subroutine MD c------------------------------------------------ c The driver for molecular dynamics subroutines c------------------------------------------------ implicit real*8 (a-h,o-z) include 'DIMENSIONS' #ifdef MPI include "mpif.h" integer IERROR,ERRCODE #endif include 'COMMON.SETUP' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' include 'COMMON.TIME1' double precision cm(3),L(3),vcm(3) #ifdef VOUT double precision v_work(maxres6),v_transf(maxres6) #endif integer ilen,rstcount external ilen character*50 tytul common /gucio/ cm integer itime c #ifdef MPI if (ilen(tmpdir).gt.0) & call copy_to_tmp(pref_orig(:ilen(pref_orig))//"_" & //liczba(:ilen(liczba))//'.rst') #else if (ilen(tmpdir).gt.0) & call copy_to_tmp(pref_orig(:ilen(pref_orig))//"_"//'.rst') #endif t_MDsetup=0.0d0 t_langsetup=0.0d0 t_MD=0.0d0 t_enegrad=0.0d0 t_sdsetup=0.0d0 write (iout,'(20(1h=),a20,20(1h=))') "MD calculation started" #ifdef MPI tt0=MPI_Wtime() #else tt0 = tcpu() #endif c Determine the inverse of the inertia matrix. call setup_MD_matrices c Initialize MD call init_MD #ifdef MPI t_MDsetup = MPI_Wtime()-tt0 #else t_MDsetup = tcpu()-tt0 #endif rstcount=0 c Entering the MD loop #ifdef MPI tt0 = MPI_Wtime() #else tt0 = tcpu() #endif if (lang.eq.2 .or. lang.eq.3) then #ifndef LANG0 call setup_fricmat if (lang.eq.2) then call sd_verlet_p_setup else call sd_verlet_ciccotti_setup endif do i=1,dimen3 do j=1,dimen3 pfric0_mat(i,j,0)=pfric_mat(i,j) afric0_mat(i,j,0)=afric_mat(i,j) vfric0_mat(i,j,0)=vfric_mat(i,j) prand0_mat(i,j,0)=prand_mat(i,j) vrand0_mat1(i,j,0)=vrand_mat1(i,j) vrand0_mat2(i,j,0)=vrand_mat2(i,j) enddo enddo flag_stoch(0)=.true. do i=1,maxflag_stoch flag_stoch(i)=.false. enddo #else write (iout,*) & "LANG=2 or 3 NOT SUPPORTED. Recompile without -DLANG0" #ifdef MPI call MPI_Abort(MPI_COMM_WORLD,IERROR,ERRCODE) #endif stop #endif else if (lang.eq.1 .or. lang.eq.4) then call setup_fricmat endif #ifdef MPI t_langsetup=MPI_Wtime()-tt0 tt0=MPI_Wtime() #else t_langsetup=tcpu()-tt0 tt0=tcpu() #endif do itime=1,n_timestep rstcount=rstcount+1 if (lang.gt.0 .and. surfarea .and. & mod(itime,reset_fricmat).eq.0) then if (lang.eq.2 .or. lang.eq.3) then #ifndef LANG0 call setup_fricmat if (lang.eq.2) then call sd_verlet_p_setup else call sd_verlet_ciccotti_setup endif do i=1,dimen3 do j=1,dimen3 pfric0_mat(i,j,0)=pfric_mat(i,j) afric0_mat(i,j,0)=afric_mat(i,j) vfric0_mat(i,j,0)=vfric_mat(i,j) prand0_mat(i,j,0)=prand_mat(i,j) vrand0_mat1(i,j,0)=vrand_mat1(i,j) vrand0_mat2(i,j,0)=vrand_mat2(i,j) enddo enddo flag_stoch(0)=.true. do i=1,maxflag_stoch flag_stoch(i)=.false. enddo #endif else if (lang.eq.1 .or. lang.eq.4) then call setup_fricmat endif write (iout,'(a,i10)') & "Friction matrix reset based on surface area, itime",itime endif if (reset_vel .and. tbf .and. lang.eq.0 & .and. mod(itime,count_reset_vel).eq.0) then call random_vel write(iout,'(a,f20.2)') & "Velocities reset to random values, time",totT do i=0,2*nres do j=1,3 d_t_old(j,i)=d_t(j,i) enddo enddo endif if (reset_moment .and. mod(itime,count_reset_moment).eq.0) then call inertia_tensor call vcm_vel(vcm) do j=1,3 d_t(j,0)=d_t(j,0)-vcm(j) enddo call kinetic(EK) kinetic_T=2.0d0/(dimen3*Rb)*EK scalfac=dsqrt(T_bath/kinetic_T) write(iout,'(a,f20.2)') "Momenta zeroed out, time",totT do i=0,2*nres do j=1,3 d_t_old(j,i)=scalfac*d_t(j,i) enddo enddo endif if (lang.ne.4) then if (RESPA) then c Time-reversible RESPA algorithm c (Tuckerman et al., J. Chem. Phys., 97, 1990, 1992) call RESPA_step(itime) else c Variable time step algorithm. call velverlet_step(itime) endif else #ifdef BROWN call brown_step(itime) #else print *,"Brown dynamics not here!" #ifdef MPI call MPI_Abort(MPI_COMM_WORLD,IERROR,ERRCODE) #endif stop #endif endif if (ntwe.ne.0) then if (mod(itime,ntwe).eq.0) call statout(itime) #ifdef VOUT do j=1,3 v_work(j)=d_t(j,0) enddo ind=3 do i=nnt,nct-1 do j=1,3 ind=ind+1 v_work(ind)=d_t(j,i) enddo enddo do i=nnt,nct if (itype(i).ne.10) then do j=1,3 ind=ind+1 v_work(ind)=d_t(j,i+nres) enddo endif enddo write (66,'(80f10.5)') & ((d_t(j,i),j=1,3),i=0,nres-1),((d_t(j,i+nres),j=1,3),i=1,nres) do i=1,ind v_transf(i)=0.0d0 do j=1,ind v_transf(i)=v_transf(i)+gvec(j,i)*v_work(j) enddo v_transf(i)= v_transf(i)*dsqrt(geigen(i)) enddo write (67,'(80f10.5)') (v_transf(i),i=1,ind) #endif endif if (mod(itime,ntwx).eq.0) then write (tytul,'("time",f8.2)') totT if(mdpdb) then call pdbout(potE,tytul,ipdb) else call cartout(totT) endif endif if (rstcount.eq.1000.or.itime.eq.n_timestep) then open(irest2,file=rest2name,status='unknown') write(irest2,*) totT,EK,potE,totE,t_bath do i=1,2*nres write (irest2,'(3e15.5)') (d_t(j,i),j=1,3) enddo do i=1,2*nres write (irest2,'(3e15.5)') (dc(j,i),j=1,3) enddo close(irest2) rstcount=0 endif enddo #ifdef MPI t_MD=MPI_Wtime()-tt0 #else t_MD=tcpu()-tt0 #endif write (iout,'(//35(1h=),a10,35(1h=)/10(/a40,1pe15.5))') & ' Timing ', & 'MD calculations setup:',t_MDsetup, & 'Energy & gradient evaluation:',t_enegrad, & 'Stochastic MD setup:',t_langsetup, & 'Stochastic MD step setup:',t_sdsetup, & 'MD steps:',t_MD write (iout,'(/28(1h=),a25,27(1h=))') & ' End of MD calculation ' #ifdef TIMING_ENE write (iout,*) "time for etotal",t_etotal," elong",t_elong, & " eshort",t_eshort write (iout,*) "time_fric",time_fric," time_stoch",time_stoch, & " time_fricmatmult",time_fricmatmult," time_fsample ", & time_fsample #endif return end c------------------------------------------------------------------------------- subroutine velverlet_step(itime) c------------------------------------------------------------------------------- c Perform a single velocity Verlet step; the time step can be rescaled if c increments in accelerations exceed the threshold c------------------------------------------------------------------------------- implicit real*8 (a-h,o-z) include 'DIMENSIONS' #ifdef MPI include 'mpif.h' integer ierror,ierrcode #endif include 'COMMON.SETUP' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' include 'COMMON.TIME1' include 'COMMON.MUCA' double precision vcm(3),incr(3) double precision cm(3),L(3) integer ilen,count,rstcount external ilen character*50 tytul integer maxcount_scale /20/ common /gucio/ cm double precision stochforcvec(MAXRES6) common /stochcalc/ stochforcvec integer itime logical scale double precision HNose1,HNose,HNose_nh,H,vtnp(maxres6) double precision vtnp_(maxres6),vtnp_a(maxres6) c scale=.true. icount_scale=0 if (lang.eq.1) then call sddir_precalc else if (lang.eq.2 .or. lang.eq.3) then #ifndef LANG0 call stochastic_force(stochforcvec) #else write (iout,*) & "LANG=2 or 3 NOT SUPPORTED. Recompile without -DLANG0" #ifdef MPI call MPI_Abort(MPI_COMM_WORLD,IERROR,ERRCODE) #endif stop #endif endif itime_scal=0 do while (scale) icount_scale=icount_scale+1 if (icount_scale.gt.maxcount_scale) then write (iout,*) & "ERROR: too many attempts at scaling down the time step. ", & "amax=",amax,"epdrift=",epdrift, & "damax=",damax,"edriftmax=",edriftmax, & "d_time=",d_time call flush(iout) #ifdef MPI call MPI_Abort(MPI_COMM_WORLD,IERROR,IERRCODE) #endif stop endif c First step of the velocity Verlet algorithm if (lang.eq.2) then #ifndef LANG0 call sd_verlet1 #endif else if (lang.eq.3) then #ifndef LANG0 call sd_verlet1_ciccotti #endif else if (lang.eq.1) then call sddir_verlet1 else if (tnp1) then call tnp1_step1 else if (tnp) then call tnp_step1 else if (tnh) then call nhcint(EK,scale_nh,wdti,wdti2,wdti4,wdti8) do i=0,2*nres do j=1,3 d_t_old(j,i)=d_t_old(j,i)*scale_nh enddo enddo endif call verlet1 endif c Build the chain from the newly calculated coordinates call chainbuild_cart if (rattle) call rattle1 if (ntwe.ne.0) then if (large.and. mod(itime,ntwe).eq.0) then write (iout,*) "Cartesian and internal coordinates: step 1" call cartprint call intout write (iout,*) "dC" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(dc(j,i),j=1,3), & (dc(j,i+nres),j=1,3) enddo write (iout,*) "Accelerations" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_a(j,i),j=1,3), & (d_a(j,i+nres),j=1,3) enddo write (iout,*) "Velocities, step 1" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo endif endif #ifdef MPI tt0 = MPI_Wtime() #else tt0 = tcpu() #endif c Calculate energy and forces call zerograd call etotal(potEcomp) #ifdef TIMING_ENE #ifdef MPI t_etotal=t_etotal+MPI_Wtime()-tt0 #else t_etotal=t_etotal+tcpu()-tt0 #endif #endif E_old=potE potE=potEcomp(0)-potEcomp(20) call cartgrad c Get the new accelerations call lagrangian #ifdef MPI t_enegrad=t_enegrad+MPI_Wtime()-tt0 #else t_enegrad=t_enegrad+tcpu()-tt0 #endif c Determine maximum acceleration and scale down the timestep if needed call max_accel amax=amax/(itime_scal+1)**2 call predict_edrift(epdrift) if (amax/(itime_scal+1).gt.damax .or. epdrift.gt.edriftmax) then