Skip to main content

tpcntrl (c39b2)

Temperature and pressure control


by Guillaume Lamoureux (Guillaume.Lamoureux@umontreal.ca)
and Wei Jiang (wjiang@mcs.anl.gov)
and Benoit Roux (Roux@uchicago.edu)


The TPCONTROL command specifies the thermodynamic ensemble to be
simulated with "DYNA VV2", using extended dynamics: Nose-Hoover
equations for constant volume and constant temperature (the NVT
ensemble) and Anderson-Hoover equations for constant pressure and
temperature (the NPT ensemble). It allows multiple thermostats.
"DYNA VV2" is a velocity-Verlet algorithm created to simulate
efficiently the motion of Drude oscillators (created by the DRUDE
command), and it understand the special nature of the Drude
oscillators. The algorithm works for non-polarizable force fields as
well. It is totally distinct from "DYNA VVER".

See J. Chem. Phys. 119, 3025-3039 (2003) for more details.


* Syntax | Syntax of the TPCONTROL command
* Description | Description of the TPCONTROL command
* Dynamics | Molecular dynamics with TPCONTROL
* Examples | Usage examples of the TPCONTROL command


Top
Syntax of the TPCONTROL command


TPCOntrol [NTHErmostats integer] [CMDAmping real] [NSTEps integer] -
nther{ thermostat-spec } -
[ barostat-spec ]

TPCOntrol OFF


thermostat-spec::= THERmostat integer [TREF real] <TAU real|QREF real> -
[TOLScf real] [MAXScf integer] -
atom-selection

atom-selection::= (» select .)

barostat-spec::= BAROstat [PREF real] <BTAU real|WREF real> <FULL|ZONLY>


Top
Description of the TPCONTROL command


-------------------------------------------------------------------
Keyword Default Purpose
-------------------------------------------------------------------
NTHErm 1 The number of separate thermostats (or heat baths).
For each one, a "thermostat-spec" sequence should
be given.

CMDAmping ZERO The friction constant (in 1/ps) to damp the motion
of the center of mass of the system. Any nonzero
value will create a separate thermostat (which
will actually be a "heat sink" instead of a "heat
bath") coupled to the three degrees of freedom
associated with the motion of the center of mass
of the whole system.

NSTEPs 1 (20) The number of sub-steps each timestep is divided to
integrate the thermostat variables. The default
value is 1 for non-polarizable systems (that is,
the multi-timestep approach is usually not needed),
and 20 for polarizable systems (that is, whenever
Drude oscillators are present).

TREF 298.15 The temperature of each thermostat (in Kelvins).

TAU/QREF Either TAU or QREF can be specified. TAU is the
characteristic response time for each thermostat
(in ps), and QREF is the inertia factor of the
thermostat (in kcal*AKMA**2). If the TAU keyword
is used, QREF is computed from QREF = Nf*kT*TAU**2
(where Nf is the number of degrees of freedom
coupled to the thermostat and kT is the
temperature of the thermostat).

LANG/NHGAM/NHGAMD Keyword LANG denotes Langevin theremostat is in use.
Nose-Hoover keyword TAU can NOT appear with LANG in
a TPCONTROL command line. NHGAM is friction strength on
center of mass of a Drude pair and NHGAMD applies to
relative motion of a Drude pair. CMDAMPING or NSTEPS
keyword is NOT necessary either.

TOLSCF 1e-5 The tolerance on the root-mean-square force on the
Drude oscillators (in kcal/Angstrom), used for the
iterative solution of the induced dipoles (if TREF
is zero for the thermostats coupled to Drude
particles).

MAXSCF 50 The maximum number of iterations for the iterative
solution of the induced dipoles (if TREF is zero
for the thermostats coupled to Drude particles).

PREF 1.0 The pressure of the barostat (in Atmospheres).

BTAU/WREF The characteristic response time for the barostat
(in ps). WREF (in kcal*AKMA**2) can be specified
directly, or computed from WREF = (sum_i
Nf_i*kT_i)*BTAU**2, where index i identifies the
thermostat, and runs from 1 to NTHER. BTAU is
size- and temperature-independent.

