# What is the recommended value for tau-p and tau-t with GROMACS2023.0 when using PR and NH **URL:** \ **Category:** User discussions\ **Created:** [February 20, 2023, 3:22pm UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840 "2023-02-20T15:22:03Z")\ **Posts on this page:** 11\ **Page:** 1
**Author:** ![Eduardneu](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/eduardneu/32/1976_2.png) [@Eduardneu](https://gromacs.bioexcel.eu/u/Eduardneu)\ **Post date:** [February 20, 2023, 3:22pm UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/1 "2023-02-20T15:22:03Z")
GROMACS version: 2023.0 GROMACS modification: No Hey there, After consulting the GROMACS 2023 [documentation](https://manual.gromacs.org/documentation/current/user-guide/mdp-options.html#mdp-nstpcouple) and this post on GitLab [Increase the default values of nsttcouple and nstpcouple (!2624) · Merge requests · GROMACS / GROMACS · GitLab](https://gitlab.com/gromacs/gromacs/-/merge_requests/2624), I noticed that if the value -1 is used for nsttcouple and nstpcouple, their default value will be adjusted to 100 (if the tau-p and tau-t values allow for this). In the GROMACS versions until 2022 both values were set to 10 if a value of -1 was selected (see GROMACS 2022 [documentation](https://manual.gromacs.org/2022/user-guide/mdp-options.html#mdp-nsttcouple). Additionally, for first-order coupling (such as Berendsen or V-rescale), the tau-p and tau-t values must be at least 5 times greater than (nstpcouple \* dt) or (nsttcoupl \* dt). For second-order coupling (such as Parrinello-Rahman or Nosé-Hoover), the tau-p and tau-t values must be at least 20 times greater than (nstpcouple \* dt) or (nsttcouple \* dt). See [Strange Fluctuations with NH+PR coupling in GROMACS 2018.3 - Redmine #2749 (#2749) · Issues · GROMACS / GROMACS · GitLab](https://gitlab.com/gromacs/gromacs/-/issues/2749) and the GROMACS 2023 [documentation](https://manual.gromacs.org/documentation/current/reference-manual/algorithms/molecular-dynamics.html#berendsen-pressure-coupling). Furthermore, when using the Parrinello-Rahman barostat and Nosé-Hoover thermostat, the tau-p value must be at least twice the value of tau-t to prevent resonance, as documented here [Strange Fluctuations with NH+PR coupling in GROMACS 2018.3 - Redmine #2749 (#2749) · Issues · GROMACS / GROMACS · GitLab](https://gitlab.com/gromacs/gromacs/-/issues/2749). Based on these guidelines, the minimum tau-t value would be 1 ns for V-rescale and 4 ns for Nosé-Hoover, respectively, as suggested by @hess here [Increase the default values of nsttcouple and nstpcouple (!2624) · Merge requests · GROMACS / GROMACS · GitLab](https://gitlab.com/gromacs/gromacs/-/merge_requests/2624). Using the Parrinello-Rahman barostat would result in a minimum tau-p of 8 ns. However, these values seem higher than what I’m used to, and I’m not sure if they could cause my simulation to blow up. I am simulating a membrane protein (G protein-coupled receptor) with a ligand in a POPC bilayer membrane. My previous parameters in the .mdp file were: ```auto tcoupl = Nose-Hoover tc_grps = PROT_LIG MEMB SOLV tau-t = 1.0 ref-t = 310 310 310 nsttcouple = 10 pcoupl = Parrinello-Rahman pcoupletype = semiisotropic tau-p = 5.0 compressibility = 4.5e-5 4.5e-5 ref-p = 1.0 1.0 nstcouple = 10 ``` I would change these parameters to: ```auto tcoupl = Nose-Hoover tc_grps = PROT_LIG MEMB SOLV tau-t = 4.0 ref-t = 310 310 310 nsttcouple = 100 pcoupl = Parrinello-Rahman pcoupletype = semiisotropic tau-p = 8.0 compressibility = 4.5e-5 4.5e-5 ref-p = 1.0 1.0 nstcouple = 100 ``` However, I am concerned that the large values for tau-t and tau-p, combined with the Nosé-Hoover and Parrinello-Rahman methods, respectively, could cause the simulation to blow up. Additionally, I am unsure if the tau-p/tau-t ratio is too small for the NH thermostat and PR barostat to avoid resonance. Intuitively, I would increase the tau-p value, but I am worried this would further increase the risk of the simulation failing. @hess during the tests to increase the default value of nsttcouple and nstpcouple, were the tau-t and tau-p values also tested? Is there any recommendation especially for the combination of NH thermostat and PR barostat? Thanks Eddie ---
**Author:** ![hess](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/hess/32/416_2.png) [@hess](https://gromacs.bioexcel.eu/u/hess)\ **Post date:** [February 21, 2023, 8:27am UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/2 "2023-02-21T08:27:12Z")
Firstly, I assume you mean picoseconds and not nanoseconds, right? Apart from the warning when tau\_p \< 2 \* tau\_t, all other comments and changes only pertain to the coupling intervals or time-step, if you prefer to call it that way. Those changes in no way affect the coupling tau values. If your tau values are small, grompp will automatically reduce the nst parameter such the integration accuracy requirements are fulfilled. If you manually set the nst values and they do not obey the integration accuracy criteria, grompp will generate a warning. So you can (still) set tau\_t and tau\_p as you like. The risk of the system blowing up is always with too small values of tau when using second order coupling, never with too large values. I think tau\_t=1 ps is on the small side, as this adds temperature oscillations with a period of 1 ps. I would recommend a value of at least 2 ps. But 1 ps often also works fine. Nowadays the recommendation is to use the v-rescale thermostat and the c-rescale barostat. These have non of the disadvantages of second order coupling and have no disadvantages themselves. ---