c Maximum acceleration or maximum predicted energy drift exceeded, rescale the time step scale=.true. ifac_time=dmax1(dlog(amax/damax),dlog(epdrift/edriftmax)) & /dlog(2.0d0)+1 itime_scal=itime_scal+ifac_time c fac_time=dmin1(damax/amax,0.5d0) fac_time=0.5d0**ifac_time d_time=d_time*fac_time if (lang.eq.2 .or. lang.eq.3) then #ifndef LANG0 c write (iout,*) "Calling sd_verlet_setup: 1" c Rescale the stochastic forces and recalculate or restore c the matrices of tinker integrator if (itime_scal.gt.maxflag_stoch) then if (large) write (iout,'(a,i5,a)') & "Calculate matrices for stochastic step;", & " itime_scal ",itime_scal if (lang.eq.2) then call sd_verlet_p_setup else call sd_verlet_ciccotti_setup endif write (iout,'(2a,i3,a,i3,1h.)') & "Warning: cannot store matrices for stochastic", & " integration because the index",itime_scal, & " is greater than",maxflag_stoch write (iout,'(2a)')"Increase MAXFLAG_STOCH or use direct", & " integration Langevin algorithm for better efficiency." else if (flag_stoch(itime_scal)) then if (large) write (iout,'(a,i5,a,l1)') & "Restore matrices for stochastic step; itime_scal ", & itime_scal," flag ",flag_stoch(itime_scal) do i=1,dimen3 do j=1,dimen3 pfric_mat(i,j)=pfric0_mat(i,j,itime_scal) afric_mat(i,j)=afric0_mat(i,j,itime_scal) vfric_mat(i,j)=vfric0_mat(i,j,itime_scal) prand_mat(i,j)=prand0_mat(i,j,itime_scal) vrand_mat1(i,j)=vrand0_mat1(i,j,itime_scal) vrand_mat2(i,j)=vrand0_mat2(i,j,itime_scal) enddo enddo else if (large) write (iout,'(2a,i5,a,l1)') & "Calculate & store matrices for stochastic step;", & " itime_scal ",itime_scal," flag ",flag_stoch(itime_scal) if (lang.eq.2) then call sd_verlet_p_setup else call sd_verlet_ciccotti_setup endif flag_stoch(ifac_time)=.true. do i=1,dimen3 do j=1,dimen3 pfric0_mat(i,j,itime_scal)=pfric_mat(i,j) afric0_mat(i,j,itime_scal)=afric_mat(i,j) vfric0_mat(i,j,itime_scal)=vfric_mat(i,j) prand0_mat(i,j,itime_scal)=prand_mat(i,j) vrand0_mat1(i,j,itime_scal)=vrand_mat1(i,j) vrand0_mat2(i,j,itime_scal)=vrand_mat2(i,j) enddo enddo endif fac_time=1.0d0/dsqrt(fac_time) do i=1,dimen3 stochforcvec(i)=fac_time*stochforcvec(i) enddo #endif else if (lang.eq.1) then c Rescale the accelerations due to stochastic forces fac_time=1.0d0/dsqrt(fac_time) do i=1,dimen3 d_as_work(i)=d_as_work(i)*fac_time enddo endif if (large) write (iout,'(a,i10,a,f8.6,a,i3,a,i3)') & "itime",itime," Timestep scaled down to ", & d_time," ifac_time",ifac_time," itime_scal",itime_scal else c Second step of the velocity Verlet algorithm if (lang.eq.2) then #ifndef LANG0 call sd_verlet2 #endif else if (lang.eq.3) then #ifndef LANG0 call sd_verlet2_ciccotti #endif else if (lang.eq.1) then call sddir_verlet2 else if (tnp1) then call tnp1_step2 else if (tnp) then call tnp_step2 else call verlet2 if (tnh) then call kinetic(EK) call nhcint(EK,scale_nh,wdti,wdti2,wdti4,wdti8) do i=0,2*nres do j=1,3 d_t(j,i)=d_t(j,i)*scale_nh enddo enddo endif endif if (rattle) call rattle2 totT=totT+d_time if (d_time.ne.d_time0) then d_time=d_time0 #ifndef LANG0 if (lang.eq.2 .or. lang.eq.3) then if (large) write (iout,'(a)') & "Restore original matrices for stochastic step" c write (iout,*) "Calling sd_verlet_setup: 2" c Restore the matrices of tinker integrator if the time step has been restored do i=1,dimen3 do j=1,dimen3 pfric_mat(i,j)=pfric0_mat(i,j,0) afric_mat(i,j)=afric0_mat(i,j,0) vfric_mat(i,j)=vfric0_mat(i,j,0) prand_mat(i,j)=prand0_mat(i,j,0) vrand_mat1(i,j)=vrand0_mat1(i,j,0) vrand_mat2(i,j)=vrand0_mat2(i,j,0) enddo enddo endif #endif endif scale=.false. endif enddo c Calculate the kinetic and the total energy and the kinetic temperature if (tnp .or. tnp1) then do i=0,2*nres do j=1,3 d_t_old(j,i)=d_t(j,i) d_t(j,i)=d_t(j,i)/s_np enddo enddo endif call kinetic(EK) totE=EK+potE c diagnostics c call kinetic1(EK1) c write (iout,*) "step",itime," EK",EK," EK1",EK1 c end diagnostics c Couple the system to Berendsen bath if needed if (tbf .and. lang.eq.0) then call verlet_bath endif kinetic_T=2.0d0/(dimen3*Rb)*EK c Backup the coordinates, velocities, and accelerations do i=0,2*nres do j=1,3 dc_old(j,i)=dc(j,i) if(.not.(tnp .or. tnp1)) d_t_old(j,i)=d_t(j,i) d_a_old(j,i)=d_a(j,i) enddo enddo if (ntwe.ne.0) then if (mod(itime,ntwe).eq.0) then if(tnp .or. tnp1) then HNose1=Hnose(EK,s_np,potE,pi_np,Q_np,t_bath,dimen3) H=(HNose1-H0)*s_np cd write (iout,'(a,10f)') "hhh",EK,s_np,potE,pi_np,H0 cd & ,EK+potE+pi_np**2/(2*Q_np)+dimen3*0.001986d0*t_bath*log(s_np) cd write (iout,*) "HHH H=",H,abs(HNose1-H0)/H0 hhh=h endif if(tnh) then HNose1=Hnose_nh(EK,potE) H=HNose1-H0 hhh=h cd write (iout,*) "HHH H=",H,abs(HNose1-H0)/H0 endif if (large) then itnp=0 do j=1,3 itnp=itnp+1 vtnp(itnp)=d_t(j,0) enddo do i=nnt,nct-1 do j=1,3 itnp=itnp+1 vtnp(itnp)=d_t(j,i) enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 itnp=itnp+1 vtnp(itnp)=d_t(j,inres) enddo endif enddo c Transform velocities from UNRES coordinate space to cartesian and Gvec c eigenvector space do i=1,dimen3 vtnp_(i)=0.0d0 vtnp_a(i)=0.0d0 do j=1,dimen3 vtnp_(i)=vtnp_(i)+Gvec(j,i)*vtnp(j) vtnp_a(i)=vtnp_a(i)+A(i,j)*vtnp(j) enddo vtnp_(i)=vtnp_(i)*dsqrt(geigen(i)) enddo do i=1,dimen3 write (iout,'("WWW",i3,3f10.5)') i,vtnp(i),vtnp_(i),vtnp_a(i) enddo write (iout,*) "Velocities, step 2" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo endif endif endif return end c------------------------------------------------------------------------------- subroutine RESPA_step(itime) c------------------------------------------------------------------------------- c Perform a single RESPA step. c------------------------------------------------------------------------------- implicit real*8 (a-h,o-z) include 'DIMENSIONS' #ifdef MPI include 'mpif.h' integer IERROR,ERRCODE #endif include 'COMMON.SETUP' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' include 'COMMON.TIME1' double precision energia_short(0:n_ene), & energia_long(0:n_ene) double precision cm(3),L(3),vcm(3),incr(3) double precision dc_old0(3,0:maxres2),d_t_old0(3,0:maxres2), & d_a_old0(3,0:maxres2) integer ilen,count,rstcount external ilen character*50 tytul integer maxcount_scale /10/ common /gucio/ cm,energia_short double precision stochforcvec(MAXRES6) common /stochcalc/ stochforcvec integer itime logical scale double precision vtnp(maxres6), vtnp_(maxres6), vtnp_a(maxres6) common /cipiszcze/ itt itt=itime if (ntwe.ne.0) then if (large.and. mod(itime,ntwe).eq.0) then write (iout,*) "***************** RESPA itime",itime write (iout,*) "Cartesian and internal coordinates: step 0" c call cartprint call pdbout(0.0d0,"cipiszcze",iout) call intout write (iout,*) "Accelerations from long-range forces" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_a(j,i),j=1,3), & (d_a(j,i+nres),j=1,3) enddo write (iout,*) "Velocities, step 0" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo endif endif c c Perform the initial RESPA step (increment velocities) c write (iout,*) "*********************** RESPA ini" if (tnp1) then call tnp_respa_step1 else if (tnp) then call tnp_respa_step1 else if (tnh.and..not.xiresp) then call nhcint(EK,scale_nh,wdti,wdti2,wdti4,wdti8) do i=0,2*nres do j=1,3 d_t(j,i)=d_t(j,i)*scale_nh enddo enddo endif call RESPA_vel endif cd if(tnp .or. tnp1) then cd write (iout,'(a,3f)') "EE1 NP S, pi",totT, s_np, pi_np cd endif if (ntwe.ne.0) then if (mod(itime,ntwe).eq.0 .and. large) then write (iout,*) "Velocities, end" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo endif endif c Compute the short-range forces #ifdef MPI tt0 =MPI_Wtime() #else tt0 = tcpu() #endif C 7/2/2009 commented out c call zerograd c call etotal_short(energia_short) if (tnp.or.tnp1) potE=energia_short(0) c call cartgrad c call lagrangian C 7/2/2009 Copy accelerations due to short-lange forces from previous MD step do i=0,2*nres do j=1,3 d_a(j,i)=d_a_short(j,i) enddo enddo if (ntwe.ne.0) then if (large.and. mod(itime,ntwe).eq.0) then write (iout,*) "energia_short",energia_short(0) write (iout,*) "Accelerations from short-range forces" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_a(j,i),j=1,3), & (d_a(j,i+nres),j=1,3) enddo endif endif #ifdef MPI t_enegrad=t_enegrad+MPI_Wtime()-tt0 #else t_enegrad=t_enegrad+tcpu()-tt0 #endif do i=0,2*nres do j=1,3 dc_old(j,i)=dc(j,i) if(.not.