Biasing and analysis methods

All of the biasing and analysis methods implemented (abf, harmonic, histogram and metadynamics) recognize the following options:

• name $<$ Identifier for the bias $>$
  Context: colvar bias
  Acceptable Values: string
  Default Value: $<$ type of bias$><$ bias index$>$
  Description: This string is used to identify the bias or analysis method in output messages and to name some output files.

• colvars $<$ Collective variables involved $>$
  Context: colvar bias
  Acceptable Values: space-separated list of colvar names
  Description: This option selects by name all the colvars to which this bias or analysis will be applied.

• outputEnergy $<$ Write the current bias energy to the trajectory file $>$
  Context: colvar bias
  Acceptable Values: boolean
  Default Value: off
  Description: If this option is chosen and colvarsTrajFrequency is not zero, the current value of the biasing energy will be written to the trajectory file during the simulation.

Adaptive Biasing Force

For a full description of the Adaptive Biasing Force method, see reference [20]. For details about this implementation, see references [34] and [35]. When publishing research that makes use of this functionality, please cite references [20] and [35].

An alternate usage of this feature is the application of custom tabulated biasing potentials to one or more colvars. See inputPrefix and updateBias below.

ABF is based on the thermodynamic integration (TI) scheme for computing free energy profiles. The free energy as a function of a set of collective variables ${\mbox{\boldmath {$\xi$}}}=(\xi_{i})_{i\in[1,n]}$ is defined from the canonical distribution of ${\mbox{\boldmath {$\xi$}}}$ , ${\mathcal P}({\mbox{\boldmath {$\xi$}}})$ :

$$A({\mbox{\boldmath {$\xi$}}}) = -\frac{1}{\beta} \ln {\mathcal P}({\mbox{\boldmath {$\xi$}}}) + A_0$$ (48)

In the TI formalism, the free energy is obtained from its gradient, which is generally calculated in the form of the average of a force ${\mbox{\boldmath {$F$}}}_\xi$ exerted on ${\mbox{\boldmath {$\xi$}}}$ , taken over an iso- ${\mbox{\boldmath {$\xi$}}}$ surface:

$${\mbox{\boldmath {$\nabla$}}}_\xi A({\mbox{\boldmath {$\xi$}}}) =... ...t\langle -{\mbox{\boldmath {$F$}}}_\xi \right\rangle_{\mbox{\boldmath {$\xi$}}}$$ (49)

Several formulae that take the form of (50) have been proposed. This implementation relies partly on the classic formulation [15], and partly on a more versatile scheme originating in a work by Ruiz-Montero et al. [61], generalized by den Otter [21] and extended to multiple variables by Ciccotti et al. [18]. Consider a system subject to constraints of the form $\sigma_{k}({\mbox{\boldmath {$x$}}}) = 0$ . Let ( ${\mbox{\boldmath {$v$}}}_{i})_{i\in[1,n]}$ be arbitrarily chosen vector fields ( $\mathbb{R}^{3N}\rightarrow\mathbb{R}^{3N}$ ) verifying, for all $i$ , $j$ , and $k$ :

$${\mbox{\boldmath {$v$}}}_{i} \cdot \mbox{\boldmath$\nabla_{\!\!x}\,$}\xi_{j} = \delta_{ij}$$ (50)

$${\mbox{\boldmath {$v$}}}_{i} \cdot \mbox{\boldmath$\nabla_{\!\!x}\,$}\sigma_{k} =$$ 0 (51)

then the following holds [18]:

$$\frac{\partial A}{\partial \xi_{i}} = \left\langle {\mbox{\boldma... ...\,$}\cdot {\mbox{\boldmath {$v$}}}_{i} \right\rangle_{\mbox{\boldmath {$\xi$}}}$$ (52)

where $V$ is the potential energy function. ${\mbox{\boldmath {$v$}}}_{i}$ can be interpreted as the direction along which the force acting on variable $\xi_{i}$ is measured, whereas the second term in the average corresponds to the geometric entropy contribution that appears as a Jacobian correction in the classic formalism [15]. Condition (51) states that the direction along which the system force on $\xi_{i}$ is measured is orthogonal to the gradient of $\xi_{j}$ , which means that the force measured on $\xi_{i}$ does not act on $\xi_{j}$ .

Equation (52) implies that constraint forces are orthogonal to the directions along which the free energy gradient is measured, so that the measurement is effectively performed on unconstrained degrees of freedom. In NAMD, constraints are typically applied to the lengths of bonds involving hydrogen atoms, for example in TIP3P water molecules (parameter rigidBonds, section 5.6.1).

In the framework of ABF, ${\bf F}_\xi$ is accumulated in bins of finite size $\delta \xi$ , thereby providing an estimate of the free energy gradient according to equation (50). The biasing force applied along the collective variables to overcome free energy barriers is calculated as:

$${\bf F}^{\rm ABF} = \alpha(N_\xi) \times \mbox{\boldmath$\nabla_{\!\!x}\,$} \widetilde A({\mbox{\boldmath {$\xi$}}})$$ (53)

where $\nabla_{\!\!x}\,$ $\widetilde A$ denotes the current estimate of the free energy gradient at the current point ${\mbox{\boldmath {$\xi$}}}$ in the collective variable subspace, and $\alpha(N_\xi)$ is a scaling factor that is ramped from 0 to 1 as the local number of samples $N_\xi$ increases to prevent nonequilibrium effects in the early phase of the simulation, when the gradient estimate has a large variance. See the fullSamples parameter below for details.

