CORREL: principal inertial axis reorientation

Joined: Sep 2003
Posts: 8,499
Forum Member

OP
Forum Member
Joined: Sep 2003
Posts: 8,499 
The following example illustrates some CORREL usage, calculating autocorrelation functions for the principal axis of the moment of inertia of a group of selected atoms, as well as that for an endtoend vector.
* reorientation of the principal axis from moment of inertia tensor * arg N == last file no. *
! INIT FOR SUCROSE NEUTRON CONFIG IN CUBIC WATER BOX stream initbox30.str
calc mxt @N * 100 calc tot @mxt / 4
correl maxt @mxt noupdate maxs 32 ! VECTORS TO EXTRACT enter pv inertia sele segid S end enter ee vect xyz S 1 C4 S 2 C4 ! STORAGE FOR CORR FXNS; ORDER OF DECLARATION MATTERS enter ci1 zero edit ci1 total @tot enter ci2 dupl ci1 enter ce1 dupl ci1 enter ce2 dupl ci1
! OPEN THE INPUT TRAJ FILES, 1 THRU N set num 1 set uni 31 label olp open unit @uni write file name dyn@num.trj incr uni by 1 incr num by 1 if num .le. @N goto olp
! PROCESS THE TRAJ FILES, FILE THE TIME SERIES WITH DATA traj firstu 31 nunit @N
! NORMALIZE VECTORS AND COMPUTE P1 AND P2 AUTOCORR FXNS mantim pv norm ! NORM corfun pv pv p1 total @tot ! P1 mantim ci1 copy corr ! STORE RESULT (corr) IN ci1 corfun pv pv p2 total @tot ! P2 mantim ci2 copy corr ! STORE
mantim ee norm corfun ee ee p1 total @tot mantim ce1 copy corr corfun ee ee p2 total @tot mantim ce2 copy corr
! THE EDIT COMMAND COMBINES ci1 ci2 ce1 ce2; WRITTEN TO ONE FILE edit ci1 veccod 4 open unit 12 write card name mireo.cor write ci1 unit 12 dumb time * dummy * end stop
Rick Venable computational chemist



Re: CORREL: principal inertial axis reorientation

Joined: Feb 2005
Posts: 7
Forum Member

Forum Member
Joined: Feb 2005
Posts: 7 
Rick, would you please to add some explanations how one should get all 3 rotational ACF?
thank you.



Re: CORREL: principal inertial axis reorientation

Joined: Sep 2003
Posts: 8,499
Forum Member

OP
Forum Member
Joined: Sep 2003
Posts: 8,499 
We investigated that, and found that the interchange of the 2 minor axes was a severe problem with no easy algorithmic solution; these sudden nonphysical shifts of 90 deg made computing correlation functions impractical for the molecules we were investigating. I suggesting picking surrogate vectors for the VECT time series type, and compute the time correlation function of those vectors instead.
Rick Venable computational chemist



Re: CORREL: principal inertial axis reorientation

Joined: Sep 2003
Posts: 4,794 Likes: 2
Forum Member

Forum Member
Joined: Sep 2003
Posts: 4,794 Likes: 2 
In recent versions (c31b1 and later I believe) there is an option to get all three principal axes in correl; the code that corrects for spurious interchanges does however have a couple of silly bugs (fixed in the next release), but as Rick points out it is not always possible to spot these swaps correctly (in particular if you are using a large SKIP value in the analysis). In our limited experience with small proteins axis swapping has not been a problem.
Lennart Nilsson Karolinska Institutet Stockholm, Sweden



Re: CORREL: principal inertial axis reorientation

Joined: Sep 2003
Posts: 8,499
Forum Member

OP
Forum Member
Joined: Sep 2003
Posts: 8,499 
With a reasonable interval between time points, it should be okay for large, fairly rigid macromolecules; it's more of a problem for short peptides, oligosaccharides, lipids, neat alkanes, neat ethers, etc.
Rick Venable computational chemist



Re: CORREL: principal inertial axis reorientation

Joined: Jul 2004
Posts: 165
Forum Member

Forum Member
Joined: Jul 2004
Posts: 165 
dear rick, I got the following column for the PN vector of the lipid head group. The 1st column should be a time, but i am not so sure about the 2nd and 3rd column? thank you for your help. 0.000000 1.000000 1.000000 5.000000 0.971475 0.918035 10.000000 0.960319 0.888238 15.000000 0.951557 0.865780 20.000000 0.943885 0.846703 25.000000 0.937154 0.830445 30.000000 0.930919 0.815697 35.000000 0.925124 0.802287 40.000000 0.919751 0.790125 45.000000 0.914676 0.778877 50.000000 0.909856 0.768361 55.000000 0.905271 0.758557 60.000000 0.900799 0.749059



Re: CORREL: principal inertial axis reorientation

Joined: Sep 2003
Posts: 8,499
Forum Member

OP
Forum Member
Joined: Sep 2003
Posts: 8,499 
They are the P1 and P2 time correlation functions, respectively; the latter is most often the one of interest for time decay.



Re: CORREL: principal inertial axis reorientation

Joined: Jul 2004
Posts: 165
Forum Member

Forum Member
Joined: Jul 2004
Posts: 165 
thank rick,
could you please give me more hint?
one could estimate the correlation time by an exponential fit to the graph, right? but my simulation time is still too short (50 ns) to get the decay to zero. how should I do to get the reasonable estimation to represent relaxation time for the molecule.
thank you.



Re: CORREL: principal inertial axis reorientation

Joined: Sep 2003
Posts: 8,499
Forum Member

OP
Forum Member
Joined: Sep 2003
Posts: 8,499 
That vector does not have unrestricted motion in all directions, and should not decay to zero; it will have a long time plateau value, which can be interesting in some cases. For example, with the CH vectors in lipids, the plateau value is the SCD value, i.e. the deuterium order parameter.
The relaxation time is not the decay to zero, but to 1/e (roughly 0.38).
For the PN vector, the decay will likely be a multiexponential; there a several types of motion that contribute, possibly on overlapping time scales. These are not rigid bodies, and this vector probably represents changes in rotamer states in the alpha chain as much as anything. The vector would not represent whole molecule motions much at all, as the head group and chain motions are fairly uncoupled.



Re: CORREL: principal inertial axis reorientation

Joined: Jul 2004
Posts: 165
Forum Member

Forum Member
Joined: Jul 2004
Posts: 165 
thank you rick,
as I plan to monitor rotational motion in different part of the lipid molecule,i.e., head, glycerol backbone and the tails.
Based on what you say, CH vector at a various position of the lipid tails would be more interesting to discuss than CCend of the tail. correct?