(tnp .or. tnp1)) d_t_old(j,i)=d_t(j,i) d_a_old(j,i)=d_a(j,i) enddo enddo c 6/30/08 A-MTS: attempt at increasing the split number do i=0,2*nres do j=1,3 dc_old0(j,i)=dc_old(j,i) d_t_old0(j,i)=d_t_old(j,i) d_a_old0(j,i)=d_a_old(j,i) enddo enddo if (ntime_split.gt.ntime_split0) ntime_split=ntime_split/2 if (ntime_split.lt.ntime_split0) ntime_split=ntime_split0 c scale=.true. d_time0=d_time do while (scale) scale=.false. c write (iout,*) "itime",itime," ntime_split",ntime_split c Split the time step d_time=d_time0/ntime_split c Perform the short-range RESPA steps (velocity Verlet increments of c positions and velocities using short-range forces) c write (iout,*) "*********************** RESPA split" do itsplit=1,ntime_split if (lang.eq.1) then call sddir_precalc else if (lang.eq.2 .or. lang.eq.3) then #ifndef LANG0 call stochastic_force(stochforcvec) #else write (iout,*) & "LANG=2 or 3 NOT SUPPORTED. Recompile without -DLANG0" #ifdef MPI call MPI_Abort(MPI_COMM_WORLD,IERROR,ERRCODE) #endif stop #endif endif c First step of the velocity Verlet algorithm if (lang.eq.2) then #ifndef LANG0 call sd_verlet1 #endif else if (lang.eq.3) then #ifndef LANG0 call sd_verlet1_ciccotti #endif else if (lang.eq.1) then call sddir_verlet1 else if (tnp1) then call tnp1_respa_i_step1 else if (tnp) then call tnp_respa_i_step1 else if (tnh.and.xiresp) then call kinetic(EK) call nhcint(EK,scale_nh,wdtii,wdtii2,wdtii4,wdtii8) do i=0,2*nres do j=1,3 d_t_old(j,i)=d_t_old(j,i)*scale_nh enddo enddo cd write(iout,*) "SSS1",itsplit,EK,scale_nh endif call verlet1 endif c Build the chain from the newly calculated coordinates call chainbuild_cart if (rattle) call rattle1 if (ntwe.ne.0) then if (large.and. mod(itime,ntwe).eq.0) then write (iout,*) "***** ITSPLIT",itsplit write (iout,*) "Cartesian and internal coordinates: step 1" call pdbout(0.0d0,"cipiszcze",iout) c call cartprint call intout write (iout,*) "Velocities, step 1" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo endif endif #ifdef MPI tt0 = MPI_Wtime() #else tt0 = tcpu() #endif c Calculate energy and forces call zerograd call etotal_short(energia_short) E_old=potE potE=energia_short(0) #ifdef TIMING_ENE #ifdef MPI t_eshort=t_eshort+MPI_Wtime()-tt0 #else t_eshort=t_eshort+tcpu()-tt0 #endif #endif call cartgrad c Get the new accelerations call lagrangian C 7/2/2009 Copy accelerations due to short-lange forces to an auxiliary array do i=0,2*nres do j=1,3 d_a_short(j,i)=d_a(j,i) enddo enddo if (ntwe.ne.0) then if (large.and. mod(itime,ntwe).eq.0) then write (iout,*)"energia_short",energia_short(0) write (iout,*) "Accelerations from short-range forces" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_a(j,i),j=1,3), & (d_a(j,i+nres),j=1,3) enddo endif endif c 6/30/08 A-MTS c Determine maximum acceleration and scale down the timestep if needed call max_accel amax=amax/ntime_split**2 call predict_edrift(epdrift) if (ntwe.gt.0 .and. large .and. mod(itime,ntwe).eq.0) & write (iout,*) "amax",amax," damax",damax, & " epdrift",epdrift," epdriftmax",epdriftmax c Exit loop and try with increased split number if the change of c acceleration is too big if (amax.gt.damax .or. epdrift.gt.edriftmax) then if (ntime_split.lt.maxtime_split) then scale=.true. ntime_split=ntime_split*2 do i=0,2*nres do j=1,3 dc_old(j,i)=dc_old0(j,i) d_t_old(j,i)=d_t_old0(j,i) d_a_old(j,i)=d_a_old0(j,i) enddo enddo write (iout,*) "acceleration/energy drift too large",amax, & epdrift," split increased to ",ntime_split," itime",itime, & " itsplit",itsplit exit else write (iout,*) & "Uh-hu. Bumpy landscape. Maximum splitting number", & maxtime_split, & " already reached!!! Trying to carry on!" endif endif #ifdef MPI t_enegrad=t_enegrad+MPI_Wtime()-tt0 #else t_enegrad=t_enegrad+tcpu()-tt0 #endif c Second step of the velocity Verlet algorithm if (lang.eq.2) then #ifndef LANG0 call sd_verlet2 #endif else if (lang.eq.3) then #ifndef LANG0 call sd_verlet2_ciccotti #endif else if (lang.eq.1) then call sddir_verlet2 else if (tnp1) then call tnp1_respa_i_step2 else if (tnp) then call tnp_respa_i_step2 else call verlet2 if (tnh.and.xiresp) then call kinetic(EK) call nhcint(EK,scale_nh,wdtii,wdtii2,wdtii4,wdtii8) do i=0,2*nres do j=1,3 d_t(j,i)=d_t(j,i)*scale_nh enddo enddo cd write(iout,*) "SSS2",itsplit,EK,scale_nh endif endif if (rattle) call rattle2 c Backup the coordinates, velocities, and accelerations if (tnp .or. tnp1) then do i=0,2*nres do j=1,3 d_t_old(j,i)=d_t(j,i) if (tnp) d_t(j,i)=d_t(j,i)/s_np if (tnp1) d_t(j,i)=d_t(j,i)/s_np enddo enddo endif do i=0,2*nres do j=1,3 dc_old(j,i)=dc(j,i) if(.not.(tnp .or. tnp1)) d_t_old(j,i)=d_t(j,i) d_a_old(j,i)=d_a(j,i) enddo enddo enddo enddo ! while scale c Restore the time step d_time=d_time0 c Compute long-range forces #ifdef MPI tt0 =MPI_Wtime() #else tt0 = tcpu() #endif call zerograd call etotal_long(energia_long) E_long=energia_long(0) potE=energia_short(0)+energia_long(0) #ifdef TIMING_ENE #ifdef MPI t_elong=t_elong+MPI_Wtime()-tt0 #else t_elong=t_elong+tcpu()-tt0 #endif #endif call cartgrad call lagrangian #ifdef MPI t_enegrad=t_enegrad+MPI_Wtime()-tt0 #else t_enegrad=t_enegrad+tcpu()-tt0 #endif c Compute accelerations from long-range forces if (ntwe.ne.0) then if (large.and. mod(itime,ntwe).eq.0) then write (iout,*) "energia_long",energia_long(0) write (iout,*) "Cartesian and internal coordinates: step 2" c call cartprint call pdbout(0.0d0,"cipiszcze",iout) call intout write (iout,*) "Accelerations from long-range forces" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_a(j,i),j=1,3), & (d_a(j,i+nres),j=1,3) enddo write (iout,*) "Velocities, step 2" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo endif endif c Compute the final RESPA step (increment velocities) c write (iout,*) "*********************** RESPA fin" if (tnp1) then call tnp_respa_step2 else if (tnp) then call tnp_respa_step2 else call RESPA_vel if (tnh.and..not.xiresp) then call kinetic(EK) call nhcint(EK,scale_nh,wdti,wdti2,wdti4,wdti8) do i=0,2*nres do j=1,3 d_t(j,i)=d_t(j,i)*scale_nh enddo enddo endif endif if (tnp .or. tnp1) then do i=0,2*nres do j=1,3 d_t(j,i)=d_t_old(j,i)/s_np enddo enddo endif c Compute the complete potential energy do i=0,n_ene potEcomp(i)=energia_short(i)+energia_long(i) enddo potE=potEcomp(0)-potEcomp(20) c potE=energia_short(0)+energia_long(0) totT=totT+d_time c Calculate the kinetic and the total energy and the kinetic temperature call kinetic(EK) totE=EK+potE c Couple the system to Berendsen bath if needed if (tbf .and. lang.eq.0) then call verlet_bath endif kinetic_T=2.0d0/(dimen3*Rb)*EK c Backup the coordinates, velocities, and accelerations if (ntwe.ne.0) then if (mod(itime,ntwe).eq.0 .and. large) then write (iout,*) "Velocities, end" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo endif if (mod(itime,ntwe).eq.0) then if(tnp .or. tnp1) then #ifndef G77 write (iout,'(a3,7f20.10)') "TTT",EK,s_np,potE,pi_np,Csplit, & E_long,energia_short(0) #else write (iout,'(a3,7f20.10)') "TTT",EK,s_np,potE,pi_np,Csplit, & E_long,energia_short(0) #endif HNose1=Hnose(EK,s_np,potE,pi_np,Q_np,t_bath,dimen3) H=(HNose1-H0)*s_np cd write (iout,'(a,10f)') "hhh",EK,s_np,potE,pi_np,H0 cd & ,EK+potE+pi_np**2/(2*Q_np)+dimen3*0.001986d0*t_bath*log(s_np) cd write (iout,*) "HHH H=",H,abs(HNose1-H0)/H0 hhh=h cd write (iout,'(a,3f)') "EE2 NP S, pi",totT, s_np, pi_np endif if(tnh) then HNose1=Hnose_nh(EK,potE) H=HNose1-H0 cd write (iout,*) "HHH H=",H,abs(HNose1-H0)/H0 hhh=h endif if (large) then itnp=0 do j=1,3 itnp=itnp+1 vtnp(itnp)=d_t(j,0) enddo do i=nnt,nct-1 do j=1,3 itnp=itnp+1 vtnp(itnp)=d_t(j,i) enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 itnp=itnp+1 vtnp(itnp)=d_t(j,inres) enddo endif enddo c Transform velocities from UNRES coordinate space to cartesian and Gvec c eigenvector space do i=1,dimen3 vtnp_(i)=0.0d0 vtnp_a(i)=0.0d0 do j=1,dimen3 vtnp_(i)=vtnp_(i)+Gvec(j,i)*vtnp(j) vtnp_a(i)=vtnp_a(i)+A(i,j)*vtnp(j) enddo vtnp_(i)=vtnp_(i)*dsqrt(geigen(i)) enddo do i=1,dimen3 write (iout,'("WWW",i3,3f10.5)') i,vtnp(i),vtnp_(i),vtnp_a(i) enddo endif endif endif return end c--------------------------------------------------------------------- subroutine RESPA_vel c First and last RESPA step (incrementing velocities using long-range c forces). implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' do j=1,3 d_t(j,0)=d_t(j,0)+0.5d0*d_a(j,0)*d_time enddo do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t(j,i)+0.5d0*d_a(j,i)*d_time enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t(j,inres)+0.5d0*d_a(j,inres)*d_time enddo endif enddo return end c----------------------------------------------------------------- subroutine verlet1 c Applying velocity Verlet algorithm - step 1 to coordinates implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision adt,adt2 #ifdef DEBUG write (iout,*) "VELVERLET1 START: DC" do i=0,nres write (iout,'(i3,3f10.5,5x,3f10.5)') i,(dc(j,i),j=1,3), & (dc(j,i+nres),j=1,3) enddo #endif do j=1,3 adt=d_a_old(j,0)*d_time adt2=0.5d0*adt dc(j,0)=dc_old(j,0)+(d_t_old(j,0)+adt2)*d_time d_t_new(j,0)=d_t_old(j,0)+adt2 d_t(j,0)=d_t_old(j,0)+adt enddo do i=nnt,nct-1 do j=1,3 adt=d_a_old(j,i)*d_time adt2=0.5d0*adt dc(j,i)=dc_old(j,i)+(d_t_old(j,i)+adt2)*d_time d_t_new(j,i)=d_t_old(j,i)+adt2 