As sampling of the phase space proceeds, the estimate $\nabla_{\!\!x}\,$ $\widetilde A$ is progressively refined. The biasing force introduced in the equations of motion guarantees that in the bin centered around ${\mbox{\boldmath {$\xi$}}}$ , the forces acting along the selected collective variables average to zero over time. Eventually, as the undelying free energy surface is canceled by the adaptive bias, evolution of the system along ${\mbox{\boldmath {$\xi$}}}$ is governed mainly by diffusion. Although this implementation of ABF can in principle be used in arbitrary dimension, a higher-dimension collective variable space is likely to result in sampling difficulties. Most commonly, the number of variables is one or two.

ABF requirements on collective variables

1. Only linear combinations of colvar components can be used in ABF calculations.
2. Availability of system forces is necessary. The following colvar components can be used in ABF calculations: distance, distance_xy, distance_z, angle, dihedral, gyration, rmsd and eigenvector. Atom groups may not be replaced by dummy atoms, unless they are excluded from the force measurement by specifying oneSiteSystemForce, if available.
3. Mutual orthogonality of colvars. In a multidimensional ABF calculation, equation (51) must be satisfied for any two colvars $\xi_{i}$ and $\xi_{j}$ . Various cases fulfill this orthogonality condition:
   • $\xi_{i}$ and $\xi_{j}$ are based on non-overlapping sets of atoms.
   • atoms involved in the force measurement on $\xi_{i}$ do not participate in the definition of $\xi_{j}$ . This can be obtained using the option oneSiteSystemForce of the distance, angle, and dihedral components (example: Ramachandran angles $\phi$ , $\psi$ ).
   • $\xi_{i}$ and $\xi_{j}$ are orthogonal by construction. Useful cases are the sum and difference of two components, or distance_z and distance_xy using the same axis.
4. Mutual orthogonality of components: when several components are combined into a colvar, it is assumed that their vectors ${\mbox{\boldmath {$v$}}}_{i}$ (equation (53)) are mutually orthogonal. The cases described for colvars in the previous paragraph apply.
5. Orthogonality of colvars and constraints: equation 52 can be satisfied in two simple ways, if either no constrained atoms are involved in the force measurement (see point 3 above) or pairs of atoms joined by a constrained bond are part of an atom group which only intervenes through its center (center of mass or geometric center) in the force measurement. In the latter case, the contributions of the two atoms to the left-hand side of equation 52 cancel out. For example, all atoms of a rigid TIP3P water molecule can safely be included in an atom group used in a distance component.

Parameters for ABF

ABF depends on parameters from collective variables to define the grid on which free energy gradients are computed. In the direction of each colvar, the grid ranges from lowerBoundary to upperBoundary, and the bin width (grid spacing) is set by the width parameter (see 10.2.1). The following specific parameters can be set in the ABF configuration block (in addition to generic bias parameters such as colvars - section 10.5):

• fullSamples $<$ Number of samples in a bin prior to application of the ABF $>$
  Context: abf
  Acceptable Values: positive integer
  Default Value: 200
  Description: To avoid nonequilibrium effects due to large fluctuations of the force exerted along the colvars, it is recommended to apply a biasing force only after a the estimate has started converging. If fullSamples is non-zero, the applied biasing force is scaled by a factor $\alpha(N_\xi)$ between 0 and 1. If the number of samples $N_\xi$ in the current bin is higher than fullSamples, the factor is one. If it is less than half of fullSamples, the factor is zero and no bias is applied. Between those two thresholds, the factor follows a linear ramp from 0 to 1: $\alpha(N_\xi) =(2N_\xi/\mathrm{fullSamples})-1$ .

• maxForce $<$ Maximum magnitude of the ABF force $>$
  Context: abf
  Acceptable Values: positive decimals (one per colvar)
  Default Value: disabled
  Description: This option enforces a cap on the magnitude of the biasing force effectively applied by this ABF bias on each colvar. This can be useful in the presence of singularities in the PMF such as hard walls, where the discretization of the average force becomes very inaccurate, causing the colvar's diffusion to get ``stuck'' at the singularity. To enable this cap, provide one non-negative value for each colvar. The unit of force is kcal/mol divided by the colvar unit.

• hideJacobian $<$ Remove geometric entropy term from calculated free energy gradient? $>$
  Context: abf
  Acceptable Values: boolean
  Default Value: no
  Description: In a few special cases, most notably distance-based variables, an alternate definition of the potential of mean force is traditionally used, which excludes the Jacobian term describing the effect of geometric entropy on the distribution of the variable. This results, for example, in particle-particle potentials of mean force being flat at large separations. Setting this parameter to yes causes the output data to follow that convention, by removing this contribution from the output gradients while applying internally the corresponding correction to ensure uniform sampling. It is not allowed for colvars with multiple components.

• outputFreq $<$ Frequency (in timesteps) at which ABF data files are refreshed $>$
  Context: abf
  Acceptable Values: positive integer
  Default Value: Colvar module restart frequency
  Description: The files containing the free energy gradient estimate and sampling histogram (and the PMF in one-dimensional calculations) are written on disk at the given time interval.

• historyFreq $<$ Frequency (in timesteps) at which ABF history files are accumulated $>$
  Context: abf
  Acceptable Values: positive integer
  Default Value: 0
  Description: If this number is non-zero, the free energy gradient estimate and sampling histogram (and the PMF in one-dimensional calculations) are appended to files on disk at the given time interval. History file names use the same prefix as output files, with ``.hist'' appended.