**Author:** ![Eduardneu](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/eduardneu/32/1976_2.png) [@Eduardneu](https://gromacs.bioexcel.eu/u/Eduardneu)\ **Post date:** [February 21, 2023, 9:44am UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/3 "2023-02-21T09:44:52Z")
Thank you very much for your insightful answer! Yes, I meant **pico** seconds not nanoseconds. Perfect than I will try v-rescale as a thermostat and c-rescale as a barostat. For the combination of v-rescale as thermostat and c-rescale as barostat is there any condition for the ratio of tau\_p/tau\_t that needs to be fulfilled (such as tau\_p \>= 2\* tau\_t when using NH and PR)? I would really like to use the increased values for nsttcouple and nstpcouple so I would increase my tau\_p and tau\_p value respectively. Regarding c-rescale, I could find the option in the in the GROMACS 2023 [documentation](https://manual.gromacs.org/documentation/current/user-guide/mdp-options.html#pressure-coupling) but I couldn’t find any information about the algorithm in the current reference [manual](https://manual.gromacs.org/current/reference-manual/algorithms/molecular-dynamics.html#pressure-coupling). I assume though it is a first order coupling. Based on that, first order couplings should be at least 5 times higher than (nstpcouple \*dt) or (nsttcouple \* dt) I would assume that with a time step of 2 femtoseconds the tau\_p and tau\_t should be at least 1 **pico** second. Would it be advisable to use the same value for tau\_p as for tau\_t? From my experience the tau\_p value was always a multiple larger than the tau\_t. But to be fair I never knew the explicit reason for that and went with the values generated and recommended by [CHARMM-GUI](https://pubs.acs.org/doi/10.1021/acs.jctc.5b00935). Also, I recall reading, that higher values for tau\_p and tau\_t are preferred because they reduce the “bias” that is introduced by the barostat and thermostat. But again, I do not really know which value for tau\_p and tau\_t would be considered too high or too low. I basically just follow the recommendations in the GROMACS manual to determine the smallest accepted value. Is there any guideline or a rule of thumb to determine the “right” value for tau\_p and tau\_t depending of the corresponding barostat and thermostat, besides the recommendation for the smallest accepted values in the GROMACS manual? I searched all the literature I could find but did not get a satisfying answers. Based on you input I would adjust my mdp file as follows: ```auto dt = 0.002 tcoupl = v-rescale tc_grps = PROT_LIG MEMB SOLV tau-t = 1.0 ref-t = 310 310 310 nsttcouple = -1 # It should be adjusted to 100 pcoupl = c-rescale pcoupletype = semiisotropic tau-p = 1.0 compressibility = 4.5e-5 4.5e-5 ref-p = 1.0 1.0 nstcouple = -1 # It should be adjusted to 100 ``` Key question: which tau\_p and tau\_t values are recommended to run a simulation of a membrane bound protein using v-rescale as thermostat and c-rescale as a barostat using GROMACS 2023.0? ---
**Author:** ![hess](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/hess/32/416_2.png) [@hess](https://gromacs.bioexcel.eu/u/hess)\ **Post date:** [February 21, 2023, 11:18am UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/4 "2023-02-21T11:18:43Z")
C-rescale is described, including reference, in the manual here: [https://manual.gromacs.org/current/reference-manual/algorithms/molecular-dynamics.html#stochastic-cell-rescaling](https://manual.gromacs.org/current/reference-manual/algorithms/molecular-dynamics.html#stochastic-cell-rescaling) First order coupling algorithms usually do not have issues with stability due to oscillations, so choosing the value for tau is not so critical. But tau\_t needs to be small enough to effectively remove energy drift due to integration errors. tau\_p should not be too small, as scaling the volume too fast might cause instabilities. My advice would be tau\_t=1 ps and tau\_p=5 ps for nearly all atomistic simulations. ---
**Author:** ![Eduardneu](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/eduardneu/32/1976_2.png) [@Eduardneu](https://gromacs.bioexcel.eu/u/Eduardneu)\ **Post date:** [February 21, 2023, 12:06pm UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/5 "2023-02-21T12:06:05Z")
Thanks again for your helpful answers and your clear recommendation. Much appreciated! ---