FULL/ZONLY To allow FULL relaxation of ALL the Crystal's degrees
of freedom or to allow scaling along the Z-direction
only (ZONLY)

OFF Turns the temperature and pressure control off.


-------------------------------------------------------------------
1) NTHER

For a non-polarizable system, separate thermostats could be used for
the fast-moving solvent (water molecules) and for the slower-moving
solute (protein). For a polarizable system generated with the DRUDE
command, an additional thermostat, set at a very low temperature,
should be used for the Drude oscillators. The "DYNA VV2" command will
recognize the special nature of the Drude particles and apply the
thermostat to the atom-Drude vibrations instead of the Drude
translations.


-------------------------------------------------------------------
2) CMDAMPING

To avoid any acceleration of the center of mass in "DYNA VV2", this
procedure is preferable to the NTRFRQ keyword in DYNA. Use a nonzero
CMDAMPING only if necessary: provided the force calculation is
accurate enough, the "DYNA VV2" should not induce any significant
translation of the center of mass.

CMDAMPING does not apply to the rotation around the center of mass
that may develop for a vacuum simulation or for a system with with
spherically symmetric boundary conditions (such as defined in SSBP).


-------------------------------------------------------------------
3) NSTEPS

The multi-timestep approach this parameter refers to is different from
the conventional MTS-RESPA approach. It does not separate the fast
and slow components of the forces on the particles, but instead uses a
higher-oder integration scheme for the thermostat variables. This
approach is useful when fast degrees of freedom (such as Drude
oscillators) are coupled to a low-temperature heat bath. For properly
chosen values for the mass of the Drude oscillators and the force
constant of atom-Drude bonds, it allows to use 1 or 2 fs timesteps.
The smaller the QREF is, the more such a multi-timestep approach is
needed.

The computational overhead of this multi-timestep integration scheme
is small (and scales as O(N)), because the atomic forces do not need
to be recomputed. However, it is not negligible if the system is
relatively small, and one may want to use NSTEPS as small as possible
(as long as it has no systematic effect on the properties of the
simulation).


-------------------------------------------------------------------
4) TREF

TREF is set to the "real" temperature of the system for all
thermostats coupled to "real" atoms, and should be set to a very low
temperature (typically, 1.0 K) for a thermostat coupled to Drude
particles. Such a low temperature will maintain the oscillators close
to the self-consistent field regime, and will improve the stability of
the simulation.

If TREF is zero (actually, if it's less than 1e-8 K) for a thermostat
coupled to Drude oscillators, the "DYNA VV2" command will solve the
positions of the Drude oscillators at every time step using an
iterative procedure. This procedure is very inefficient and should be
used for testing purposes only.

The TEMPerature output for the "DYNA VV2" command corresponds to the
kinetic temperature of the first selection (coupled to the first
thermostat). For a complete output of the temperature, use the IUNO
unit specified in DYNA.


-------------------------------------------------------------------
5) TAU/QREF

The inertia factor QREF of each thermostat should be tuned so that the
thermostat is following the natural temperature fluctuations of Nf
degrees of freedom at temperature T. If QREF is too small, the
temperature of the system will be controlled on too short a time
scale, and the temperature fluctuations will be abnormal (that is, not
typical of the canonical ensemble). If QREF is too large, the
temperature control is inefficient, and some modes of motion of the
system may not be properly thermalized. The kinetic temperatures each
thermostat is controlling are printed in the IUNO unit specified in
"DYNA VV2", and their distribution should be checked to see if the
QREF's are too small.

If a thermostat is coupled to Drude oscillators, the QREF value can be
as low as the order of the multi-timestep integration scheme (NSTEPS)
allows it. For the oscillators, as long as the temperature is low
enough compared to the actual temperature of the system, the
temperature fluctuations are meaningless.

TAU is roughly size and temperature-independent, and is safer to use
than QREF, which should be scaled with the size of the system and the
temperature of the heat bath.

-------------------------------------------------------------------
6) LANG/NHGAM/NHGAMD

Langevin thermostat can be used to replace Nose-Hoover thermostat.
Especially for thermodynamic simulation with Drude oscillators Dual Langevin
thermostating is strongly recommended (see the following examples). It
achieves more samplings and stability.