• inputPrefix $<$ Filename prefix for reading ABF data $>$
  Context: abf
  Acceptable Values: list of strings
  Description: If this parameter is set, for each item in the list, ABF tries to read a gradient and a sampling files named $<$ inputPrefix$>$ .grad and $<$ inputPrefix$>$ .count. This is done at startup and sets the initial state of the ABF algorithm. The data from all provided files is combined appropriately. Also, the grid definition (min and max values, width) need not be the same that for the current run. This command is useful to piece together data from simulations in different regions of collective variable space, or change the colvar boundary values and widths. Note that it is not recommended to use it to switch to a smaller width, as that will leave some bins empty in the finer data grid. This option is NOT compatible with reading the data from a restart file (colvarsInput option of the NAMD config file).

• applyBias $<$ Apply the ABF bias? $>$
  Context: abf
  Acceptable Values: boolean
  Default Value: yes
  Description: If this is set to no, the calculation proceeds normally but the adaptive biasing force is not applied. Data is still collected to compute the free energy gradient. This is mostly intended for testing purposes, and should not be used in routine simulations.

• updateBias $<$ Update the ABF bias? $>$
  Context: abf
  Acceptable Values: boolean
  Default Value: yes
  Description: If this is set to no, the initial biasing force (e.g. read from a restart file or through inputPrefix) is not updated during the simulation. As a result, a constant bias is applied. This can be used to apply a custom, tabulated biasing potential to any combination of colvars. To that effect, one should prepare a gradient file containing the gradient of the potential to be applied (negative of the bias force), and a count file containing only values greater than fullSamples. These files must match the grid parameters of the colvars.

Multiple-replica ABF

• shared $<$ Apply multiple-replica ABF, sharing force samples among the replicas? $>$
  Context: abf
  Acceptable Values: boolean
  Default Value: no
  Description: This is command requires that NAMD be compiled and executed with multiple-replica support. If shared is set to yes, the system force samples will be synchronized among all replicas at intervals defined by sharedFreq. Thus, it is as if system force samples among all replicas are gathered in a single shared buffer, which why the algorithm is referred to as shared ABF. Shared ABF allows all replicas to benefit from the sampling done by other replicas and can lead to faster convergence of the biasing force.

• sharedFreq $<$ Frequency (in timesteps) at which force samples are synchronized among the replicas $>$
  Context: abf
  Acceptable Values: positive integer
  Default Value: outputFreq
  Description: In the current implementation of shared ABF, each replica maintains a separate buffer of system force samples that determine the biasing force. Every sharedFreq steps, the replicas communicate the samples that have been gathered since the last synchronization time, ensuring all replicas apply a similar biasing force.

Output files

The ABF bias produces the following files, all in multicolumn ASCII format:

• outputName.grad: current estimate of the free energy gradient (grid), in multicolumn;
• outputName.count: total number of samples collected, on the same grid;
• outputName.pmf: only for one-dimensional calculations, integrated free energy profile or PMF.

If several ABF biases are defined concurrently, their name is inserted to produce unique filenames for output, as in outputName.abf1.grad. This should not be done routinely and could lead to meaningless results: only do it if you know what you are doing!

If the colvar space has been partitioned into sections (windows) in which independent ABF simulations have been run, the resulting data can be merged using the inputPrefix option described above (a run of 0 steps is enough).

Post-processing: reconstructing a multidimensional free energy surface

If a one-dimensional calculation is performed, the estimated free energy gradient is automatically integrated and a potential of mean force is written under the file name <outputName>.pmf, in a plain text format that can be read by most data plotting and analysis programs (e.g. gnuplot).

In dimension 2 or greater, integrating the discretized gradient becomes non-trivial. The standalone utility abf_integrate is provided to perform that task. abf_integrate reads the gradient data and uses it to perform a Monte-Carlo (M-C) simulation in discretized collective variable space (specifically, on the same grid used by ABF to discretize the free energy gradient). By default, a history-dependent bias (similar in spirit to metadynamics) is used: at each M-C step, the bias at the current position is incremented by a preset amount (the hill height). Upon convergence, this bias counteracts optimally the underlying gradient; it is negated to obtain the estimate of the free energy surface.

abf_integrate is invoked using the command-line:

integrate <gradient_file> [-n <nsteps>] [-t <temp>] [-m (0|1)]
                          [-h <hill_height>] [-f <factor>]

The gradient file name is provided first, followed by other parameters in any order. They are described below, with their default value in square brackets:

• -n: number of M-C steps to be performed; by default, a minimal number of steps is chosen based on the size of the grid, and the integration runs until a convergence criterion is satisfied (based on the RMSD between the target gradient and the real PMF gradient)
• -t: temperature for M-C sampling (unrelated to the simulation temperature) [500 K]
• -m: use metadynamics-like biased sampling? (0 = false) [1]
• -h: increment for the history-dependent bias (``hill height'') [0.01 kcal/mol]
• -f: if non-zero, this factor is used to scale the increment stepwise in the second half of the M-C sampling to refine the free energy estimate [0.5]

Using the default values of all parameters should give reasonable results in most cases.

abf_integrate produces the following output files:

• <gradient_file>.pmf: computed free energy surface
• <gradient_file>.histo: histogram of M-C sampling (not usable in a straightforward way if the history-dependent bias has been applied)
• <gradient_file>.est: estimated gradient of the calculated free energy surface (from finite differences)
• <gradient_file>.dev: deviation between the user-provided numerical gradient and the actual gradient of the calculated free energy surface. The RMS norm of this vector field is used as a convergence criteria and displayed periodically during the integration.

