150 |
|
increasing time step. For each time step, the dotted line is total |
151 |
|
energy using the DLM integrator, and the solid line comes from the |
152 |
|
quaternion integrator. The larger time step plots are shifted up |
153 |
< |
from the true energy baseline for clarity.} \label{timestep} |
153 |
> |
from the true energy baseline for clarity.} |
154 |
> |
\label{methodFig:timestep} |
155 |
|
\end{figure} |
156 |
|
|
157 |
< |
In Fig.~\ref{timestep}, the resulting energy drift at various time |
158 |
< |
steps for both the DLM and quaternion integration schemes is |
159 |
< |
compared. All of the 1000 molecule water simulations started with |
160 |
< |
the same configuration, and the only difference was the method for |
161 |
< |
handling rotational motion. At time steps of 0.1 and 0.5 fs, both |
162 |
< |
methods for propagating molecule rotation conserve energy fairly |
163 |
< |
well, with the quaternion method showing a slight energy drift over |
164 |
< |
time in the 0.5 fs time step simulation. At time steps of 1 and 2 |
165 |
< |
fs, the energy conservation benefits of the DLM method are clearly |
166 |
< |
demonstrated. Thus, while maintaining the same degree of energy |
167 |
< |
conservation, one can take considerably longer time steps, leading |
168 |
< |
to an overall reduction in computation time. |
157 |
> |
In Fig.~\ref{methodFig:timestep}, the resulting energy drift at |
158 |
> |
various time steps for both the DLM and quaternion integration |
159 |
> |
schemes is compared. All of the 1000 molecule water simulations |
160 |
> |
started with the same configuration, and the only difference was the |
161 |
> |
method for handling rotational motion. At time steps of 0.1 and 0.5 |
162 |
> |
fs, both methods for propagating molecule rotation conserve energy |
163 |
> |
fairly well, with the quaternion method showing a slight energy |
164 |
> |
drift over time in the 0.5 fs time step simulation. At time steps of |
165 |
> |
1 and 2 fs, the energy conservation benefits of the DLM method are |
166 |
> |
clearly demonstrated. Thus, while maintaining the same degree of |
167 |
> |
energy conservation, one can take considerably longer time steps, |
168 |
> |
leading to an overall reduction in computation time. |
169 |
|
|
170 |
|
\subsection{\label{methodSection:NVT}Nos\'{e}-Hoover Thermostatting} |
171 |
|
|
172 |
< |
The Nos\'e-Hoover equations of motion are given by\cite{Hoover85} |
172 |
> |
The Nos\'e-Hoover equations of motion are given by\cite{Hoover1985} |
173 |
|
\begin{eqnarray} |
174 |
|
\dot{{\bf r}} & = & {\bf v}, \\ |
175 |
|
\dot{{\bf v}} & = & \frac{{\bf f}}{m} - \chi {\bf v} ,\\ |
285 |
|
|
286 |
|
The Nos\'e-Hoover algorithm is known to conserve a Hamiltonian for |
287 |
|
the extended system that is, to within a constant, identical to the |
288 |
< |
Helmholtz free energy,\cite{melchionna93} |
288 |
> |
Helmholtz free energy,\cite{Melchionna1993} |
289 |
|
\begin{equation} |
290 |
|
H_{\mathrm{NVT}} = V + K + f k_B T_{\mathrm{target}} \left( |
291 |
|
\frac{\tau_{T}^2 \chi^2(t)}{2} + \int_{0}^{t} \chi(t^\prime) |
296 |
|
last column of the {\tt .stat} file to allow checks on the quality |
297 |
|
of the integration. |
298 |
|
|
298 |
– |
Bond constraints are applied at the end of both the {\tt moveA} and |
299 |
– |
{\tt moveB} portions of the algorithm. Details on the constraint |
300 |
– |
algorithms are given in section \ref{oopseSec:rattle}. |
301 |
– |
|
299 |
|
\subsection{\label{methodSection:NPTi}Constant-pressure integration with |
300 |
|
isotropic box deformations (NPTi)} |
301 |
|
|
302 |
|
To carry out isobaric-isothermal ensemble calculations {\sc oopse} |
303 |
|