-------------------------------------------------------------------
7) TOLSCF, MAXSCF

If the iterative procedure cannot meet the tolerance criterion on the
gradient in less than MAXSCF iterations, it will print a short warning
message.


-------------------------------------------------------------------
8) BTAU/WREF

Similar to TAU/QREF, but for the barostat. The volume fluctuations
should be looked at to make sure BTAU is not too small. It BTAU is
too large, the volume will equilibrate too slowly and the volume
fluctuations will be underestimated (which may have thermodynamic
consequences if the volume of the system is relatively small).


Top
Molecular dynamics with TPCONTROL


The only molecular dynamics command using the information set by
TPCONTROL is "DYNA VV2".

TPCONTROL shares some data structures with the NOSE command, but
should not be used with "DYNA NOSE". The only two keywords "DYNA VV2"
recognizes from "DYNA NOSE" are IUNO and NSNOS. (What VV2 writes in
unit IUNO every NSNOS steps is not the same, however.) Similarly, the
"DYNA VV2" command ignores the constant-pressure options of "DYNA
CPT". For "DYNA VV2", all the information concerning temperature and
pressure control should come from TPCONTROL.

With TPCONTROL on, the restart file written by "DYNA VV2" contains
additional information about the constraint forces applied by "SHAKE".
Such a restart file can be read back only by "DYNA VV2 RESTART" (and
with TPCONTROL on).

With TPCONTROL off, "DYNA VV2" is performing a constant energy,
constant volume simulation (and it ignores the special nature of the
Drude oscillators, if any).


Top
Usage examples of the TPCONTROL command


All the following examples should be called immediately before calling
the "DYNA VV2" command:

OPEN WRITE CARD UNIT 62 NAME @name
DYNA VV2 ... -
IUNO 62 NSNOS 100


For a NVT simulation of a non-polarizable system:

TPCONTROL NTHER 1 -
THER 1 TREF 298.15 TAU 0.1 SELECT ALL END

For a NPT simulation of a non-polarizable system:

TPCONTROL NTHER 1 -
THER 1 TREF 298.15 TAU 0.1 SELECT ALL END -
BARO PREF 1.00 BTAU 0.2

To use separate thermostats for solvent and solute:

TPCONTROL NTHER 2 -
THER 1 TREF 298.15 TAU 0.1 SELECT SEGID WATE END -
THER 2 TREF 298.15 TAU 0.2 SELECT SEGID SOLU END -
BARO PREF 1.00 BTAU 0.2

For a NPT simulation with dual Nose_Hoover for Drude oscillator systems:

TPCONTROL NTHER 2 NSTEP 50 -
THER 1 TREF 298.15 TAU 0.1 SELECT .NOT. TYPE D* END -
THER 2 TREF 1.00 TAU 0.005 SELECT TYPE D* END -
BARO PREF 1.00 BTAU 0.2

For a NPT simulation with dual Langevin for Drude oscillator systems:

TPCONTROL NTHER 2 NHGAM 5.0 NHGAMD 10.0 -
THER 1 TREF 298.15 LANG SELECT .NOT. TYPE D* END -
THER 2 TREF 1.00 LANG SELECT TYPE D* END -
BARO PREF 1.00 BTAU 0.2

For a NPT simulation with dual Nose-Hoover for Drude oscillators, using
an iterative procedure to solve the induced dipoles (TREF=0):

TPCONTROL NTHER 2 NSTEP 50 -
THER 1 TREF 298.15 TAU 0.1 SELECT .NOT. TYPE D* END -
THER 2 TREF 0.00 TAU 0.005 SELECT TYPE D* END -
BARO PREF 1.00 BTAU 0.2

For a NPT Crystal simulation with dual Nose Hoover for Drude oscillators:

TPCONTROL NTHER 2 CMDAMP 10.0 NSTEP 20 -
THER 1 TREF 298.15 TAU 0.1 SELECT .NOT. TYPE D* END -
THER 2 TREF 1.00 TAU 0.005 SELECT TYPE D* END -
BARO PREF 1.00 BTAU 0.2 DSCY FULL