Note: Typically, the ``deviation'' vector field does not vanish as the integration converges. This happens because the numerical estimate of the gradient does not exactly derive from a potential, due to numerical approximations used to obtain it (finite sampling and discretization on a grid).

Metadynamics

The metadynamics method uses a history-dependent potential [43] that generalizes to any type of colvars the conformational flooding [28] and local elevation [36] methods, originally formulated to use as colvars the principal components of a covariance matrix or a set of dihedral angles, respectively. The metadynamics potential on the colvars ${\mbox{\boldmath {$\xi$}}} = (\xi_{1}, \xi_{2}, \ldots, \xi_{N_{\mathrm{cv}}})$ is defined as:

$$V_{\mathrm{meta}}({\mbox{\boldmath {$\xi$}}}) \; = \; { \sum_{t' ... ...t(-\frac{(\xi_{i}-\xi_{i}(t'))^{2}}{2\delta_{\xi_{i}}^{2}}\right) } }\mathrm{,}$$ (54)

where $V_{\mathrm{meta}}$ is the history-dependent potential acting on the current values of the colvars ${\mbox{\boldmath {$\xi$}}}$ , and depends only parametrically on the previous values of the colvars. $V_{\mathrm{meta}}$ is constructed as a sum of $N_{\mathrm{cv}}$ -dimensional repulsive Gaussian ``hills'', whose height is a chosen energy constant $W$ , and whose centers are the previously explored configurations $\left({\mbox{\boldmath {$\xi$}}}(\delta{}t), {\mbox{\boldmath {$\xi$}}}(2\delta{}t), \ldots\right)$ . Each Gaussian functions has a width of approximately $2\delta_{\xi_{i}}$ along the direction of the $i$ -th colvar.

During the simulation, the system evolves towards the nearest minimum of the ``effective'' potential of mean force $\tilde{A}({\mbox{\boldmath {$\xi$}}})$ , which is the sum of the ``real'' underlying potential of mean force $A({\mbox{\boldmath {$\xi$}}})$ and the the metadynamics potential $V_{\mathrm{meta}}({\mbox{\boldmath {$\xi$}}})$ . Therefore, at any given time the probability of observing the configuration ${\mbox{\boldmath {$\xi^{*}$}}}$ is proportional to $\exp\left(-\tilde{A}({\mbox{\boldmath {$\xi^{*}$}}})/\kappa_{\mathrm{B}}T\right)$ : this is also the probability that a new Gaussian ``hill'' is added at that configuration. If the simulation is run for a sufficiently long time, each local minimum is canceled out by the sum of the Gaussian ``hill'' functions. At that stage the the ``effective'' potential of mean force $\tilde{A}({\mbox{\boldmath {$\xi$}}})$ is constant, and $-V_{\mathrm{meta}}({\mbox{\boldmath {$\xi$}}})$ is an accurate estimator of the ``real'' potential of mean force $A({\mbox{\boldmath {$\xi$}}})$ , save for an additive constant:

$$A({\mbox{\boldmath {$\xi$}}}) \; \simeq \; { -V_{\mathrm{meta}}({\mbox{\boldmath {$\xi$}}}) + K }$$ (55)

Assuming that the set of collective variables includes all relevant degrees of freedom, the predicted error of the estimate is a simple function of the correlation times of the colvars $\tau_{\xi_{i}}$ , and of the user-defined parameters $W$ , $\delta_{\xi_{i}}$ and $\delta{}t$ [14]. In typical applications, a good rule of thumb can be to choose the ratio $W/\delta{}t$ much smaller than $\kappa_{\mathrm{B}}T/\tau_{{\mbox{\boldmath {$\xi$}}}}$ , where $\tau_{{\mbox{\boldmath {$\xi$}}}}$ is the longest among ${\mbox{\boldmath {$\xi$}}}$ 's correlation times: $\delta_{\xi_{i}}$ then dictates the resolution of the calculated PMF.

To enable a metadynamics calculation, a metadynamics block must be defined in the colvars configuration file. Its only mandatory keyword is the colvars option listing all the variables involved: multidimensional PMFs are obtained by the same metadynamics instance applied to all the colvars.

The parameters $W$ and $\delta{}t$ are specified by the keywords hillWeight and newHillFrequency, respectively. The values of these options are optimal for colvars with correlation times $\tau_{{\mbox{\boldmath {$\xi$}}}}$ in the range of a few thousand simulation steps, typical of many biomolecular simulations:

• hillWeight $<$ Height of each hill (kcal/mol) $>$
  Context: metadynamics
  Acceptable Values: positive decimal
  Default Value: 0.01
  Description: This option sets the height $W$ of the hills that are added during this run. Lower values provide more accurate sampling at the price of longer simulation times to complete a PMF calculation.

• newHillFrequency $<$ Frequency of hill creation $>$
  Context: metadynamics
  Acceptable Values: positive integer
  Default Value: 1000
  Description: This option sets the number of integration steps after which a new hill is added to the metadynamics potential. Its value determines the parameter $\delta{}t$ in eq. 55. Higher values provide more accurate sampling at the price of longer simulation times to complete a PMF calculation.

It is the user's responsibility to either leave hillWeight and newHillFrequency at their default values, or to change them to match the specifics of each system. The parameter $\delta_{\xi_{i}}$ is instead defined as approximately half the width of the corresponding colvar $\xi_{i}$ (see 10.2.1).

