fixed. This change resulted in some reorganization of common routines used by KQUAD, KSEXT, KOCT, and CCBEND, so small differences in results may be seen.
The CALIBRATION factor on HKICK, VKICK, EHKICK, and EVKICK and the HCALIBRATION and VCALIBRATION factors on HVKICK and EHVKICK was applied twice in the computation of the response matrices for trajectories. That is, the actual calibration factor would be the square of the given factor.
A memory bug was found in CSBEND that would sometimes cause a crash when REFERENCE_CORRECTION=1 if the moments_output command was given. This was reported by G. Penn (LBNL).

As pointed out by G. Penn (LBNL), the lengthError parameter in the closed_orbit output file was always zero, even when fixed_length=0. This was fixed.

Added the FACTOR parameter for the FMULT element, allowing all multipole components to be scaled by a common factor. Also added the UNTILTED_MATRIX parameter, which can be used to force elegant to numerically compute the matrix for the untilted element; see the manual page for discussion.
Added the PARALLEL_INTEGRATION parameter to IBSCATTER, which permits disabling the default parallelization of s-dependent integrals for intrabeam scattering.
The WATCH element has a new parameter AUTO_REFERENCE that allows automatically determining the reference frequency from the rf cavities.

The rf_setup command now does a better job of setting the rf frequency in the presence of a closed orbit, which may have a different length than the nominal orbit. It also includes the ability to offset the rf phase, which allows the user to compensate for small differences between the calculation of the energy loss per turn from radiation integrals and tracking. These improvements were inspired by questions from G. Penn (LBNL).
The save_lattice command can now write the entire lattice as a single beamline, if output_seq=2. This was requested by forum user blanco-garcia.

Added the multiply_np_by_cores parameter to bunched_beam and bunched_beam_moments. If non-zero, then the number of particles generated is n_particles_per_bunch times the number of working cores.