implements the Melchionna modifications to the |
304 |
< |
Nos\'e-Hoover-Andersen equations of motion,\cite{melchionna93} |
304 |
> |
Nos\'e-Hoover-Andersen equations of motion,\cite{Melchionna1993} |
305 |
|
|
306 |
|
\begin{eqnarray} |
307 |
|
\dot{{\bf r}} & = & {\bf v} + \eta \left( {\bf r} - {\bf R}_0 \right), \\ |
620 |
|
standard NPT ensemble with a different pressure control strategy |
621 |
|
|
622 |
|
\begin{equation} |
623 |
< |
\.{\overleftrightarrow{{\eta _{\alpha \beta}}}}=\left\{\begin{array}{ll} |
623 |
> |
\dot {\overleftrightarrow{\eta}} _{\alpha \beta}=\left\{\begin{array}{ll} |
624 |
|
\frac{{V(P_{\alpha \beta } - P_{{\rm{target}}} )}}{{\tau_{\rm{B}}^{\rm{2}} fk_B T_{{\rm{target}}} }} |
625 |
< |
& \mbox{if \[ \alpha = \beta = z)$}\\ |
625 |
> |
& \mbox{if $ \alpha = \beta = z$}\\ |
626 |
|
0 & \mbox{otherwise}\\ |
627 |
|
\end{array} |
628 |
|
\right. |
646 |
|
pressure. The equation of motion for cell size control tensor, |
647 |
|
$\eta$, in $NP\gamma T$ is |
648 |
|
\begin{equation} |
649 |
< |
\.{\overleftrightarrow{{\eta _{\alpha \beta}}}}=\left\{\begin{array}{ll} |
649 |
> |
\dot {\overleftrightarrow{\eta}} _{\alpha \beta}=\left\{\begin{array}{ll} |
650 |
|
- A_{xy} (\gamma _\alpha - \gamma _{{\rm{target}}} ) & \mbox{$\alpha = \beta = x$ or $\alpha = \beta = y$}\\ |
651 |
|
\frac{{V(P_{\alpha \beta } - P_{{\rm{target}}} )}}{{\tau _{\rm{B}}^{\rm{2}} fk_B T_{{\rm{target}}}}} & \mbox{$\alpha = \beta = z$} \\ |
652 |
|
0 & \mbox{$\alpha \ne \beta$} \\ |
653 |
+ |
\end{array} |
654 |
|
\right. |
655 |
|
\end{equation} |
656 |
|
where $ \gamma _{{\rm{target}}}$ is the external surface tension and |
657 |
|
the instantaneous surface tensor $\gamma _\alpha$ is given by |
658 |
|
\begin{equation} |
659 |
< |
\gamma _\alpha = - h_z |
660 |
< |
(\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\leftrightarrow$}} |
663 |
< |
\over P} _{\alpha \alpha } - P_{{\rm{target}}} ) |
659 |
> |
\gamma _\alpha = - h_z( \overleftrightarrow{P} _{\alpha \alpha } |
660 |
> |
- P_{{\rm{target}}} ) |
661 |
|
\label{methodEquation:instantaneousSurfaceTensor} |
662 |
|
\end{equation} |
663 |
|
|
669 |
|
integrator is a special case of $NP\gamma T$ if the surface tension |
670 |
|
$\gamma$ is set to zero. |
671 |
|
|
672 |
< |
%\section{\label{methodSection:constraintMethod}Constraint Method} |
672 |
> |
\section{\label{methodSection:zcons}Z-Constraint Method} |
673 |
|
|
674 |
< |
%\subsection{\label{methodSection:bondConstraint}Bond Constraint for Rigid Body} |
674 |
> |
Based on the fluctuation-dissipation theorem, a force |
675 |
> |
auto-correlation method was developed by Roux and Karplus to |
676 |
> |
investigate the dynamics of ions inside ion channels\cite{Roux1991}. |
677 |
> |
The time-dependent friction coefficient can be calculated from the |
678 |
> |
deviation of the instantaneous force from its mean force. |
679 |
> |
\begin{equation} |
680 |
> |
\xi(z,t)=\langle\delta F(z,t)\delta F(z,0)\rangle/k_{B}T, |
681 |
> |
\end{equation} |
682 |
> |
where% |
683 |
> |
\begin{equation} |
684 |
> |
\delta F(z,t)=F(z,t)-\langle F(z,t)\rangle. |
685 |
> |
\end{equation} |
686 |
|
|
687 |
< |
%\subsection{\label{methodSection:zcons}Z-constraint Method} |
687 |
> |
If the time-dependent friction decays rapidly, the static friction |
688 |
> |
coefficient can be approximated by |
689 |
> |
\begin{equation} |
690 |
> |
\xi_{\text{static}}(z)=\int_{0}^{\infty}\langle\delta F(z,t)\delta |
691 |
> |
F(z,0)\rangle dt. |
692 |
> |
\end{equation} |
693 |
> |
Allowing diffusion constant to then be calculated through the |
694 |
> |
Einstein relation:\cite{Marrink1994} |
695 |
> |
\begin{equation} |
696 |
> |