Output files

When interpolating grids are enabled (default behavior), the PMF is written every colvarsRestartFrequency steps to the file outputName.pmf. The following two options allow to control this behavior and to visually track statistical convergence:

• writeFreeEnergyFile $<$ Periodically write the PMF for visualization $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: on
  Description: When useGrids and this option are on, the PMF is written every colvarsRestartFrequency steps.

• saveFreeEnergyFile $<$ Keep all the PMF files $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: off
  Description: When writeFreeEnergyFile and this option are on, the step number is included in the file name. Activating this option can be useful to follow more closely the convergence of the simulation, by comparing PMFs separated by short times.

Note: when Gaussian hills are deposited near lowerBoundary or upperBoundary (see 10.2.1) and interpolating grids are used (default behavior), their truncation can give rise to accumulating errors. In these cases, as a measure of fault-tolerance all Gaussian hills near the boundaries are included in the output state file, and are recalculated analytically whenever the colvar falls outside the grid's boundaries. (Such measure protects the accuracy of the calculation, and can only be disabled by hardLowerBoundary or hardUpperBoundary.) To avoid gradual loss of performance and growth of the state file, either one of the following solutions is recommended:

• enabling the option expandBoundaries, so that the grid's boundaries are automatically recalculated whenever necessary; the resulting .pmf will have its abscissas expanded accordingly;
• setting lowerWall and upperWall well within the interval delimited by lowerBoundary and upperBoundary.

Performance tuning

The following options control the computational cost of metadynamics calculations, but do not affect results. Default values are chosen to minimize such cost with no loss of accuracy.

• useGrids $<$ Interpolate the hills with grids $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: on
  Description: This option discretizes all hills for improved performance, accumulating their energy and their gradients on two separate grids of equal spacing. Grids are defined by the values of lowerBoundary, upperBoundary and width for each colvar. Currently, this option is implemented for all types of variables except the non-scalar types (distanceDir or orientation). If expandBoundaries is defined in one of the colvars, grids are automatically expanded along the direction of that colvar.

• hillWidth $<$ Relative width of the hills $>$
  Context: metadynamics
  Acceptable Values: positive decimal
  Default Value: $\sqrt{2\pi}/2$
  Description: Along each colvar, the width of each Gaussian hill ( $2\delta_{\xi_{i}}$ ) is given by the product between this number and the colvar's width. The default value gives hills whose volume is the product of $W$ times the width of all colvars. For a smoother visualization of the free energy plot, decrease width and increase hillWidth in the same proportion. Note: when useGrids is on (default in most cases), values smaller than 1 should be avoided to avoid discretization errors.

• rebinGrids $<$ Recompute the grids when reading a state file $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: off
  Description: When restarting from a state file, the grid's parameters (boundaries and widths) saved in the state file override those in the configuration file. Enabling this option forces the grids to match those in the current configuration file.

Well-tempered metadynamics

The following options define the configuration for the ``well-tempered'' metadynamics approach [4]:

• wellTempered $<$ Perform well-tempered metadynamics $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: off
  Description: If enabled, this flag causes well-tempered metadynamics as described by Barducci et al.[4] to be performed, rather than standard metadynamics. The parameter biasTemperature is then required.This feature was contributed by Li Li (Luthey-Schulten group, Departement of Chemistry, UIUC).

• biasTemperature $<$ Temperature bias for well-tempered metadynamics $>$
  Context: metadynamics
  Acceptable Values: positive decimal
  Description: When running metadynamics in the long time limit, collective variable space is sampled to a modified temperature $T+\Delta T$ . In conventional metadynamics, the temperature ``boost'' $\Delta T$ would constantly increases with time. Instead, in well-tempered metadynamics $\Delta T$ must be defined by the user via biasTemperature. If dumpFreeEnergyFile is enabled, the written PMF includes the scaling factor $(T+\Delta T)/\Delta T$ [4]. A careful choice of $\Delta T$ determines the sampling and convergence rate, and is hence crucial to the success of a well-tempered metadynamics simulation.

Multiple-replicas metadynamics

The following options define metadynamics calculations with more than one replica:

• multipleReplicas $<$ Multiple replicas metadynamics $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: off
  Description: If this option is on, multiple (independent) replica of the same system can be run at the same time, and their hills will be combined to obtain a single PMF [59]. Replicas are identified by the value of replicaID. Communication is done by files: each replica must be able to read the files created by the others, whose paths are communicated through the file replicasRegistry. This file, and the files listed in it, are read every replicaUpdateFrequency steps. Every time the colvars state file is written (colvarsRestartFrequency), the file:
  ``outputName.colvars.name.replicaID.state'' is also written, containing the state of the metadynamics bias for replicaID. In the time steps between colvarsRestartFrequency, new hills are temporarily written to the file:
  ``outputName.colvars.name.replicaID.hills'', which serves as communication buffer. These files are only required for communication, and may be deleted after a new MD run is started with a different outputName.

• replicaID $<$ Set the identifier for this replica $>$
  Context: metadynamics
  Acceptable Values: string
  Description: If multipleReplicas is on, this option sets a unique identifier for this replica. All replicas should use identical collective variable configurations, except for the value of this option.

• replicasRegistry $<$ Multiple replicas database file $>$
  Context: metadynamics
  Acceptable Values: UNIX filename
  Default Value: ``name.replica_files.txt''
  Description: If multipleReplicas is on, this option sets the path to the replicas' database file.

