Organized by GPT-6-Astra. SPONGE 2.0 is under active, iterative development; features, parameters, and documentation may change. Consult the GitHub repository’s docs and skills for the latest version information.

Source snapshot for this page.

Advanced Parameters

Device and Parallelism

Parameter	Type	Default	Description
device	string	auto-detected	GPU device ID
device_optimized_block	int	-	GPU block size optimization

Domain Decomposition

MPI domain decomposition parameters are read from the [DOM_DEC] scope:

[DOM_DEC]
update_interval = 100
split_nx = 2
split_ny = 2
split_nz = 1

Parameter	Scope	Type	Default	Description
update_interval	DOM_DEC	int	100	Domain-decomposition refresh interval
split_nx	DOM_DEC	int	auto	Number of splits in X
split_ny	DOM_DEC	int	auto	Number of splits in Y
split_nz	DOM_DEC	int	auto	Number of splits in Z

When the split counts are omitted, SPONGE chooses them automatically from the PP MPI size and box geometry. Current MPI domain decomposition only supports orthogonal boxes.

Minimization

Minimization-specific parameters use the [minimization] scope:

mode = "minimization"

[minimization]
dynamic_dt = 1
max_move = 0.1
beta1 = 0.9
epsilon = 1e-4

Parameter	Scope	Type	Default	Description
dynamic_dt	minimization	int	1	Use the ADAM-like dynamic-step minimizer
max_move	minimization	float	0.1	Maximum coordinate displacement per step in angstrom
momentum_keep	minimization	float	0.0 when dynamic_dt = 0	Momentum retention factor in gradient-descent mode
beta1	minimization	float	0.9	First ADAM parameter used when dynamic_dt != 0
beta2	minimization	float	0.9 in the current source	Second ADAM parameter used when dynamic_dt != 0
epsilon	minimization	float	1e-4	Numerical stabilizer used when dynamic_dt != 0

When dynamic_dt != 0, SPONGE internally resets dt to 3e-4 ps. When dynamic_dt = 0, it uses a tiny fixed dt and the gradient-descent path.

Rerun Mode

rerun uses the normal top-level trajectory file keys plus several rerun-specific flat keys:

mode = "rerun"
crd = "traj.dat"
box = "traj.box"
vel = "traj.vel"
frame_limit = 1000
rerun_start = 0
rerun_strip = 0
rerun_need_box_update = 0

Parameter	Scope	Type	Default	Description
crd	top-level	string	required	Input coordinate trajectory for rerun
box	top-level	string	required	Input box trajectory for rerun
vel	top-level	string	optional	Input velocity trajectory for rerun
frame_limit	top-level	int	-	Maximum number of rerun frames
rerun_frame_limit	top-level	int	alias of frame_limit	Compatibility alias
rerun_start	top-level	int	0	Starting frame offset
rerun_strip	top-level	int	0	Number of frames skipped between processed frames
rerun_need_box_update	top-level	int	0	Whether to update the box deformation tensor during rerun

The current source uses flat rerun_* keys rather than a [rerun] TOML section.

Wall Constraints

Hard Walls

Hard-wall limits are configured through the [hard_wall] prefix:

[hard_wall]
x_low = 0.0
x_high = 40.0

Parameter	Scope	Type	Description
x_low, x_high	hard_wall	float	Hard-wall position in the X direction
y_low, y_high	hard_wall	float	Hard-wall position in the Y direction
z_low, z_high	hard_wall	float	Hard-wall position in the Z direction

Hard walls cannot be used in npt mode.

Soft Walls

Soft walls are enabled by providing a configuration file:

[soft_walls]
in_file = "soft_walls.txt"

Parameter	Scope	Type	Description
in_file	soft_walls	string	Soft-wall configuration file path

ReaxFF Reactive Force Field

Configured via [REAXFF] section:

[REAXFF]
in_file = "ffield.reax.cho"
type_in_file = "atom_types.txt"

Parameter	Type	Description
in_file	string	ReaxFF parameter file
type_in_file	string	Atom type mapping file

Generalized Born

Generalized Born parameters are configured through the [gb] scope:

[gb]
in_file = "gb.txt"
epsilon = 78.5
radii_offset = 0.09
radii_cutoff = 25.0

Parameter	Scope	Type	Default	Description
in_file	gb	string	-	Native GB parameter file
epsilon	gb	float	78.5	Relative dielectric constant
radii_offset	gb	float	0.09	Offset subtracted from self radii
radii_cutoff	gb	float	cutoff	Cutoff used when building effective Born radii

LJ Soft Core

If a native soft-core LJ parameter file is provided through LJ_soft_core_in_file, SPONGE also reads several top-level alchemical parameters:

Parameter	Scope	Type	Default	Description
lambda_lj	top-level	float	required	Soft-core LJ coupling parameter
soft_core_alpha	top-level	float	0.5	Soft-core alpha parameter
soft_core_powfer	top-level	float	1.0	Soft-core power parameter
soft_core_sigma	top-level	float	3.0	Soft-core sigma parameter
soft_core_sigma_min	top-level	float	0.0	Lower bound for the effective sigma term

These keys are top-level in the current source, not nested under [LJ_soft_core].

Custom Forces

Parameter	Scope	Type	Description
in_file	pairwise_force	string	Pairwise-force configuration file
in_file	listed_forces	string	Listed-force configuration file

Output Mapping For Periodic Molecules

Parameter	Scope	Type	Default	Description
force_whole_output	top-level	bool	false	Force whole-molecule output mapping for periodic systems
make_output_whole	top-level	string	-	Extra connectivity edges written as atom_i-atom_j pairs to help rebuild whole molecules

These keys are only meaningful when pbc = true.

Solvent LJ Shortcut

Parameter	Scope	Type	Default	Description
solvent_LJ	top-level	bool	auto	Enable the water-specific solvent LJ shortcut

On GPU builds, if solvent_LJ is omitted, SPONGE may enable it automatically for suitable water-only tail residues. Setting the key explicitly overrides that heuristic.

Plugins

Plugins are configured by the top-level key plugin, which is a whitespace- separated list of shared-library paths. This is not currently a TOML table interface.

Control Parameters

Parameter	Type	Default	Description
command_only	bool	false	Use command-line arguments only, skip mdin file
dont_check_input	bool	false	Skip unused parameter checking
end_pause	bool	false	Pause after simulation ends