d_t(j,i)=d_t_old(j,i)+adt enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 adt=d_a_old(j,inres)*d_time adt2=0.5d0*adt dc(j,inres)=dc_old(j,inres)+(d_t_old(j,inres)+adt2)*d_time d_t_new(j,inres)=d_t_old(j,inres)+adt2 d_t(j,inres)=d_t_old(j,inres)+adt enddo endif enddo #ifdef DEBUG write (iout,*) "VELVERLET1 END: DC" do i=0,nres write (iout,'(i3,3f10.5,5x,3f10.5)') i,(dc(j,i),j=1,3), & (dc(j,i+nres),j=1,3) enddo #endif return end c--------------------------------------------------------------------- subroutine verlet2 c Step 2 of the velocity Verlet algorithm: update velocities implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' do j=1,3 d_t(j,0)=d_t_new(j,0)+0.5d0*d_a(j,0)*d_time enddo do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t_new(j,i)+0.5d0*d_a(j,i)*d_time enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t_new(j,inres)+0.5d0*d_a(j,inres)*d_time enddo endif enddo return end c----------------------------------------------------------------- subroutine sddir_precalc c Applying velocity Verlet algorithm - step 1 to coordinates implicit real*8 (a-h,o-z) include 'DIMENSIONS' #ifdef MPI include 'mpif.h' #endif include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' include 'COMMON.TIME1' double precision stochforcvec(MAXRES6) common /stochcalc/ stochforcvec c c Compute friction and stochastic forces c #ifdef MPI time00=MPI_Wtime() #else time00=tcpu() #endif call friction_force #ifdef MPI time_fric=time_fric+MPI_Wtime()-time00 time00=MPI_Wtime() #else time_fric=time_fric+tcpu()-time00 time00=tcpu() #endif call stochastic_force(stochforcvec) #ifdef MPI time_stoch=time_stoch+MPI_Wtime()-time00 #else time_stoch=time_stoch+tcpu()-time00 #endif c c Compute the acceleration due to friction forces (d_af_work) and stochastic c forces (d_as_work) c call ginv_mult(fric_work, d_af_work) call ginv_mult(stochforcvec, d_as_work) return end c--------------------------------------------------------------------- subroutine sddir_verlet1 c Applying velocity Verlet algorithm - step 1 to velocities implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' c Revised 3/31/05 AL: correlation between random contributions to c position and velocity increments included. double precision sqrt13 /0.57735026918962576451d0/ ! 1/sqrt(3) double precision adt,adt2 c c Add the contribution from BOTH friction and stochastic force to the c coordinates, but ONLY the contribution from the friction forces to velocities c do j=1,3 adt=(d_a_old(j,0)+d_af_work(j))*d_time adt2=0.5d0*adt+sqrt13*d_as_work(j)*d_time dc(j,0)=dc_old(j,0)+(d_t_old(j,0)+adt2)*d_time d_t_new(j,0)=d_t_old(j,0)+0.5d0*adt d_t(j,0)=d_t_old(j,0)+adt enddo ind=3 do i=nnt,nct-1 do j=1,3 adt=(d_a_old(j,i)+d_af_work(ind+j))*d_time adt2=0.5d0*adt+sqrt13*d_as_work(ind+j)*d_time dc(j,i)=dc_old(j,i)+(d_t_old(j,i)+adt2)*d_time d_t_new(j,i)=d_t_old(j,i)+0.5d0*adt d_t(j,i)=d_t_old(j,i)+adt enddo ind=ind+3 enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 adt=(d_a_old(j,inres)+d_af_work(ind+j))*d_time adt2=0.5d0*adt+sqrt13*d_as_work(ind+j)*d_time dc(j,inres)=dc_old(j,inres)+(d_t_old(j,inres)+adt2)*d_time d_t_new(j,inres)=d_t_old(j,inres)+0.5d0*adt d_t(j,inres)=d_t_old(j,inres)+adt enddo ind=ind+3 endif enddo return end c--------------------------------------------------------------------- subroutine sddir_verlet2 c Calculating the adjusted velocities for accelerations implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision stochforcvec(MAXRES6),d_as_work1(MAXRES6) double precision cos60 /0.5d0/, sin60 /0.86602540378443864676d0/ c Revised 3/31/05 AL: correlation between random contributions to c position and velocity increments included. c The correlation coefficients are calculated at low-friction limit. c Also, friction forces are now not calculated with new velocities. c call friction_force call stochastic_force(stochforcvec) c c Compute the acceleration due to friction forces (d_af_work) and stochastic c forces (d_as_work) c call ginv_mult(stochforcvec, d_as_work1) c c Update velocities c do j=1,3 d_t(j,0)=d_t_new(j,0)+(0.5d0*(d_a(j,0)+d_af_work(j)) & +sin60*d_as_work(j)+cos60*d_as_work1(j))*d_time enddo ind=3 do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t_new(j,i)+(0.5d0*(d_a(j,i)+d_af_work(ind+j)) & +sin60*d_as_work(ind+j)+cos60*d_as_work1(ind+j))*d_time enddo ind=ind+3 enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t_new(j,inres)+(0.5d0*(d_a(j,inres) & +d_af_work(ind+j))+sin60*d_as_work(ind+j) & +cos60*d_as_work1(ind+j))*d_time enddo ind=ind+3 endif enddo return end c--------------------------------------------------------------------- subroutine max_accel c c Find the maximum difference in the accelerations of the the sites c at the beginning and the end of the time step. c implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' double precision aux(3),accel(3),accel_old(3),dacc do j=1,3 c aux(j)=d_a(j,0)-d_a_old(j,0) accel_old(j)=d_a_old(j,0) accel(j)=d_a(j,0) enddo amax=0.0d0 do i=nnt,nct c Backbone if (i.lt.nct) then c 7/3/08 changed to asymmetric difference do j=1,3 c accel(j)=aux(j)+0.5d0*(d_a(j,i)-d_a_old(j,i)) accel_old(j)=accel_old(j)+0.5d0*d_a_old(j,i) accel(j)=accel(j)+0.5d0*d_a(j,i) c if (dabs(accel(j)).gt.amax) amax=dabs(accel(j)) if (dabs(accel(j)).gt.dabs(accel_old(j))) then dacc=dabs(accel(j)-accel_old(j)) if (dacc.gt.amax) amax=dacc endif enddo endif enddo c Side chains do j=1,3 c accel(j)=aux(j) accel_old(j)=d_a_old(j,0) accel(j)=d_a(j,0) enddo if (nnt.eq.2) then do j=1,3 accel_old(j)=accel_old(j)+d_a_old(j,1) accel(j)=accel(j)+d_a(j,1) enddo endif do i=nnt,nct if (itype(i).ne.10) then do j=1,3 c accel(j)=accel(j)+d_a(j,i+nres)-d_a_old(j,i+nres) accel_old(j)=accel_old(j)+d_a_old(j,i+nres) accel(j)=accel(j)+d_a(j,i+nres) enddo endif do j=1,3 c if (dabs(accel(j)).gt.amax) amax=dabs(accel(j)) if (dabs(accel(j)).gt.dabs(accel_old(j))) then dacc=dabs(accel(j)-accel_old(j)) if (dacc.gt.amax) amax=dacc endif enddo do j=1,3 accel_old(j)=accel_old(j)+d_a_old(j,i) accel(j)=accel(j)+d_a(j,i) c aux(j)=aux(j)+d_a(j,i)-d_a_old(j,i) enddo enddo return end c--------------------------------------------------------------------- subroutine predict_edrift(epdrift) c c Predict the drift of the potential energy c implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.MUCA' double precision epdrift,epdriftij c Drift of the potential energy epdrift=0.0d0 do i=nnt,nct c Backbone if (i.lt.nct) then do j=1,3 epdriftij=dabs((d_a(j,i)-d_a_old(j,i))*gcart(j,i)) if (lmuca) epdriftij=epdriftij*factor c write (iout,*) "back",i,j,epdriftij if (epdriftij.gt.epdrift) epdrift=epdriftij enddo endif c Side chains if (itype(i).ne.10) then do j=1,3 epdriftij= & dabs((d_a(j,i+nres)-d_a_old(j,i+nres))*gxcart(j,i)) if (lmuca) epdriftij=epdriftij*factor c write (iout,*) "side",i,j,epdriftij if (epdriftij.gt.epdrift) epdrift=epdriftij enddo endif enddo epdrift=0.5d0*epdrift*d_time*d_time c write (iout,*) "epdrift",epdrift return end c----------------------------------------------------------------------- subroutine verlet_bath c c Coupling to the thermostat by using the Berendsen algorithm c implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision T_half,fact c T_half=2.0d0/(dimen3*Rb)*EK fact=dsqrt(1.0d0+(d_time/tau_bath)*(t_bath/T_half-1.0d0)) c write(iout,*) "T_half", T_half c write(iout,*) "EK", EK c write(iout,*) "fact", fact do j=1,3 d_t(j,0)=fact*d_t(j,0) enddo do i=nnt,nct-1 do j=1,3 d_t(j,i)=fact*d_t(j,i) enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=fact*d_t(j,inres) enddo endif enddo return end c--------------------------------------------------------- subroutine init_MD c Set up the initial conditions of a MD simulation implicit real*8 (a-h,o-z) include 'DIMENSIONS' #ifdef MP include 'mpif.h' character*16 form integer IERROR,ERRCODE #endif include 'COMMON.SETUP' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' include 'COMMON.REMD' real*8 energia_long(0:n_ene), & energia_short(0:n_ene),vcm(3),incr(3),E_short double precision cm(3),L(3),xv,sigv,lowb,highb double precision varia(maxvar) character*256 qstr integer ilen external ilen character*50 tytul logical file_exist common /gucio/ cm d_time0=d_time c write(iout,*) "d_time", d_time c Compute the standard deviations of stochastic forces for Langevin dynamics c if the friction coefficients do not depend on surface area if (lang.gt.0 .and. .not.surfarea) then do i=nnt,nct-1 stdforcp(i)=stdfp*dsqrt(gamp) enddo do i=nnt,nct stdforcsc(i)=stdfsc(itype(i))*dsqrt(gamsc(itype(i))) enddo endif c Open the pdb file for snapshotshots #ifdef MPI if(mdpdb) then if (ilen(tmpdir).gt.0) & call copy_to_tmp(pref_orig(:ilen(pref_orig))//"_MD"// & liczba(:ilen(liczba))//".pdb") open(ipdb, & file=prefix(:ilen(prefix))//"_MD"//liczba(:ilen(liczba)) & //".pdb") else #ifdef NOXDR if (ilen(tmpdir).gt.0 .and. (me.eq.king .or. .not.traj1file)) & call copy_to_tmp(pref_orig(:ilen(pref_orig))//"_MD"// & liczba(:ilen(liczba))//".x") cartname=prefix(:ilen(prefix))//"_MD"//liczba(:ilen(liczba)) & //".x" #else if (ilen(tmpdir).gt.0 .and. (me.eq.king .or. .not.traj1file)) & call copy_to_tmp(pref_orig(:ilen(pref_orig))//"_MD"// & liczba(:ilen(liczba))//".cx") cartname=prefix(:ilen(prefix))//"_MD"//liczba(:ilen(liczba)) & //".cx" #endif endif #else if(mdpdb) then if (ilen(tmpdir).gt.0) & call copy_to_tmp(pref_orig(:ilen(pref_orig))//"_MD.pdb") open(ipdb,file=prefix(:ilen(prefix))//"_MD.pdb") else if (ilen(tmpdir).gt.0) & call copy_to_tmp(pref_orig(:ilen(pref_orig))//"_MD.cx") cartname=prefix(:ilen(prefix))//"_MD.cx" endif #endif if (usampl) then write (qstr,'(256(1h ))') ipos=1 do i=1,nfrag iq = qinfrag(i,iset)*10 iw = wfrag(i,iset)/100 if (iw.gt.0) then if(me.eq.king.or..not.out1file) & write (iout,*) "Frag",qinfrag(i,iset),wfrag(i,iset),iq,iw write (qstr(ipos:ipos+6),'(2h_f,i1,1h_,i1,1h_,i1)') i,iq,iw ipos=ipos+7 endif enddo do i=1,npair iq = qinpair(i,iset)*10 iw = wpair(i,iset)/100 if (iw.gt.0) then if(me.eq.king.or..not.out1file) & write (iout,*) "Pair",i,qinpair(i,iset),wpair(i,iset),iq,iw write (qstr(ipos:ipos+6),'(2h_p,i1,1h_,i1,1h_,i1)') i,iq,iw ipos=ipos+7 endif enddo c pdbname=pdbname(:ilen(pdbname)-4)//qstr(:ipos-1)//'.pdb' #ifdef NOXDR c cartname=cartname(:ilen(cartname)-2)//qstr(:ipos-1)//'.x' #else c cartname=cartname(:ilen(cartname)-3)//qstr(:ipos-1)//'.cx' #endif c statname=statname(:ilen(statname)-5)//qstr(:ipos-1)//'.stat' endif icg=1 if (rest) then if (restart1file) then if (me.eq.king) & inquire(file=mremd_rst_name,exist=file_exist) write (*,*) me," Before broadcast: file_exist",file_exist #ifdef MPI call MPI_Bcast(file_exist,1,MPI_LOGICAL,king,CG_COMM, & IERR) write (*,*) me," After broadcast: file_exist",file_exist #endif c inquire(file=mremd_rst_name,exist=file_exist) if(me.eq.king.or..not.out1file) & write(iout,*) "Initial state read by master and distributed" else if (ilen(tmpdir).gt.0) & call copy_to_tmp(pref_orig(:ilen(pref_orig))//'_' & //liczba(:ilen(liczba))//'.rst') inquire(file=rest2name,exist=file_exist) endif if(file_exist) then if(.not.restart1file) then if(me.eq.king.or..not.out1file) & write(iout,*) "Initial state will be read from file ", & rest2name(:ilen(rest2name)) call readrst endif call rescale_weights(t_bath) else if(me.eq.king.or..not.out1file)then if (restart1file) then write(iout,*) "File ",mremd_rst_name(:ilen(mremd_rst_name)), & " does not exist" else write(iout,*) "File ",rest2name(:ilen(rest2name)), & " does not exist" endif write(iout,*) "Initial velocities randomly generated" endif call random_vel totT=0.0d0 endif else c Generate initial velocities if(me.eq.king.or..not.out1file) & write(iout,*) "Initial velocities randomly generated" call random_vel totT=0.0d0 endif c rest2name = prefix(:ilen(prefix))//'.rst' if(me.eq.king.or..not.out1file)then write (iout,*) "Initial velocities" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo call flush(iout) c Zeroing the total angular momentum of the system write(iout,*) "Calling the zero-angular & momentum subroutine" endif call inertia_tensor c Getting the potential energy and forces and velocities and accelerations call vcm_vel(vcm) c write (iout,*) "velocity of the center of the mass:" c write (iout,*) (vcm(j),j=1,3) do j=1,3 d_t(j,0)=d_t(j,0)-vcm(j) enddo c Removing the velocity of the center of mass call vcm_vel(vcm) if(me.eq.king.or..not.out1file)then write (iout,*) "vcm right after adjustment:" write (iout,*) (vcm(j),j=1,3) call flush(iout) endif if (.not.rest) then call chainbuild if(iranconf.ne.0) then if (overlapsc) then print *, 'Calling OVERLAP_SC' call overlap_sc(fail) endif if (searchsc) then call sc_move(2,nres-1,10,1d10,nft_sc,etot) print *,'SC_move',nft_sc,etot if(me.eq.king.or..not.out1file) & write(iout,*) 'SC_move',nft_sc,etot endif if(dccart)then print *, 'Calling MINIM_DC' call minim_dc(etot,iretcode,nfun) else call geom_to_var(nvar,varia) print *,'Calling MINIMIZE.' call minimize(etot,varia,iretcode,nfun) call var_to_geom(nvar,varia) endif if(me.eq.king.or..not.out1file) & write(iout,*) 'SUMSL return code is',iretcode,' eval ',nfun endif endif call chainbuild_cart call kinetic(EK) if (tbf) then call verlet_bath endif kinetic_T=2.0d0/(dimen3*Rb)*EK if(me.eq.king.or..not.out1file)then call cartprint call intout endif #ifdef MPI tt0=MPI_Wtime() #else tt0=tcpu() #endif call zerograd call etotal(potEcomp) #ifdef TIMING_ENE #ifdef MPI t_etotal=t_etotal+MPI_Wtime()-tt0 #else t_etotal=t_etotal+tcpu()-tt0 #endif #endif potE=potEcomp(0) if(tnp .or. tnp1) then s_np=1.0 pi_np=0.0 HNose1=Hnose(EK,s_np,potE,pi_np,Q_np,t_bath,dimen3) H0=Hnose1 write(iout,*) 'H0= ',H0 endif if(tnh) then HNose1=Hnose_nh(EK,potE) H0=HNose1 write (iout,*) 'H0= ',H0 endif if (hmc.gt.0) then hmc_acc=0 hmc_etot=potE+EK if(me.eq.king.or..not.out1file) & write(iout,*) 'HMC',hmc_etot,potE,EK do i=1,2*nres do j=1,3 dc_hmc(j,i)=dc(j,i) enddo enddo endif call cartgrad call lagrangian call max_accel if (amax*d_time .gt. dvmax) then d_time=d_time*dvmax/amax if(me.eq.king.or..not.out1file) write (iout,*) & "Time step reduced to",d_time, & " because of too large initial acceleration." endif if(me.eq.king.or..not.out1file)then write(iout,*) "Potential energy and its components" call enerprint(potEcomp) c write(iout,*) (potEcomp(i),i=0,n_ene) endif potE=potEcomp(0)-potEcomp(20) totE=EK+potE itime=0 if (ntwe.ne.0) call statout(itime) if(me.eq.king.or..not.out1file) & write (iout,'(/a/3(a25,1pe14.5/))') "Initial:", & " Kinetic energy",EK," potential energy",potE, & " total energy",totE," maximum acceleration ", & amax if (large) then write (iout,*) "Initial coordinates" do i=1,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(c(j,i),j=1,3), & (c(j,i+nres),j=1,3) enddo write (iout,*) "Initial dC" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(dc(j,i),j=1,3), & (dc(j,i+nres),j=1,3) enddo write (iout,*) "Initial velocities" write (iout,"(13x,' backbone ',23x,' side chain')") do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_t(j,i),j=1,3), & (d_t(j,i+nres),j=1,3) enddo write (iout,*) "Initial accelerations" do i=0,nres c write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_a(j,i),j=1,3), write (iout,'(i3,3f15.10,3x,3f15.10)') i,(d_a(j,i),j=1,3), & (d_a(j,i+nres),j=1,3) enddo endif do i=0,2*nres do j=1,3 dc_old(j,i)=dc(j,i) d_t_old(j,i)=d_t(j,i) d_a_old(j,i)=d_a(j,i) enddo c write (iout,*) "dc_old",i,(dc_old(j,i),j=1,3) enddo if (RESPA) then #ifdef MPI tt0 =MPI_Wtime() #else tt0 = tcpu() #endif call zerograd call etotal_short(energia_short) #ifdef TIMING_ENE #ifdef MPI t_eshort=t_eshort+MPI_Wtime()-tt0 #else t_eshort=t_eshort+tcpu()-tt0 #endif #endif if(tnp .or. tnp1) then E_short=energia_short(0) HNose1=Hnose(EK,s_np,E_short,pi_np,Q_np,t_bath,dimen3) Csplit=Hnose1 c Csplit =110 c_new_var_csplit Csplit=H0-E_long c Csplit = H0-energia_short(0) write(iout,*) 'Csplit= ',Csplit endif call cartgrad call lagrangian if(.not.out1file .and. large) then write (iout,*) "energia_long",energia_long(0), & " energia_short",energia_short(0), & " total",energia_long(0)+energia_short(0) write (iout,*) "Initial fast-force accelerations" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_a(j,i),j=1,3), & (d_a(j,i+nres),j=1,3) enddo endif C 7/2/2009 Copy accelerations due to short-lange forces to an auxiliary array do i=0,2*nres do j=1,3 d_a_short(j,i)=d_a(j,i) enddo enddo #ifdef MPI tt0=MPI_Wtime() #else tt0=tcpu() #endif call zerograd call etotal_long(energia_long) #ifdef TIMING_ENE #ifdef MPI t_elong=t_elong+MPI_Wtime()-tt0 #else t_elong=t_elong+tcpu()-tt0 #endif #endif call cartgrad call lagrangian if(.not.out1file .and. large) then write (iout,*) "energia_long",energia_long(0) write (iout,*) "Initial slow-force accelerations" do i=0,nres write (iout,'(i3,3f10.5,3x,3f10.5)') i,(d_a(j,i),j=1,3), & (d_a(j,i+nres),j=1,3) enddo endif #ifdef MPI t_enegrad=t_enegrad+MPI_Wtime()-tt0 #else t_enegrad=t_enegrad+tcpu()-tt0 #endif endif return end c----------------------------------------------------------- subroutine random_vel implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' include 'COMMON.TIME1' double precision xv,sigv,lowb,highb c Generate random velocities from Gaussian distribution of mean 0 and std of KT/m c First generate velocities in the eigenspace of the G matrix c write (iout,*) "Calling random_vel dimen dimen3",dimen,dimen3 c call flush(iout) c write (iout,*) "RANDOM_VEL dimen",dimen xv=0.0d0 ii=0 do i=1,dimen do k=1,3 ii=ii+1 sigv=dsqrt((Rb*t_bath)/geigen(i)) lowb=-5*sigv highb=5*sigv d_t_work_new(ii)=anorm_distr(xv,sigv,lowb,highb) c