• replicaUpdateFrequency $<$ How often hills are communicated between replicas $>$
  Context: metadynamics
  Acceptable Values: positive integer
  Default Value: newHillFrequency
  Description: If multipleReplicas is on, this option sets the number of steps after which each replica (re)reads the other replicas' files. The lowest meaningful value of this number is newHillFrequency. If access to the file system is significantly affecting the simulation performance, this number can be increased, at the price of reduced synchronization between replicas. Values higher than colvarsRestartFrequency may not improve performance significantly.

• dumpPartialFreeEnergyFile $<$ Periodically write the contribution to the PMF from this replica $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: on
  Description: When multipleReplicas is on, tje file outputName.pmf contains the combined PMF from all replicas. Enabling this option produces an additional file outputName.partial.pmf, which can be useful to quickly monitor the contribution of each replica to the PMF. The requirements for this option are the same as dumpFreeEnergyFile.

Compatibility and post-processing

The following options may be useful only for applications that go beyond the calculation of a PMF by metadynamics:

• name $<$ Name of this metadynamics instance $>$
  Context: metadynamics
  Acceptable Values: string
  Default Value: ``meta'' + rank number
  Description: This option sets the name for this metadynamics instance. While it is not advisable to use more than one metadynamics instance within the same simulation, this allows to distinguish each instance from the others. If there is more than one metadynamics instance, the name of this bias is included in the metadynamics output file names, such as e.g. the .pmf file.

• keepHills $<$ Write each individual hill to the state file $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: off
  Description: When useGrids and this option are on, all hills are saved to the state file in their analytic form, alongside their grids. This makes it possible to later use exact analytic Gaussians for rebinGrids. To only keep track of the history of the added hills, writeHillsTrajectory is preferable.

• writeHillsTrajectory $<$ Write a log of new hills $>$
  Context: metadynamics
  Acceptable Values: boolean
  Default Value: on
  Description: If this option is on, a logfile is written by the metadynamics bias, with the name ``outputName.colvars.$<$ name$>$ .hills.traj'', which can be useful to follow the time series of the hills. When multipleReplicas is on, its name changes to
  ``outputName.colvars.$<$ name$>$ .$<$ replicaID$>$ .hills.traj''. This file can be used to quickly visualize the positions of all added hills, in case newHillFrequency does not coincide with colvarsRestartFrequency.

Harmonic restraints

The harmonic biasing method may be used to enforce fixed or moving restraints, including variants of Steered and Targeted MD. Within energy minimization runs, it allows for restrained minimization, e.g. to calculate relaxed potential energy surfaces. In the context of the colvars module, harmonic potentials are meant according to their textbook definition: $$V({\mbox{\boldmath {$x$}}}) = \frac{1}{2} k ({\mbox{\boldmath {$x$}}} - {\mbox{\boldmath {$x_0$}}})^2$$ . Note that this differs from harmonic bond and angle potentials in common force fields, where the factor of one half is typically omitted, resulting in a non-standard definition of the force constant. The restraint energy is reported by NAMD under the MISC title. A harmonic restraint is set up by a harmonic {...} block, which may contain (in addition to the standard option colvars) the following keywords:

• forceConstant $<$ Scaled force constant (kcal/mol) $>$
  Context: harmonic
  Acceptable Values: positive decimal
  Default Value: 1.0
  Description: This defines a scaled force constant for the harmonic potential. To ensure consistency for multidimensional restraints, it is divided internally by the square of the specific width for each colvar involved (which is 1 by default), so that all colvars are effectively dimensionless and of commensurate size. For instance, setting a scaled force constant of 10 kcal/mol acting on two colvars, an angle with a width of 5 degrees and a distance with a width of 0.5 Å, will apply actual force constants of 0.4 kcal/mol$\times$ degree$^{-2}$ for the angle and 40 kcal/mol/Å$^2$ for the distance.

• centers $<$ Initial harmonic restraint centers $>$
  Context: harmonic
  Acceptable Values: space-separated list of colvar values
  Description: The centers (equilibrium values) of the restraint are entered here. The number of values must be the number of requested colvars. Each value is a decimal number if the corresponding colvar returns a scalar, a ``(x, y, z)'' triplet if it returns a unit vector or a vector, and a ``q0, q1, q2, q3)'' quadruplet if it returns a rotational quaternion. If a colvar has periodicities or symmetries, its closest image to the restraint center is considered when calculating the harmonic potential.

Tip: A complex set of restraints can be applied to a system, by defining several colvars, and applying one or more harmonic restraints to different groups of colvars. In some cases, dozens of colvars can be defined, but their value may not be relevant: to limit the size of the colvars trajectory file, it may be wise to disable outputValue for such ``ancillary'' variables, and leave it enabled only for ``relevant'' ones.

Moving restraints: steered molecular dynamics

The following options allow to change gradually the centers of the harmonic restraints during a simulations. When the centers are changed continuously, a steered MD in a collective variable space is carried out.

• targetCenters $<$ Steer the restraint centers towards these targets $>$
  Context: harmonic
  Acceptable Values: space-separated list of colvar values
  Description: When defined, the current centers will be moved towards these values during the simulation. By default, the centers are moved over a total of targetNumSteps steps by a linear interpolation, in the spirit of Steered MD. If targetNumStages is set to a nonzero value, the change is performed in discrete stages, lasting targetNumSteps steps each. This second mode may be used to sample successive windows in the context of an Umbrella Sampling simulation. When continuing a simulation run, the centers specified in the configuration file $<$ colvarsConfig$>$ are overridden by those saved in the restart file $<$ colvarsInput$>$ . To perform Steered MD in an arbitrary space of colvars, it is sufficient to use this option and enable outputAppliedForce within each of the colvars involved.