**Author:** ![gxf1212](https://avatars.discourse-cdn.com/v4/letter/g/cab0a1/32.png) [@gxf1212](https://gromacs.bioexcel.eu/u/gxf1212)\ **Post date:** [July 11, 2024, 1:09pm UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/6 "2024-07-11T13:09:21Z")
Hi! Very helpful explanation. But I wonder if you have any suggestions on tau\_t and tau\_p for coarse grained simulations using v-rescale as a thermostat and c-rescale as a barostat. And can I still use compressibility = 4.5e-5 for my 230k-atom system that is polymers solvated with martini3 water? Many thanks, Xufan The whole .mdp file is title = production run ;;;define = -DPOSRES ; position restrain ; Run parameters integrator = md ; leap-frog integrator nsteps = 100000000 ; 20 \* 100000000 = 2000 ns dt = 0.02 ; 20 fs, 30 fs explodes ; Output control nstxtcout = 50000 ; save coordinates every 500.0 ps nstvout = 0 ; save velocities every 1.0 ps nstenergy = 0 ; save energies every 100.0 ps nstlog = 50000 ; update log file every 500.0 ps ; Bond parameters continuation = yes ; Restarting after NVT constraint\_algorithm = lincs ; holonomic constraints constraints = h-bonds ; none bonds (even heavy atom-H bonds) constrained lincs\_iter = 1 ; accuracy of LINCS lincs\_order = 4 ; also related to accuracy ; Neighborsearching cutoff-scheme = Verlet ns\_type = grid ; search neighboring grid cells nstlist = 20 ; 20 fs, largely irrelevant with Verlet scheme verlet-buffer-tolerance = 0.005 ; nonbonded parameters rcoulomb = 1.1 ; short-range electrostatic cutoff (in nm) epsilon\_r = 15 ; 2.5 (with polarizable water) rvdw = 1.1 ; short-range van der Waals cutoff (in nm) vdw\_type = cutoff vdw-modifier = Potential-shift-verlet ; Electrostatics coulombtype = PME ; Particle Mesh Ewald for long-range electrostatics pme\_order = 4 ; cubic interpolation fourierspacing = 0.16 ; grid spacing for FFT ; Temperature coupling is on tcoupl = V-rescale ; modified Berendsen thermostat tc-grps = Solute W\_Ion ; two coupling groups - more accurate tau\_t = 1.0 1.0 ; time constant, in ps ref\_t = 300 300 ; reference temperature, one for each group, in K ; Pressure coupling is on pcoupl = C-rescale ; Pressure coupling on in NPT pcoupltype = isotropic ; uniform scaling of box vectors tau\_p = 1.0 ; time constant, in ps ref\_p = 1.0 ; reference pressure, in bar compressibility = 4.5e-5 ; isothermal compressibility of water, bar^-1 ; Periodic boundary conditions pbc = xyz ; 3-D PBC ; Dispersion correction DispCorr = EnerPres ; account for cut-off vdW scheme ; Velocity generation gen\_vel = no ; com refcoord\_scaling = com comm-mode = Linear ; Remove center of mass translation ; comm-grps = Protein Non-Protein ; group(s) for center of mass motion removal, default is the whole system ---
**Author:** ![hess](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/hess/32/416_2.png) [@hess](https://gromacs.bioexcel.eu/u/hess)\ **Post date:** [August 6, 2024, 1:35pm UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/7 "2024-08-06T13:35:52Z")
I don’t know what the compressibility of Martini models is, ask on the Martini forum/list. I would think one would want to increase tau\_t and tau\_p by roughly the factor of the difference in time step. But then one needs to check that the actual temperature is sufficiently close to the reference temperature. ---
**Author:** ![flor](https://avatars.discourse-cdn.com/v4/letter/f/a87d85/32.png) [@flor](https://gromacs.bioexcel.eu/u/flor)\ **Post date:** [May 9, 2025, 7:44am UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/8 "2025-05-09T07:44:51Z")
Hi, I wonder if you have any idea for tau\_t, tau\_p for simulations in high temperature: 363K and high pressure: 5000 and +/-1000 bar? Thanks in advanced ---
**Author:** ![hess](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/hess/32/416_2.png) [@hess](https://gromacs.bioexcel.eu/u/hess)\ **Post date:** [May 12, 2025, 1:40pm UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/9 "2025-05-12T13:40:43Z")
The state point should not affect the choice of coupling parameters much. I would still suggest tau\_t=1 and tau\_p=5 with the rescale algorithms. ---
**Author:** ![flor](https://avatars.discourse-cdn.com/v4/letter/f/a87d85/32.png) [@flor](https://gromacs.bioexcel.eu/u/flor)\ **Post date:** [May 12, 2025, 2:02pm UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/10 "2025-05-12T14:02:57Z")
I used tau\_t=0.1 and tau\_p=1 :( ---
**Author:** ![hess](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/hess/32/416_2.png) [@hess](https://gromacs.bioexcel.eu/u/hess)\ **Post date:** [May 12, 2025, 2:24pm UTC](https://gromacs.bioexcel.eu/t/what-is-the-recommended-value-for-tau-p-and-tau-t-with-gromacs2023-0-when-using-pr-and-nh/5840/11 "2025-05-12T14:24:13Z")
That should also be fine. You might just loose a little bit of performance. The time dependence of the fluctuations of the kinetic energy and the pressure will be lost, but in most cases you are not interested in this.