write (iout,*) "i",i," ii",ii," geigen",geigen(i), c & " d_t_work_new",d_t_work_new(ii) enddo enddo call flush(iout) c diagnostics c Ek1=0.0d0 c ii=0 c do i=1,dimen c do k=1,3 c ii=ii+1 c Ek1=Ek1+0.5d0*geigen(i)*d_t_work_new(ii)**2 c enddo c enddo c write (iout,*) "Ek from eigenvectors",Ek1 c end diagnostics c Transform velocities to UNRES coordinate space do k=0,2 do i=1,dimen ind=(i-1)*3+k+1 d_t_work(ind)=0.0d0 do j=1,dimen d_t_work(ind)=d_t_work(ind) & +Gvec(i,j)*d_t_work_new((j-1)*3+k+1) enddo c write (iout,*) "i",i," ind",ind," d_t_work",d_t_work(ind) c call flush(iout) enddo enddo c Transfer to the d_t vector do j=1,3 d_t(j,0)=d_t_work(j) enddo ind=3 do i=nnt,nct-1 do j=1,3 ind=ind+1 d_t(j,i)=d_t_work(ind) enddo enddo c do i=0,nres-1 c write (iout,*) "d_t",i,(d_t(j,i),j=1,3) c enddo c call flush(iout) do i=nnt,nct if (itype(i).ne.10) then do j=1,3 ind=ind+1 d_t(j,i+nres)=d_t_work(ind) enddo endif enddo c call kinetic(EK) c write (iout,*) "Kinetic energy",Ek,EK1," kinetic temperature", c & 2.0d0/(dimen3*Rb)*EK,2.0d0/(dimen3*Rb)*EK1 c call flush(iout) return end #ifndef LANG0 c----------------------------------------------------------- subroutine sd_verlet_p_setup c Sets up the parameters of stochastic Verlet algorithm implicit real*8 (a-h,o-z) include 'DIMENSIONS' #ifdef MPI include 'mpif.h' #endif include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' include 'COMMON.TIME1' double precision emgdt(MAXRES6), & pterm,vterm,rho,rhoc,vsig, & pfric_vec(MAXRES6),vfric_vec(MAXRES6), & afric_vec(MAXRES6),prand_vec(MAXRES6), & vrand_vec1(MAXRES6),vrand_vec2(MAXRES6) logical lprn /.false./ double precision zero /1.0d-8/, gdt_radius /0.05d0/ double precision ktm #ifdef MPI tt0 = MPI_Wtime() #else tt0 = tcpu() #endif c c AL 8/17/04 Code adapted from tinker c c Get the frictional and random terms for stochastic dynamics in the c eigenspace of mass-scaled UNRES friction matrix c do i = 1, dimen gdt = fricgam(i) * d_time c c Stochastic dynamics reduces to simple MD for zero friction c if (gdt .le. zero) then pfric_vec(i) = 1.0d0 vfric_vec(i) = d_time afric_vec(i) = 0.5d0 * d_time * d_time prand_vec(i) = 0.0d0 vrand_vec1(i) = 0.0d0 vrand_vec2(i) = 0.0d0 c c Analytical expressions when friction coefficient is large c else if (gdt .ge. gdt_radius) then egdt = dexp(-gdt) pfric_vec(i) = egdt vfric_vec(i) = (1.0d0-egdt) / fricgam(i) afric_vec(i) = (d_time-vfric_vec(i)) / fricgam(i) pterm = 2.0d0*gdt - 3.0d0 + (4.0d0-egdt)*egdt vterm = 1.0d0 - egdt**2 rho = (1.0d0-egdt)**2 / sqrt(pterm*vterm) c c Use series expansions when friction coefficient is small c else gdt2 = gdt * gdt gdt3 = gdt * gdt2 gdt4 = gdt2 * gdt2 gdt5 = gdt2 * gdt3 gdt6 = gdt3 * gdt3 gdt7 = gdt3 * gdt4 gdt8 = gdt4 * gdt4 gdt9 = gdt4 * gdt5 afric_vec(i) = (gdt2/2.0d0 - gdt3/6.0d0 + gdt4/24.0d0 & - gdt5/120.0d0 + gdt6/720.0d0 & - gdt7/5040.0d0 + gdt8/40320.0d0 & - gdt9/362880.0d0) / fricgam(i)**2 vfric_vec(i) = d_time - fricgam(i)*afric_vec(i) pfric_vec(i) = 1.0d0 - fricgam(i)*vfric_vec(i) pterm = 2.0d0*gdt3/3.0d0 - gdt4/2.0d0 & + 7.0d0*gdt5/30.0d0 - gdt6/12.0d0 & + 31.0d0*gdt7/1260.0d0 - gdt8/160.0d0 & + 127.0d0*gdt9/90720.0d0 vterm = 2.0d0*gdt - 2.0d0*gdt2 + 4.0d0*gdt3/3.0d0 & - 2.0d0*gdt4/3.0d0 + 4.0d0*gdt5/15.0d0 & - 4.0d0*gdt6/45.0d0 + 8.0d0*gdt7/315.0d0 & - 2.0d0*gdt8/315.0d0 + 4.0d0*gdt9/2835.0d0 rho = sqrt(3.0d0) * (0.5d0 - 3.0d0*gdt/16.0d0 & - 17.0d0*gdt2/1280.0d0 & + 17.0d0*gdt3/6144.0d0 & + 40967.0d0*gdt4/34406400.0d0 & - 57203.0d0*gdt5/275251200.0d0 & - 1429487.0d0*gdt6/13212057600.0d0) end if c c Compute the scaling factors of random terms for the nonzero friction case c ktm = 0.5d0*d_time/fricgam(i) psig = dsqrt(ktm*pterm) / fricgam(i) vsig = dsqrt(ktm*vterm) rhoc = dsqrt(1.0d0 - rho*rho) prand_vec(i) = psig vrand_vec1(i) = vsig * rho vrand_vec2(i) = vsig * rhoc end if end do if (lprn) then write (iout,*) & "pfric_vec, vfric_vec, afric_vec, prand_vec, vrand_vec1,", & " vrand_vec2" do i=1,dimen write (iout,'(i5,6e15.5)') i,pfric_vec(i),vfric_vec(i), & afric_vec(i),prand_vec(i),vrand_vec1(i),vrand_vec2(i) enddo endif c c Transform from the eigenspace of mass-scaled friction matrix to UNRES variables c #ifndef LANG0 call eigtransf(dimen,maxres2,mt3,mt2,pfric_vec,pfric_mat) call eigtransf(dimen,maxres2,mt3,mt2,vfric_vec,vfric_mat) call eigtransf(dimen,maxres2,mt3,mt2,afric_vec,afric_mat) call eigtransf(dimen,maxres2,mt3,mt1,prand_vec,prand_mat) call eigtransf(dimen,maxres2,mt3,mt1,vrand_vec1,vrand_mat1) call eigtransf(dimen,maxres2,mt3,mt1,vrand_vec2,vrand_mat2) #endif #ifdef MPI t_sdsetup=t_sdsetup+MPI_Wtime() #else t_sdsetup=t_sdsetup+tcpu()-tt0 #endif return end c------------------------------------------------------------- subroutine eigtransf1(n,ndim,ab,d,c) implicit none integer n,ndim double precision ab(ndim,ndim,n),c(ndim,n),d(ndim) integer i,j,k do i=1,n do j=1,n c(i,j)=0.0d0 do k=1,n c(i,j)=c(i,j)+ab(k,j,i)*d(k) enddo enddo enddo return end c------------------------------------------------------------- subroutine eigtransf(n,ndim,a,b,d,c) implicit none integer n,ndim double precision a(ndim,n),b(ndim,n),c(ndim,n),d(ndim) integer i,j,k do i=1,n do j=1,n c(i,j)=0.0d0 do k=1,n c(i,j)=c(i,j)+a(i,k)*b(k,j)*d(k) enddo enddo enddo return end c------------------------------------------------------------- subroutine sd_verlet1 c Applying stochastic velocity Verlet algorithm - step 1 to velocities implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision stochforcvec(MAXRES6) common /stochcalc/ stochforcvec logical lprn /.false./ c write (iout,*) "dc_old" c do i=0,nres c write (iout,'(i5,3f10.5,5x,3f10.5)') c & i,(dc_old(j,i),j=1,3),(dc_old(j,i+nres),j=1,3) c enddo do j=1,3 dc_work(j)=dc_old(j,0) d_t_work(j)=d_t_old(j,0) d_a_work(j)=d_a_old(j,0) enddo ind=3 do i=nnt,nct-1 do j=1,3 dc_work(ind+j)=dc_old(j,i) d_t_work(ind+j)=d_t_old(j,i) d_a_work(ind+j)=d_a_old(j,i) enddo ind=ind+3 enddo do i=nnt,nct if (itype(i).ne.10) then do j=1,3 dc_work(ind+j)=dc_old(j,i+nres) d_t_work(ind+j)=d_t_old(j,i+nres) d_a_work(ind+j)=d_a_old(j,i+nres) enddo ind=ind+3 endif enddo #ifndef LANG0 if (lprn) then write (iout,*) & "pfric_mat, vfric_mat, afric_mat, prand_mat, vrand_mat1,", & " vrand_mat2" do i=1,dimen do j=1,dimen write (iout,'(2i5,6e15.5)') i,j,pfric_mat(i,j), & vfric_mat(i,j),afric_mat(i,j), & prand_mat(i,j),vrand_mat1(i,j),vrand_mat2(i,j) enddo enddo endif do i=1,dimen ddt1=0.0d0 ddt2=0.0d0 do j=1,dimen dc_work(i)=dc_work(i)+vfric_mat(i,j)*d_t_work(j) & +afric_mat(i,j)*d_a_work(j)+prand_mat(i,j)*stochforcvec(j) ddt1=ddt1+pfric_mat(i,j)*d_t_work(j) ddt2=ddt2+vfric_mat(i,j)*d_a_work(j) enddo d_t_work_new(i)=ddt1+0.5d0*ddt2 d_t_work(i)=ddt1+ddt2 enddo #endif do j=1,3 dc(j,0)=dc_work(j) d_t(j,0)=d_t_work(j) enddo ind=3 do i=nnt,nct-1 do j=1,3 dc(j,i)=dc_work(ind+j) d_t(j,i)=d_t_work(ind+j) enddo ind=ind+3 enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 dc(j,inres)=dc_work(ind+j) d_t(j,inres)=d_t_work(ind+j) enddo ind=ind+3 endif enddo return end c-------------------------------------------------------------------------- subroutine sd_verlet2 c Calculating the adjusted velocities for accelerations implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision stochforcvec(MAXRES6),stochforcvecV(MAXRES6) common /stochcalc/ stochforcvec c c Compute the stochastic forces which contribute to velocity change c call stochastic_force(stochforcvecV) #ifndef LANG0 do i=1,dimen ddt1=0.0d0 ddt2=0.0d0 do j=1,dimen ddt1=ddt1+vfric_mat(i,j)*d_a_work(j) ddt2=ddt2+vrand_mat1(i,j)*stochforcvec(j)+ & vrand_mat2(i,j)*stochforcvecV(j) enddo d_t_work(i)=d_t_work_new(i)+0.5d0*ddt1+ddt2 enddo #endif do j=1,3 d_t(j,0)=d_t_work(j) enddo ind=3 do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t_work(ind+j) enddo ind=ind+3 enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t_work(ind+j) enddo ind=ind+3 endif enddo return end c----------------------------------------------------------- subroutine sd_verlet_ciccotti_setup c Sets up the parameters of stochastic velocity Verlet algorithmi; Ciccotti's c version implicit real*8 (a-h,o-z) include 'DIMENSIONS' #ifdef MPI include 'mpif.h' #endif include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' include 'COMMON.TIME1' double precision emgdt(MAXRES6), & pterm,vterm,rho,rhoc,vsig, & pfric_vec(MAXRES6),vfric_vec(MAXRES6), & afric_vec(MAXRES6),prand_vec(MAXRES6), & vrand_vec1(MAXRES6),vrand_vec2(MAXRES6) logical lprn /.false./ double precision zero /1.0d-8/, gdt_radius /0.05d0/ double precision ktm #ifdef MPI tt0 = MPI_Wtime() #else tt0 = tcpu() #endif c c AL 8/17/04 Code adapted from tinker c c Get the frictional and random terms for stochastic dynamics in the c eigenspace of mass-scaled UNRES friction matrix c do i = 1, dimen write (iout,*) "i",i," fricgam",fricgam(i) gdt = fricgam(i) * d_time c c Stochastic dynamics reduces to simple MD for zero friction c if (gdt .le. zero) then pfric_vec(i) = 1.0d0 vfric_vec(i) = d_time afric_vec(i) = 0.5d0*d_time*d_time prand_vec(i) = afric_vec(i) vrand_vec2(i) = vfric_vec(i) c c Analytical expressions when friction coefficient is large c else egdt = dexp(-gdt) pfric_vec(i) = egdt vfric_vec(i) = dexp(-0.5d0*gdt)*d_time afric_vec(i) = 0.5d0*dexp(-0.25d0*gdt)*d_time*d_time prand_vec(i) = afric_vec(i) vrand_vec2(i) = vfric_vec(i) c c Compute the scaling factors of random terms for the nonzero friction case c c ktm = 0.5d0*d_time/fricgam(i) c psig = dsqrt(ktm*pterm) / fricgam(i) c vsig = dsqrt(ktm*vterm) c prand_vec(i) = psig*afric_vec(i) c vrand_vec2(i) = vsig*vfric_vec(i) end if end do if (lprn) then write (iout,*) & "pfric_vec, vfric_vec, afric_vec, prand_vec, vrand_vec1,", & " vrand_vec2" do i=1,dimen write (iout,'(i5,6e15.5)') i,pfric_vec(i),vfric_vec(i), & afric_vec(i),prand_vec(i),vrand_vec1(i),vrand_vec2(i) enddo endif c c Transform from the eigenspace of mass-scaled friction matrix to UNRES variables c call eigtransf(dimen,maxres2,mt3,mt2,pfric_vec,pfric_mat) call eigtransf(dimen,maxres2,mt3,mt2,vfric_vec,vfric_mat) call eigtransf(dimen,maxres2,mt3,mt2,afric_vec,afric_mat) call eigtransf(dimen,maxres2,mt3,mt1,prand_vec,prand_mat) call eigtransf(dimen,maxres2,mt3,mt1,vrand_vec2,vrand_mat2) #ifdef MPI t_sdsetup=t_sdsetup+MPI_Wtime() #else t_sdsetup=t_sdsetup+tcpu()-tt0 #endif return end c------------------------------------------------------------- subroutine sd_verlet1_ciccotti c Applying stochastic velocity Verlet algorithm - step 1 to velocities implicit real*8 (a-h,o-z) include 'DIMENSIONS' #ifdef MPI include 'mpif.h' #endif include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision stochforcvec(MAXRES6) common /stochcalc/ stochforcvec logical lprn /.false./ c write (iout,*) "dc_old" c do i=0,nres c write (iout,'(i5,3f10.5,5x,3f10.5)') c & i,(dc_old(j,i),j=1,3),(dc_old(j,i+nres),j=1,3) c enddo do j=1,3 dc_work(j)=dc_old(j,0) d_t_work(j)=d_t_old(j,0) d_a_work(j)=d_a_old(j,0) enddo ind=3 do i=nnt,nct-1 do j=1,3 dc_work(ind+j)=dc_old(j,i) d_t_work(ind+j)=d_t_old(j,i) d_a_work(ind+j)=d_a_old(j,i) enddo ind=ind+3 enddo do i=nnt,nct if (itype(i).ne.10) then do j=1,3 dc_work(ind+j)=dc_old(j,i+nres) d_t_work(ind+j)=d_t_old(j,i+nres) d_a_work(ind+j)=d_a_old(j,i+nres) enddo ind=ind+3 endif enddo #ifndef LANG0 if (lprn) then write (iout,*) & "pfric_mat, vfric_mat, afric_mat, prand_mat, vrand_mat1,", & " vrand_mat2" do i=1,dimen do j=1,dimen write (iout,'(2i5,6e15.5)') i,j,pfric_mat(i,j), & vfric_mat(i,j),afric_mat(i,j), & prand_mat(i,j),vrand_mat1(i,j),vrand_mat2(i,j) enddo enddo endif do i=1,dimen ddt1=0.0d0 ddt2=0.0d0 do j=1,dimen dc_work(i)=dc_work(i)+vfric_mat(i,j)*d_t_work(j) & +afric_mat(i,j)*d_a_work(j)+prand_mat(i,j)*stochforcvec(j) ddt1=ddt1+pfric_mat(i,j)*d_t_work(j) ddt2=ddt2+vfric_mat(i,j)*d_a_work(j) enddo d_t_work_new(i)=ddt1+0.5d0*ddt2 d_t_work(i)=ddt1+ddt2 enddo #endif do j=1,3 dc(j,0)=dc_work(j) d_t(j,0)=d_t_work(j) enddo ind=3 do i=nnt,nct-1 do j=1,3 dc(j,i)=dc_work(ind+j) d_t(j,i)=d_t_work(ind+j) enddo ind=ind+3 enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 dc(j,inres)=dc_work(ind+j) d_t(j,inres)=d_t_work(ind+j) enddo ind=ind+3 endif enddo return end c-------------------------------------------------------------------------- subroutine sd_verlet2_ciccotti c Calculating the adjusted velocities for accelerations implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' #ifndef LANG0 include 'COMMON.LANGEVIN' #else include 'COMMON.LANGEVIN.lang0' #endif include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision stochforcvec(MAXRES6),stochforcvecV(MAXRES6) common /stochcalc/ stochforcvec c c Compute the stochastic forces which contribute to velocity change c call stochastic_force(stochforcvecV) #ifndef LANG0 do i=1,dimen ddt1=0.0d0 ddt2=0.0d0 do j=1,dimen ddt1=ddt1+vfric_mat(i,j)*d_a_work(j) c ddt2=ddt2+vrand_mat2(i,j)*stochforcvecV(j) ddt2=ddt2+vrand_mat2(i,j)*stochforcvec(j) enddo d_t_work(i)=d_t_work_new(i)+0.5d0*ddt1+ddt2 enddo #endif do j=1,3 d_t(j,0)=d_t_work(j) enddo ind=3 do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t_work(ind+j) enddo ind=ind+3 enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t_work(ind+j) enddo ind=ind+3 endif enddo return end #endif c------------------------------------------------------ double precision function HNose(ek,s,e,pi,Q,t_bath,dimenl) implicit none double precision ek,s,e,pi,Q,t_bath,Rb integer dimenl Rb=0.001986d0 HNose=ek+e+pi**2/(2*Q)+dimenl*Rb*t_bath*log(s) c print '(6f15.5,i5,a2,2f15.5)',ek,s,e,pi,Q,t_bath,dimenl,"--", c & pi**2/(2*Q),dimenl*Rb*t_bath*log(s) return end c----------------------------------------------------------------- double precision function HNose_nh(eki,e) implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.MD' HNose_nh=eki+e+dimen3*Rb*t_bath*xlogs(1)+qmass(1)*vlogs(1)**2/2 do i=2,nnos HNose_nh=HNose_nh+qmass(i)*vlogs(i)**2/2+Rb*t_bath*xlogs(i) enddo c write(4,'(5e15.5)') c & vlogs(1),xlogs(1),HNose,eki,e return end c----------------------------------------------------------------- SUBROUTINE NHCINT(akin,scale,wdti,wdti2,wdti4,wdti8) implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.MD' double precision akin,gnkt,dt,aa,gkt,scale double precision wdti(maxyosh),wdti2(maxyosh), & wdti4(maxyosh),wdti8(maxyosh) integer i,iresn,iyosh,inos,nnos1 dt=d_time nnos1=nnos+1 GKT = Rb*t_bath GNKT = dimen3*GKT akin=akin*2 C THIS ROUTINE DOES THE NOSE-HOOVER PART OF THE C INTEGRATION FROM t=0 TO t=DT/2 C GET THE TOTAL KINETIC ENERGY SCALE = 1.D0 c CALL GETKINP(MASS,VX,VY,VZ,AKIN) C UPDATE THE FORCES GLOGS(1) = (AKIN - GNKT)/QMASS(1) C START THE MULTIPLE TIME STEP PROCEDURE DO IRESN = 1,NRESN DO IYOSH = 1,NYOSH C UPDATE THE THERMOSTAT VELOCITIES VLOGS(NNOS) = VLOGS(NNOS) + GLOGS(NNOS)*WDTI4(IYOSH) DO INOS = 1,NNOS-1 AA = EXP(-WDTI8(IYOSH)*VLOGS(NNOS1-INOS) ) VLOGS(NNOS-INOS) = VLOGS(NNOS-INOS)*AA*AA & + WDTI4(IYOSH)*GLOGS(NNOS-INOS)*AA ENDDO C UPDATE THE PARTICLE VELOCITIES AA = EXP(-WDTI2(IYOSH)*VLOGS(1) ) SCALE = SCALE*AA C UPDATE THE FORCES GLOGS(1) = (SCALE*SCALE*AKIN - GNKT)/QMASS(1) C UPDATE THE THERMOSTAT POSITIONS DO INOS = 1,NNOS XLOGS(INOS) = XLOGS(INOS) + VLOGS(INOS)*WDTI2(IYOSH) ENDDO C UPDATE THE THERMOSTAT VELOCITIES DO INOS = 1,NNOS-1 AA = EXP(-WDTI8(IYOSH)*VLOGS(INOS+1) ) VLOGS(INOS) = VLOGS(INOS)*AA*AA & + WDTI4(IYOSH)*GLOGS(INOS)*AA GLOGS(INOS+1) = (QMASS(INOS)*VLOGS(INOS)*VLOGS(INOS) & -GKT)/QMASS(INOS+1) ENDDO VLOGS(NNOS) = VLOGS(NNOS) + GLOGS(NNOS)*WDTI4(IYOSH) ENDDO ENDDO C UPDATE THE PARTICLE VELOCITIES c outside of this subroutine c DO I = 1,N c VX(I) = VX(I)*SCALE c VY(I) = VY(I)*SCALE c VZ(I) = VZ(I)*SCALE c ENDDO RETURN END c----------------------------------------------------------------- subroutine tnp1_respa_i_step1 c Applying Nose-Poincare algorithm - step 1 to coordinates c JPSJ 70 75 (2001) S. Nose c c d_t is not updated here c implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision adt,adt2,tmp tmp=1+pi_np/(2*Q_np)*0.5*d_time s12_np=s_np*tmp**2 pistar=pi_np/tmp s12_dt=d_time/s12_np d_time_s12=d_time*0.5*s12_np do j=1,3 d_t_new(j,0)=d_t_old(j,0)+d_a_old(j,0)*d_time_s12 dc(j,0)=dc_old(j,0)+d_t_new(j,0)*s12_dt enddo do i=nnt,nct-1 do j=1,3 d_t_new(j,i)=d_t_old(j,i)+d_a_old(j,i)*d_time_s12 dc(j,i)=dc_old(j,i)+d_t_new(j,i)*s12_dt enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t_new(j,inres)=d_t_old(j,inres)+d_a_old(j,inres)*d_time_s12 dc(j,inres)=dc_old(j,inres)+d_t_new(j,inres)*s12_dt enddo endif enddo return end c--------------------------------------------------------------------- subroutine tnp1_respa_i_step2 c Step 2 of the velocity Verlet algorithm: update velocities implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision d_time_s12 do i=0,2*nres do j=1,3 d_t(j,i)=d_t_new(j,i) enddo enddo call kinetic(EK) EK=EK/s12_np**2 d_time_s12=0.5d0*s12_np*d_time do j=1,3 d_t(j,0)=d_t_new(j,0)+d_a(j,0)*d_time_s12 enddo do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t_new(j,i)+d_a(j,i)*d_time_s12 enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t_new(j,inres)+d_a(j,inres)*d_time_s12 enddo endif enddo pistar=pistar+(EK-0.5*(E_old+potE) & -dimen3*Rb*t_bath*log(s12_np)+Csplit-dimen3*Rb*t_bath)*d_time tmp=1+pistar/(2*Q_np)*0.5*d_time s_np=s12_np*tmp**2 pi_np=pistar/tmp return end c------------------------------------------------------- subroutine tnp1_step1 c Applying Nose-Poincare algorithm - step 1 to coordinates c JPSJ 70 75 (2001) S. Nose c c d_t is not updated here c implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision adt,adt2,tmp tmp=1+pi_np/(2*Q_np)*0.5*d_time s12_np=s_np*tmp**2 pistar=pi_np/tmp