• targetNumSteps $<$ Number of steps for steering $>$
  Context: harmonic
  Acceptable Values: positive integer
  Description: In single-stage (continuous) transformations, defines the number of MD steps required to move the restraint centers (or force constant) towards the values specified with targetCenters or targetForceConstant. After the target values have been reached, the centers (resp. force constant) are kept fixed. In multi-stage transformations, this sets the number of MD steps per stage.

• outputCenters $<$ Write the current centers to the trajectory file $>$
  Context: harmonic
  Acceptable Values: boolean
  Default Value: off
  Description: If this option is chosen and colvarsTrajFrequency is not zero, the positions of the restraint centers will be written to the trajectory file during the simulation. This option allows to conveniently extract the PMF from the colvars trajectory files in a steered MD calculation.

• outputAccumulatedWork $<$ Write the accumulated work of the moving restraint to the trajectory file $>$
  Context: harmonic
  Acceptable Values: boolean
  Default Value: off
  Description: If this option is chosen, targetCenters is defined, and colvarsTrajFrequency is not zero, the accumulated work from the beginning of the simulation will be written to the trajectory file. If the simulation has been continued from a previous state file, the previously accumulated work is included in the integral. This option allows to conveniently extract the PMF from the colvars trajectory files in a steered MD calculation.

Note on restarting moving restraint simulations: Information about the current step and stage of a simulation with moving restraints is stored in the restart file (state file). Thus, such simulations can be run in several chunks, and restarted directly using the same colvars configuration file. In case of a restart, the values of parameters such as targetCenters, targetNumSteps, etc. should not be changed manually.

Moving restraints: umbrella sampling

The centers of the harmonic restraints can also be changed in discrete stages: in this cases a one-dimensional umbrella sampling simulation is performed. The sampling windows in simulation are calculated in sequence. The colvars trajectory file may then be used both to evaluate the correlation times between consecutive windows, and to calculate the frequency distribution of the colvar of interest in each window. Furthermore, frequency distributions on a predefined grid can be automatically obtained by using the histogram bias (see 10.5.6).

To activate an umbrella sampling simulation, the same keywords as in the previous section can be used, with the addition of the following:

• targetNumStages $<$ Number of stages for steering $>$
  Context: harmonic
  Acceptable Values: non-negative integer
  Default Value: 0
  Description: If non-zero, sets the number of stages in which the restraint centers or force constant are changed to their target values. If zero, the change is continuous. Each stage lasts targetNumSteps MD steps. To sample both ends of the transformation, the simulation should be run for targetNumSteps $\times$ (targetNumStages + 1).

Changing force constant

The force constant of the harmonic restraint may also be changed to equilibrate [22].

• targetForceConstant $<$ Change the force constant towards this value $>$
  Context: harmonic
  Acceptable Values: positive decimal
  Description: When defined, the current forceConstant will be moved towards this value during the simulation. Time evolution of the force constant is dictated by the targetForceExponent parameter (see below). By default, the force constant is changed smoothly over a total of targetNumSteps steps. This is useful to introduce or remove restraints in a progressive manner. If targetNumStages is set to a nonzero value, the change is performed in discrete stages, lasting targetNumSteps steps each. This second mode may be used to compute the conformational free energy change associated with the restraint, within the FEP or TI formalisms. For convenience, the code provides an estimate of the free energy derivative for use in TI. A more complete free energy calculation (particularly with regard to convergence analysis), while not handled by the colvars module, can be performed by post-processing the colvars trajectory, if colvarsTrajFrequency is set to a suitably small value. It should be noted, however, that restraint free energy calculations may be handled more efficiently by an indirect route, through the determination of a PMF for the restrained coordinate.[22]

• targetForceExponent $<$ Exponent in the time-dependence of the force constant $>$
  Context: harmonic
  Acceptable Values: decimal equal to or greater than 1.0
  Default Value: 1.0
  Description: Sets the exponent, $\alpha$ , in the function used to vary the force constant as a function of time. The force is varied according to a coupling parameter $\lambda$ , raised to the power $\alpha$ : $k_\lambda = k_0 + \lambda^\alpha (k_1 - k_0)$ , where $k_0$ , $k_\lambda$ , and $k_1$ are the initial, current, and final values of the force constant. The parameter $\lambda$ evolves linearly from 0 to 1, either smoothly, or in targetNumStages equally spaced discrete stages, or according to an arbitrary schedule set with lambdaSchedule. When the initial value of the force constant is zero, an exponent greater than 1.0 distributes the effects of introducing the restraint more smoothly over time than a linear dependence, and ensures that there is no singularity in the derivative of the restraint free energy with respect to lambda. A value of 4 has been found to give good results in some tests.

• targetEquilSteps $<$ Number of steps discarded from TI estimate $>$
  Context: harmonic
  Acceptable Values: positive integer
  Description: Defines the number of steps within each stage that are considered equilibration and discarded from the restraint free energy derivative estimate reported reported in the output.

• lambdaSchedule $<$ Schedule of lambda-points for changing force constant $>$
  Context: harmonic
  Acceptable Values: list of real numbers between 0 and 1
  Description: If specified together with targetForceConstant, sets the sequence of discrete $\lambda$ values that will be used for different stages.

Linear restraints

The linear restraint biasing method is used to minimally bias a simulation. There is generally a unique strength of bias for each CV center, which means you must know the bias force constant specifically for the center of the CV. This force constant may be found by using experiment directed simulation described in section 10.5.5. Please cite Pitera and Chodera when using [58].