D(z)=\frac{k_{B}T}{\xi_{\text{static}}(z)}=\frac{(k_{B}T)^{2}}{\int_{0}^{\infty |
697 |
> |
}\langle\delta F(z,t)\delta F(z,0)\rangle dt}.% |
698 |
> |
\end{equation} |
699 |
> |
|
700 |
> |
The Z-Constraint method, which fixes the z coordinates of the |
701 |
> |
molecules with respect to the center of the mass of the system, has |
702 |
> |
been a method suggested to obtain the forces required for the force |
703 |
> |
auto-correlation calculation.\cite{Marrink1994} However, simply |
704 |
> |
resetting the coordinate will move the center of the mass of the |
705 |
> |
whole system. To avoid this problem, we reset the forces of |
706 |
> |
z-constrained molecules as well as subtract the total constraint |
707 |
> |
forces from the rest of the system after the force calculation at |
708 |
> |
each time step instead of resetting the coordinate. |
709 |
|
|
710 |
+ |
After the force calculation, define $G_\alpha$ as |
711 |
+ |
\begin{equation} |
712 |
+ |
G_{\alpha} = \sum_i F_{\alpha i}, \label{oopseEq:zc1} |
713 |
+ |
\end{equation} |
714 |
+ |
where $F_{\alpha i}$ is the force in the z direction of atom $i$ in |
715 |
+ |
z-constrained molecule $\alpha$. The forces of the z constrained |
716 |
+ |
molecule are then set to: |
717 |
+ |
\begin{equation} |
718 |
+ |
F_{\alpha i} = F_{\alpha i} - |
719 |
+ |
\frac{m_{\alpha i} G_{\alpha}}{\sum_i m_{\alpha i}}. |
720 |
+ |
\end{equation} |
721 |
+ |
Here, $m_{\alpha i}$ is the mass of atom $i$ in the z-constrained |
722 |
+ |
molecule. Having rescaled the forces, the velocities must also be |
723 |
+ |
rescaled to subtract out any center of mass velocity in the z |
724 |
+ |
direction. |
725 |
+ |
\begin{equation} |
726 |
+ |
v_{\alpha i} = v_{\alpha i} - |
727 |
+ |
\frac{\sum_i m_{\alpha i} v_{\alpha i}}{\sum_i m_{\alpha i}}, |
728 |
+ |
\end{equation} |
729 |
+ |
where $v_{\alpha i}$ is the velocity of atom $i$ in the z direction. |
730 |
+ |
Lastly, all of the accumulated z constrained forces must be |
731 |
+ |
subtracted from the system to keep the system center of mass from |
732 |
+ |
drifting. |
733 |
+ |
\begin{equation} |
734 |
+ |
F_{\beta i} = F_{\beta i} - \frac{m_{\beta i} \sum_{\alpha} |
735 |
+ |
G_{\alpha}} |
736 |
+ |
{\sum_{\beta}\sum_i m_{\beta i}}, |
737 |
+ |
\end{equation} |
738 |
+ |
where $\beta$ are all of the unconstrained molecules in the system. |
739 |
+ |
Similarly, the velocities of the unconstrained molecules must also |
740 |
+ |
be scaled. |
741 |
+ |
\begin{equation} |
742 |
+ |
v_{\beta i} = v_{\beta i} + \sum_{\alpha} |
743 |
+ |
\frac{\sum_i m_{\alpha i} v_{\alpha i}}{\sum_i m_{\alpha i}}. |
744 |
+ |
\end{equation} |
745 |
+ |
|
746 |
+ |
At the very beginning of the simulation, the molecules may not be at |
747 |
+ |
their constrained positions. To move a z-constrained molecule to its |
748 |
+ |
specified position, a simple harmonic potential is used |
749 |
+ |
\begin{equation} |
750 |
+ |
U(t)=\frac{1}{2}k_{\text{Harmonic}}(z(t)-z_{\text{cons}})^{2},% |
751 |
+ |
\end{equation} |
752 |
+ |
where $k_{\text{Harmonic}}$ is the harmonic force constant, $z(t)$ |
753 |
+ |
is the current $z$ coordinate of the center of mass of the |
754 |
+ |
constrained molecule, and $z_{\text{cons}}$ is the constrained |
755 |
+ |
position. The harmonic force operating on the z-constrained molecule |
756 |
+ |
at time $t$ can be calculated by |
757 |
+ |
\begin{equation} |
758 |
+ |
F_{z_{\text{Harmonic}}}(t)=-\frac{\partial U(t)}{\partial z(t)}= |
759 |
+ |
-k_{\text{Harmonic}}(z(t)-z_{\text{cons}}). |
760 |
+ |
\end{equation} |
761 |
+ |
|
762 |
+ |
|
763 |
|
\section{\label{methodSection:langevin}Integrators for Langevin Dynamics of Rigid Bodies} |
764 |
|
|
765 |