s12_dt=d_time/s12_np d_time_s12=d_time*0.5*s12_np do j=1,3 d_t_new(j,0)=d_t_old(j,0)+d_a_old(j,0)*d_time_s12 dc(j,0)=dc_old(j,0)+d_t_new(j,0)*s12_dt enddo do i=nnt,nct-1 do j=1,3 d_t_new(j,i)=d_t_old(j,i)+d_a_old(j,i)*d_time_s12 dc(j,i)=dc_old(j,i)+d_t_new(j,i)*s12_dt enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t_new(j,inres)=d_t_old(j,inres)+d_a_old(j,inres)*d_time_s12 dc(j,inres)=dc_old(j,inres)+d_t_new(j,inres)*s12_dt enddo endif enddo return end c--------------------------------------------------------------------- subroutine tnp1_step2 c Step 2 of the velocity Verlet algorithm: update velocities implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision d_time_s12 do i=0,2*nres do j=1,3 d_t(j,i)=d_t_new(j,i) enddo enddo call kinetic(EK) EK=EK/s12_np**2 d_time_s12=0.5d0*s12_np*d_time do j=1,3 d_t(j,0)=d_t_new(j,0)+d_a(j,0)*d_time_s12 enddo do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t_new(j,i)+d_a(j,i)*d_time_s12 enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t_new(j,inres)+d_a(j,inres)*d_time_s12 enddo endif enddo cd write(iout,*) 'pistar',pistar,EK,E_old,potE,s12_np pistar=pistar+(EK-0.5*(E_old+potE) & -dimen3*Rb*t_bath*log(s12_np)+H0-dimen3*Rb*t_bath)*d_time tmp=1+pistar/(2*Q_np)*0.5*d_time s_np=s12_np*tmp**2 pi_np=pistar/tmp return end c----------------------------------------------------------------- subroutine tnp_respa_i_step1 c Applying Nose-Poincare algorithm - step 1 to coordinates c J.Comput.Phys. 151 114 (1999) S.D.Bond B.J.Leimkuhler B.B.Laird c c d_t is not updated here, it is destroyed c implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision C_np,d_time_s,tmp,d_time_ss d_time_s=d_time*0.5*s_np ct2 d_time_s=d_time*0.5*s12_np do j=1,3 d_t_new(j,0)=d_t_old(j,0)+d_a_old(j,0)*d_time_s enddo do i=nnt,nct-1 do j=1,3 d_t_new(j,i)=d_t_old(j,i)+d_a_old(j,i)*d_time_s enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t_new(j,inres)=d_t_old(j,inres)+d_a_old(j,inres)*d_time_s enddo endif enddo do i=0,2*nres do j=1,3 d_t(j,i)=d_t_new(j,i) enddo enddo call kinetic(EK) EK=EK/s_np**2 C_np=0.5*d_time*(dimen3*Rb*t_bath*(1.0+log(s_np))-EK+potE-Csplit) & -pi_np pistar=-2.0*C_np/(1.0+sqrt(1.0-C_np*d_time/Q_np)) tmp=0.5*d_time*pistar/Q_np s12_np=s_np*(1.0+tmp)/(1.0-tmp) d_time_ss=0.5*d_time*(1.0/s12_np+1.0/s_np) ct2 d_time_ss=d_time/s12_np c d_time_ss=0.5*d_time*(1.0/sold_np+1.0/s_np) do j=1,3 dc(j,0)=dc_old(j,0)+d_t_new(j,0)*d_time_ss enddo do i=nnt,nct-1 do j=1,3 dc(j,i)=dc_old(j,i)+d_t_new(j,i)*d_time_ss enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 dc(j,inres)=dc_old(j,inres)+d_t_new(j,inres)*d_time_ss enddo endif enddo return end c--------------------------------------------------------------------- subroutine tnp_respa_i_step2 c Step 2 of the velocity Verlet algorithm: update velocities implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision d_time_s EK=EK*(s_np/s12_np)**2 HNose1=Hnose(EK,s12_np,potE,pistar,Q_np,t_bath,dimen3) pi_np=pistar+0.5*d_time*(2*EK-dimen3*Rb*t_bath & -HNose1+Csplit) cr print '(a,5f)','i_step2',EK,potE,HNose1,pi_np,E_long d_time_s=d_time*0.5*s12_np c d_time_s=d_time*0.5*s_np do j=1,3 d_t(j,0)=d_t_new(j,0)+d_a(j,0)*d_time_s enddo do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t_new(j,i)+d_a(j,i)*d_time_s enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t_new(j,inres)+d_a(j,inres)*d_time_s enddo endif enddo s_np=s12_np return end c----------------------------------------------------------------- subroutine tnp_respa_step1 c Applying Nose-Poincare algorithm - step 1 to vel for RESPA c J.Comput.Phys. 151 114 (1999) S.D.Bond B.J.Leimkuhler B.B.Laird c c d_t is not updated here, it is destroyed c implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision C_np,d_time_s,tmp,d_time_ss double precision energia(0:n_ene) d_time_s=d_time*0.5*s_np do j=1,3 d_t_old(j,0)=d_t_old(j,0)+d_a(j,0)*d_time_s enddo do i=nnt,nct-1 do j=1,3 d_t_old(j,i)=d_t_old(j,i)+d_a(j,i)*d_time_s enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t_old(j,inres)=d_t_old(j,inres)+d_a(j,inres)*d_time_s enddo endif enddo c C_np=0.5*d_time*(dimen3*Rb*t_bath*(1.0+log(s_np))-EK+potE-H0) c & -pi_np c c pistar=-2.0*C_np/(1.0+sqrt(1.0-C_np*d_time/Q_np)) c tmp=0.5*d_time*pistar/Q_np c s12_np=s_np*(1.0+tmp)/(1.0-tmp) c write(iout,*) 'tnp_respa_step1',s_np,s12_np,EK,potE,C_np,pistar,tmp ct1 pi_np=pistar c sold_np=s_np c s_np=s12_np c------------------------------------- c test of reviewer's comment pi_np=pi_np-0.5*d_time*(E_long+Csplit-H0) cr print '(a,3f)','1 pi_np,s_np',pi_np,s_np,E_long c------------------------------------- return end c--------------------------------------------------------------------- subroutine tnp_respa_step2 c Step 2 of the velocity Verlet algorithm: update velocities for RESPA implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision d_time_s ct1 s12_np=s_np ct2 pistar=pi_np ct call kinetic(EK) ct HNose1=Hnose(EK,s12_np,potE,pistar,Q_np,t_bath,dimen3) ct pi_np=pistar+0.5*d_time*(2*EK-dimen3*Rb*t_bath) ct & -0.5*d_time*(HNose1-H0) c------------------------------------- c test of reviewer's comment pi_np=pi_np-0.5*d_time*(E_long+Csplit-H0) cr print '(a,3f)','2 pi_np,s_np',pi_np,s_np,E_long c------------------------------------- d_time_s=d_time*0.5*s_np do j=1,3 d_t_old(j,0)=d_t_old(j,0)+d_a(j,0)*d_time_s enddo do i=nnt,nct-1 do j=1,3 d_t_old(j,i)=d_t_old(j,i)+d_a(j,i)*d_time_s enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t_old(j,inres)=d_t_old(j,inres)+d_a(j,inres)*d_time_s enddo endif enddo cd s_np=s12_np return end c--------------------------------------------------------------------- subroutine tnp_step1 c Applying Nose-Poincare algorithm - step 1 to coordinates c J.Comput.Phys. 151 114 (1999) S.D.Bond B.J.Leimkuhler B.B.Laird c c d_t is not updated here, it is destroyed c implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision C_np,d_time_s,tmp,d_time_ss d_time_s=d_time*0.5*s_np do j=1,3 d_t_new(j,0)=d_t_old(j,0)+d_a_old(j,0)*d_time_s enddo do i=nnt,nct-1 do j=1,3 d_t_new(j,i)=d_t_old(j,i)+d_a_old(j,i)*d_time_s enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t_new(j,inres)=d_t_old(j,inres)+d_a_old(j,inres)*d_time_s enddo endif enddo do i=0,2*nres do j=1,3 d_t(j,i)=d_t_new(j,i) enddo enddo call kinetic(EK) EK=EK/s_np**2 C_np=0.5*d_time*(dimen3*Rb*t_bath*(1.0+log(s_np))-EK+potE-H0) & -pi_np pistar=-2.0*C_np/(1.0+sqrt(1.0-C_np*d_time/Q_np)) tmp=0.5*d_time*pistar/Q_np s12_np=s_np*(1.0+tmp)/(1.0-tmp) c write(iout,*) 'tnp_step1',s_np,s12_np,EK,potE,C_np,pistar,tmp d_time_ss=0.5*d_time*(1.0/s12_np+1.0/s_np) do j=1,3 dc(j,0)=dc_old(j,0)+d_t_new(j,0)*d_time_ss enddo do i=nnt,nct-1 do j=1,3 dc(j,i)=dc_old(j,i)+d_t_new(j,i)*d_time_ss enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 dc(j,inres)=dc_old(j,inres)+d_t_new(j,inres)*d_time_ss enddo endif enddo return end c----------------------------------------------------------------- subroutine tnp_step2 c Step 2 of the velocity Verlet algorithm: update velocities implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.VAR' include 'COMMON.MD' include 'COMMON.CHAIN' include 'COMMON.DERIV' include 'COMMON.GEO' include 'COMMON.LOCAL' include 'COMMON.INTERACT' include 'COMMON.IOUNITS' include 'COMMON.NAMES' double precision d_time_s EK=EK*(s_np/s12_np)**2 HNose1=Hnose(EK,s12_np,potE,pistar,Q_np,t_bath,dimen3) pi_np=pistar+0.5*d_time*(2*EK-dimen3*Rb*t_bath) & -0.5*d_time*(HNose1-H0) cd write(iout,'(a,4f)') 'mmm',EK,potE,HNose1,pi_np d_time_s=d_time*0.5*s12_np do j=1,3 d_t(j,0)=d_t_new(j,0)+d_a(j,0)*d_time_s enddo do i=nnt,nct-1 do j=1,3 d_t(j,i)=d_t_new(j,i)+d_a(j,i)*d_time_s enddo enddo do i=nnt,nct if (itype(i).ne.10) then inres=i+nres do j=1,3 d_t(j,inres)=d_t_new(j,inres)+d_a(j,inres)*d_time_s enddo endif enddo s_np=s12_np return end subroutine hmc_test(itime) implicit real*8 (a-h,o-z) include 'DIMENSIONS' include 'COMMON.CONTROL' include 'COMMON.MD' include 'COMMON.CHAIN' hmc_acc=hmc_acc+1 delta=-(potE+EK-hmc_etot)/(Rb*t_bath) if (delta .lt. -50.0d0) then delta=0.0d0 else delta=dexp(delta) endif xxx=ran_number(0.0d0,1.0d0) if (me.eq.king .or. .not. out1file) & write(iout,'(a8,i5,6f10.4)') & 'HMC',itime,potE+EK,potE,EK,hmc_etot,delta,xxx if (delta .le. xxx) then do i=1,2*nres do j=1,3 dc(j,i)=dc_hmc(j,i) enddo enddo itime=itime-hmc totT=totThmc else if (me.eq.king .or. .not. out1file) & write(iout,*) 'HMC accepting new' totThmc=totT do i=1,2*nres do j=1,3 dc_hmc(j,i)=dc(j,i) enddo enddo endif call chainbuild_cart call random_vel do i=0,2*nres do j=1,3 d_t_old(j,i)=d_t(j,i) enddo enddo call kinetic(EK) kinetic_T=2.0d0/(dimen3*Rb)*EK call etotal(potEcomp) potE=potEcomp(0) hmc_etot=potE+EK if (me.eq.king .or. .not. out1file) & write(iout,'(a8,i5,3f10.4)')'HMC new',itime,potE+EK,potE,EK return end