ABF:Controls

Navigation: Documentation / Methods / Adaptive Biasing Force Method / ABF:Controls

---

The behavior of the ABF method is controlled by the setup provided in the [abf] and eventually [abf-walker] sections. Both sections belong to the {PMFLIB} group.

[abf]

Key	Default value	Type	Description
fmode	0	integer	The mode of the ABF method: 0 - the method is disabled 1 - the standard ABF implementation is used (four points) 2 - less accurate implementation is used (two points) If the method is enabled then at least one collective variable must be provided in the {ABF} group. The collective variable specification is described in ABF:Collective variables.
fapply_abf	on	logical	Determine if the collected mean force is applied as a bias.
feimode	1	integer	The extrapolation/interpolation mode for ABF forces: 0 - disabled 1 - linear ramp with bottom limit 2 - Gaussian kernel smoothing This option triggers further setup, which is optional and described below.
fsample	5000	integer	Every fsample MD steps, the actual information about collective variable values will be printed to the ABF output file. The name of the output file is provided in the [files] section.
frestart	off	logical	Disable/Enable restarting the ABF simulation from the previous run. The name of the restart file is provided in the [files] section. If the restart is enabled but no restart file exists, then warning is raised and the simulation continues.
frstupdate	5000	integer	Every frstupdate MD steps, the actual ABF force and other data will be printed to the ABF restart file. The name of the restart file is provided in the [files] section.
ftrjsample	0	integer	Every ftrjsample MD steps, the actual ABF acumulator and other data will be printed to the ABF trajectory file. The name of the trajectory file is provided in the [files] section. Zero interval disables the ABF trajectory writing.
fmask	off	logical	Disable/Enable ABF force weights (mask). The name of file with the ABF weights is provided in the [files] section. This option has only sense if more than one collective variable is requested and some regions in such CV space should not be sampled.
fenthalpy	off	logical	If enabled, data for enthalpy calculation are collected.
entropy	off	logical	If enabled, data for entropy (-TdS) calculation are collected.
fepotoffset	0.0	real	A constant, which is added to the potential energy (affect the enthalpy calculation).
fekinoffset	0.0	real	A constant, which is added to the kinetic energy (affect the entropy calculation).

Linear ramp

If the feimode option is equal to 1 then the following options change behavior of the linear ramp.

Key	Default value	Type	Description
fhramp_min	100	integer	Bottom limit for number of samples in a bin. If the actual number of samples in a bin is lower than this number no ABF force is applied.
fhramp_max	200	integer	Top limit for number of samples in a bin. If the actual number of samples in a bin is higher than this number then estimated ABF force is applied fully otherwise it is scaled down by the linear ramp.

The linear ramp has the following form:

$\bold F_{applied} = \frac{N_k - fhramp\_min}{fhramp\_max-fhramp\_min}\bold F_{estimated}$, if $N_k < fhramp\_max$,

otherwise the full force is applied:

$\bold F_{applied} = \bold F_{estimated}$,

where $N_k$ is the number of samples in the bin k.

Gaussian kernel smoothing

If the feimode option is equal to 2 then the following option changes behavior of the Gaussian kernel smoothing. This filter requires an additional parameter (wfac) specified for each CV.

Key	Default value	Type	Description
fsmoothupdate	5000	integer	The value determines how often (in MD steps) the smoothed mean forces are updated.

The relationships between the sampled instantaneous collective and smoothed forces is given:

$\bold F_{aplied}(\boldsymbol \xi_i) = \frac{ \sum\limits_{j=1}^{N_{bins}} K(\boldsymbol \xi_i,\boldsymbol \xi_j)\bold F_{estimated}(\boldsymbol \xi_j)} {\sum\limits_{j=1}^{N_{bins}} K(\boldsymbol \xi_i,\boldsymbol \xi_j) }$

,

with the Gaussian smoothing kernel defined as:

$K(\boldsymbol \xi_i,\boldsymbol \xi_j) = exp\left( - \frac {1}{2} \sum\limits_{k=1}^{N_{CVS}} \frac { (\xi_{i,k} - \xi_{j,k})^2} { w_{k}^2 h_{k}^2 } \right)$

,

where $h_{k}^2$ is a bin width and $w_{k}^2$ is an user provided factor (wfac), see ABF:Collective variables.

[abf-walker]

The optional section [abf-walker] provides information about access to the MWA server. The MWA extension is described in ABF:Multiple walkers approach.

Key	Default value	Type	Description
fserverkey		string	The name of a file with a server key. The server key contains information about the server name, the port number, and the server password. The file is generated by the mwa-server utility when the MWA server is started.
fserverupdate	20000	integer	The number of MD steps between exchanges of data with the MWA server. Since data are exchanged in a connect/send/receive/disconnect mode, exchanges should not occur too often.
fabortonmwaerr	on	logical	Should the running of a walker (client) abort when MWA server communication failure occurs?
fconrepeats	0	integer	The number of consecutive failed exchanges with the server that will lead to premature client termination. The zero means that no failures are allowed and the client is terminated at the first failed exchange.