< |
\subsection{\label{methodSection:temperature}Temperature Control} |
765 |
> |
%\subsection{\label{methodSection:temperature}Temperature Control} |
766 |
|
|
767 |
< |
\subsection{\label{methodSection:pressureControl}Pressure Control} |
767 |
> |
%\subsection{\label{methodSection:pressureControl}Pressure Control} |
768 |
|
|
769 |
< |
\section{\label{methodSection:hydrodynamics}Hydrodynamics} |
769 |
> |
%\section{\label{methodSection:hydrodynamics}Hydrodynamics} |
770 |
|
|
771 |
< |
%\section{\label{methodSection:coarseGrained}Coarse-Grained Modeling} |
771 |
> |
%applications of langevin dynamics |
772 |
> |
As an excellent alternative to newtonian dynamics, Langevin |
773 |
> |
dynamics, which mimics a simple heat bath with stochastic and |
774 |
> |
dissipative forces, has been applied in a variety of studies. The |
775 |
> |
stochastic treatment of the solvent enables us to carry out |
776 |
> |
substantially longer time simulation. Implicit solvent Langevin |
777 |
> |
dynamics simulation of met-enkephalin not only outperforms explicit |
778 |
> |
solvent simulation on computation efficiency, but also agrees very |
779 |
> |
well with explicit solvent simulation on dynamics |
780 |
> |
properties\cite{Shen2002}. Recently, applying Langevin dynamics with |
781 |
> |
UNRES model, Liow and his coworkers suggest that protein folding |
782 |
> |
pathways can be possibly exploited within a reasonable amount of |
783 |
> |
time\cite{Liwo2005}. The stochastic nature of the Langevin dynamics |
784 |
> |
also enhances the sampling of the system and increases the |
785 |
> |
probability of crossing energy barrier\cite{Banerjee2004, Cui2003}. |
786 |
> |
Combining Langevin dynamics with Kramers's theory, Klimov and |
787 |
> |
Thirumalai identified the free-energy barrier by studying the |
788 |
> |
viscosity dependence of the protein folding rates\cite{Klimov1997}. |
789 |
> |
In order to account for solvent induced interactions missing from |
790 |
> |
implicit solvent model, Kaya incorporated desolvation free energy |
791 |
> |
barrier into implicit coarse-grained solvent model in protein |
792 |
> |
folding/unfolding study and discovered a higher free energy barrier |
793 |
> |
between the native and denatured states. Because of its stability |
794 |
> |
against noise, Langevin dynamics is very suitable for studying |
795 |
> |
remagnetization processes in various |
796 |
> |
systems\cite{Garcia-Palacios1998,Berkov2002,Denisov2003}. For |
797 |
> |
instance, the oscillation power spectrum of nanoparticles from |
798 |
> |
Langevin dynamics simulation has the same peak frequencies for |
799 |
> |
different wave vectors,which recovers the property of magnetic |
800 |
> |
excitations in small finite structures\cite{Berkov2005a}. In an |
801 |
> |
attempt to reduce the computational cost of simulation, multiple |
802 |
> |
time stepping (MTS) methods have been introduced and have been of |
803 |
> |
great interest to macromolecule and protein |
804 |
> |
community\cite{Tuckerman1992}. Relying on the observation that |
805 |
> |
forces between distant atoms generally demonstrate slower |
806 |
> |
fluctuations than forces between close atoms, MTS method are |
807 |
> |
generally implemented by evaluating the slowly fluctuating forces |
808 |
> |
less frequently than the fast ones. Unfortunately, nonlinear |
809 |
> |
instability resulting from increasing timestep in MTS simulation |
810 |
> |
have became a critical obstruction preventing the long time |
811 |
> |
simulation. Due to the coupling to the heat bath, Langevin dynamics |
812 |
> |
has been shown to be able to damp out the resonance artifact more |
813 |
> |