• forceConstant $<$ Scaled force constant (kcal/mol) $>$
  Context: linear
  Acceptable Values: positive decimal
  Default Value: 1.0
  Description: This defines a scaled force constant for the linear bias. To ensure consistency for multidimensional restraints, it is divided internally by the specific width for each colvar involved (which is 1 by default), so that all colvars are effectively dimensionless and of commensurate size.

• centers $<$ Initial linear restraint centers $>$
  Context: linear
  Acceptable Values: space-separated list of colvar values
  Description: The centers (equilibrium values) of the restraint are entered here. The number of values must be the number of requested colvars. Each value is a decimal number if the corresponding colvar returns a scalar, a ``(x, y, z)'' triplet if it returns a unit vector or a vector, and a ``q0, q1, q2, q3)'' quadruplet if it returns a rotational quaternion. If a colvar has periodicities or symmetries, its closest image to the restraint center is considered when calculating the linear potential.

Adaptive Linear Bias/Experiment Directed Simulation

Experiment directed simulation applies a linear bias with a changing force constant. Please cite White and Voth [76] when using this feature. As opposed to that reference, the force constant here is scaled by the width corresponding to the biased colvar. In White and Voth, each force constant is scaled by the colvars set center. The bias converges to a linear bias, after which it will be the minimal possible bias. You may also stop the simulation, take the median of the force constants (ForceConst) found in the colvars trajectory file, and then apply a linear bias with that constant. All the notes about units described in sections 10.5.4 and 10.5.3 apply here as well. This is not a valid simulation of any particular statistical ensemble and is only an optimization algorithm until the bias has converged.

• centers $<$ Collective variable centers $>$
  Context: alb
  Acceptable Values: space-separated list of colvar values
  Description: The desired center (equilibrium values) which will be sought during the adaptive linear biasing. The number of values must be the number of requested colvars. Each value is a decimal number if the corresponding colvar returns a scalar, a ``(x, y, z)'' triplet if it returns a unit vector or a vector, and a ``q0, q1, q2, q3)'' quadruplet if it returns a rotational quaternion. If a colvar has periodicities or symmetries, its closest image to the restraint center is considered when calculating the linear potential.
• updateFrequency $<$ The duration of updates $>$
  Context: alb
  Acceptable Values: An integer
  Description: This is, $N$ , the number of simulation steps to use for each update to the bias. This determines how long the system requires to equilibrate after a change in force constant ($N/2$ ), how long statistics are collected for an iteration ($N/2$ ), and how quickly energy is added to the system (at most, $A/2N$ , where $A$ is the forceRange). Until the force constant has converged, the method as described is an optimization procedure and not an integration of a particular statistical ensemble. It is important that each step should be uncorrelated from the last so that iterations are independent. Therefore, $N$ should be at least twice the autocorrelation time of the collective variable. The system should also be able to dissipate energy as fast as $N/2$ , which can be done by adjusting thermostat parameters. Practically, $N$ has been tested successfully at significantly shorter than the autocorrelation time of the collective variables being biased and still converge correctly.

• forceRange $<$ The expected range of the force constant in units of energy $>$
  Context: alb
  Acceptable Values: A space-separated list of decimal numbers
  Default Value: 3 $k_b T$
  Description: This is largest magnitude of the force constant which one expects. If this parameter is too low, the simulation will not converge. If it is too high the simulation will waste time exploring values that are too large. A value of 3 $k_b T$ has worked well in the systems presented as a first choice. This parameter is dynamically adjusted over the course of a simulation. The benefit is that a bad guess for the forceRange can be corrected. However, this can lead to large amounts of energy being added over time to the system. To prevent this dynamic update, add hardForceRange yes as a parameter
• rateMax $<$ The maximum rate of change of force constant $>$
  Context: alb
  Acceptable Values: A list of space-separated real numbers
  Description: This optional parameter controls how much energy is added to the system from this bias. Tuning this separately from the updateFrequency and forceRange can allow for large bias changes but with a low rateMax prevents large energy changes that can lead to instability in the simulation.

Multidimensional histograms

The histogram feature is used to record the distribution of a set of collective variables in the form of a N-dimensional histogram. It functions as a ``collective variable bias'', and is invoked by adding a histogram block to the colvars configuration file.

In addition to the common parameters name and colvars described above, a histogram block may define the following parameter:

• outputFreq $<$ Frequency (in timesteps) at which the histogram file is refreshed $>$
  Context: histogram
  Acceptable Values: positive integer
  Default Value: Colvar module restart frequency
  Description: The file containing histogram data is written on disk at the given time interval.

Like the ABF and metadynamics biases, histogram uses parameters from the colvars to define its grid. The grid ranges from lowerBoundary to upperBoundary, and the bin width is set by the width parameter.

Scripted biases

Rather than using the biasing methods described above, it is possible to apply biases provided at run time as a Tcl script, in the spirit of TclForces.

• scriptedColvarForces $<$ Enable custom, scripted forces on colvars $>$
  Context: global
  Acceptable Values: boolean
  Default Value: off
  Description: If this flag is enabled, a Tcl procedure named calc_colvar_forces accepting one parameter should be defined by the user. It is executed at each timestep, with the current step number as parameter, between the calculation of colvars and the application of bias forces. This procedure may use the scripting interface (see 10.6) to access the values of colvars and apply forces on them, effectively defining custom collective variable biases.