efficiently\cite{Sandu1999}. |
814 |
|
|
815 |
< |
%\section{\label{methodSection:moleculeScale}Molecular-Scale Modeling} |
815 |
> |
%review rigid body dynamics |
816 |
> |
Rigid bodies are frequently involved in the modeling of different |
817 |
> |
areas, from engineering, physics, to chemistry. For example, |
818 |
> |
missiles and vehicle are usually modeled by rigid bodies. The |
819 |
> |
movement of the objects in 3D gaming engine or other physics |
820 |
> |
simulator is governed by the rigid body dynamics. In molecular |
821 |
> |
simulation, rigid body is used to simplify the model in |
822 |
> |
protein-protein docking study\cite{Gray2003}. |
823 |
> |
|
824 |
> |
It is very important to develop stable and efficient methods to |
825 |
> |
integrate the equations of motion of orientational degrees of |
826 |
> |
freedom. Euler angles are the nature choice to describe the |
827 |
> |
rotational degrees of freedom. However, due to its singularity, the |
828 |
> |
numerical integration of corresponding equations of motion is very |
829 |
> |
inefficient and inaccurate. Although an alternative integrator using |
830 |
> |
different sets of Euler angles can overcome this |
831 |
> |
difficulty\cite{Ryckaert1977, Andersen1983}, the computational |
832 |
> |
penalty and the lost of angular momentum conservation still remain. |
833 |
> |
In 1977, a singularity free representation utilizing quaternions was |
834 |
> |
developed by Evans\cite{Evans1977}. Unfortunately, this approach |
835 |
> |
suffer from the nonseparable Hamiltonian resulted from quaternion |
836 |
> |
representation, which prevents the symplectic algorithm to be |
837 |
> |
utilized. Another different approach is to apply holonomic |
838 |
> |
constraints to the atoms belonging to the rigid |
839 |
> |
body\cite{Barojas1973}. Each atom moves independently under the |
840 |
> |
normal forces deriving from potential energy and constraint forces |
841 |
> |
which are used to guarantee the rigidness. However, due to their |
842 |
> |
iterative nature, SHAKE and Rattle algorithm converge very slowly |
843 |
> |
when the number of constraint increases. |
844 |
> |
|
845 |
> |
The break through in geometric literature suggests that, in order to |
846 |
> |
develop a long-term integration scheme, one should preserve the |
847 |
> |
geometric structure of the flow. Matubayasi and Nakahara developed a |
848 |
> |
time-reversible integrator for rigid bodies in quaternion |
849 |
> |
representation. Although it is not symplectic, this integrator still |
850 |
> |
demonstrates a better long-time energy conservation than traditional |
851 |
> |
methods because of the time-reversible nature. Extending |
852 |
> |
Trotter-Suzuki to general system with a flat phase space, Miller and |
853 |
> |
his colleagues devised an novel symplectic, time-reversible and |
854 |
> |
volume-preserving integrator in quaternion representation. However, |
855 |
> |
all of the integrators in quaternion representation suffer from the |
856 |
> |
computational penalty of constructing a rotation matrix from |
857 |
> |
quaternions to evolve coordinates and velocities at every time step. |
858 |
> |
An alternative integration scheme utilizing rotation matrix directly |
859 |
> |
is RSHAKE\cite{Kol1997}, in which a conjugate momentum to rotation |
860 |
> |
matrix is introduced to re-formulate the Hamiltonian's equation and |
861 |
> |
the Hamiltonian is evolved in a constraint manifold by iteratively |
862 |
> |
satisfying the orthogonality constraint. However, RSHAKE is |
863 |
> |
inefficient because of the iterative procedure. An extremely |
864 |
> |
efficient integration scheme in rotation matrix representation, |
865 |
> |
which also preserves the same structural properties of the |
866 |
> |
Hamiltonian flow as Miller's integrator, is proposed by Dullweber, |
867 |
> |
Leimkuhler and McLachlan (DLM)\cite{Dullweber1997}. |
868 |
> |
|
869 |
> |
%review langevin/browninan dynamics for arbitrarily shaped rigid body |
870 |
> |
Combining Langevin or Brownian dynamics with rigid body dynamics, |
871 |
> |
one can study the slow processes in biomolecular systems. Modeling |
872 |
> |
the DNA as a chain of rigid spheres beads, which subject to harmonic |
873 |
> |
potentials as well as excluded volume potentials, Mielke and his |
874 |
> |
coworkers discover rapid superhelical stress generations from the |
875 |
> |
stochastic simulation of twin supercoiling DNA with response to |
876 |
> |
induced torques\cite{Mielke2004}. Membrane fusion is another key |
877 |
> |
biological process which controls a variety of physiological |
878 |
> |
functions, such as release of neurotransmitters \textit{etc}. A |
879 |
> |
typical fusion event happens on the time scale of millisecond, which |
880 |
> |
is impracticable to study using all atomistic model with newtonian |
881 |
> |
mechanics. With the help of coarse-grained rigid body model and |
882 |
> |
stochastic dynamics, the fusion pathways were exploited by many |
883 |
> |
researchers\cite{Noguchi2001,Noguchi2002,Shillcock2005}. Due to the |
884 |
> |
difficulty of numerical integration of anisotropy rotation, most of |
885 |
> |
the rigid body models are simply modeled by sphere, cylinder, |
886 |
> |
ellipsoid or other regular shapes in stochastic simulations. In an |
887 |
> |
effort to account for the diffusion anisotropy of the arbitrary |
888 |
> |
particles, Fernandes and de la Torre improved the original Brownian |
889 |
> |
dynamics simulation algorithm\cite{Ermak1978,Allison1991} by |
890 |
> |
incorporating a generalized $6\times6$ diffusion tensor and |
891 |
> |
introducing a simple rotation evolution scheme consisting of three |
892 |
> |
consecutive rotations\cite{Fernandes2002}. Unfortunately, unexpected |
893 |
> |
error and bias are introduced into the system due to the arbitrary |
894 |
> |
order of applying the noncommuting rotation |
895 |
> |
operators\cite{Beard2003}. Based on the observation the momentum |
896 |
> |
relaxation time is much less than the time step, one may ignore the |
897 |
> |
inertia in Brownian dynamics. However, assumption of the zero |
898 |
> |
average acceleration is not always true for cooperative motion which |
899 |
> |
is common in protein motion. An inertial Brownian dynamics (IBD) was |
900 |
> |
proposed to address this issue by adding an inertial correction |
901 |
> |
term\cite{Beard2001}. As a complement to IBD which has a lower bound |
902 |
> |
in time step because of the inertial relaxation time, long-time-step |
903 |
> |
inertial dynamics (LTID) can be used to investigate the inertial |
904 |
> |
behavior of the polymer segments in low friction |
905 |
> |
regime\cite{Beard2001}. LTID can also deal with the rotational |
906 |
> |
dynamics for nonskew bodies without translation-rotation coupling by |
907 |
> |
separating the translation and rotation motion and taking advantage |
908 |
> |
of the analytical solution of hydrodynamics properties. However, |
909 |
> |
typical nonskew bodies like cylinder and ellipsoid are inadequate to |
910 |
> |
represent most complex macromolecule assemblies. These intricate |
911 |
> |
molecules have been represented by a set of beads and their |
912 |
> |
hydrodynamics properties can be calculated using variant |
913 |
> |
hydrodynamic interaction tensors. |
914 |
> |
|
915 |
> |
The goal of the present work is to develop a Langevin dynamics |
916 |
> |
algorithm for arbitrary rigid particles by integrating the accurate |
917 |
> |
estimation of friction tensor from hydrodynamics theory into the |
918 |
> |
sophisticated rigid body dynamics. |
919 |
> |
|
920 |
> |
|
921 |
> |
\subsection{Friction Tensor} |
922 |
> |
|
923 |
> |
For an arbitrary rigid body moves in a fluid, it may experience |
924 |
> |
friction force $f_r$ or friction torque $\tau _r$ along the opposite |
925 |
> |
direction of the velocity $v$ or angular velocity $\omega$ at |
926 |
> |
arbitrary origin $P$, |
927 |
> |
\begin{equation} |
928 |
> |
\left( \begin{array}{l} |
929 |
> |
f_r \\ |
930 |
> |
\tau _r \\ |
931 |
> |
\end{array} \right) = - \left( {\begin{array}{*{20}c} |
932 |
> |
{\Xi _{P,t} } & {\Xi _{P,c}^T } \\ |
933 |
> |
{\Xi _{P,c} } & {\Xi _{P,r} } \\ |
934 |
> |
\end{array}} \right)\left( \begin{array}{l} |
935 |
> |
\nu \\ |
936 |
> |
\omega \\ |
937 |
> |
\end{array} \right) |
938 |
> |
\end{equation} |
939 |
> |
where $\Xi _{P,t}t$ is the translation friction tensor, $\Xi _{P,r}$ |
940 |
> |
is the rotational friction tensor and $\Xi _{P,c}$ is the |
941 |
> |
translation-rotation coupling tensor. The procedure of calculating |
942 |
> |
friction tensor using hydrodynamic tensor and comparison between |
943 |
> |
bead model and shell model were elaborated by Carrasco \textit{et |
944 |
> |
al}\cite{Carrasco1999}. An important property of the friction tensor |
945 |
> |
is that the translational friction tensor is independent of origin |
946 |
> |
while the rotational and coupling are sensitive to the choice of the |
947 |
> |
origin \cite{Brenner1967}, which can be described by |
948 |
> |
\begin{equation} |
949 |
> |
\begin{array}{c} |
950 |
> |
\Xi _{P,t} = \Xi _{O,t} = \Xi _t \\ |
951 |
> |
\Xi _{P,c} = \Xi _{O,c} - r_{OP} \times \Xi _t \\ |
952 |
> |
\Xi _{P,r} = \Xi _{O,r} - r_{OP} \times \Xi _t \times r_{OP} + \Xi _{O,c} \times r_{OP} - r_{OP} \times \Xi _{O,c}^T \\ |
953 |
> |
\end{array} |
954 |
> |
\end{equation} |
955 |
> |
Where $O$ is another origin and $r_{OP}$ is the vector joining $O$ |
956 |
> |
and $P$. It is also worthy of mention that both of translational and |
957 |
> |
rotational frictional tensors are always symmetric. In contrast, |
958 |
> |
coupling tensor is only symmetric at center of reaction: |
959 |
> |
\begin{equation} |
960 |
> |
\Xi _{R,c} = \Xi _{R,c}^T |
961 |
> |
\end{equation} |
962 |
> |
The proper location for applying friction force is the center of |
963 |
> |
reaction, at which the trace of rotational resistance tensor reaches |
964 |
> |
minimum. |
965 |
> |
|
966 |
> |
\subsection{Rigid body dynamics} |
967 |
> |
|
968 |
> |
The Hamiltonian of rigid body can be separated in terms of potential |
969 |
> |
energy $V(r,A)$ and kinetic energy $T(p,\pi)$, |
970 |
> |
\[ |
971 |
> |
H = V(r,A) + T(v,\pi ) |
972 |
> |
\] |
973 |
> |
A second-order symplectic method is now obtained by the composition |
974 |
> |
of the flow maps, |
975 |
> |
\[ |
976 |
> |
\varphi _{\Delta t} = \varphi _{\Delta t/2,V} \circ \varphi |
977 |
> |
_{\Delta t,T} \circ \varphi _{\Delta t/2,V}. |
978 |
> |
\] |
979 |
> |
Moreover, $\varphi _{\Delta t/2,V}$ can be divided into two |
980 |
> |
sub-flows which corresponding to force and torque respectively, |
981 |
> |
\[ |
982 |
> |
\varphi _{\Delta t/2,V} = \varphi _{\Delta t/2,F} \circ \varphi |
983 |
> |
_{\Delta t/2,\tau }. |
984 |
> |
\] |
985 |
> |
Since the associated operators of $\varphi _{\Delta t/2,F} $ and |
986 |
> |
$\circ \varphi _{\Delta t/2,\tau }$ are commuted, the composition |
987 |
> |
order inside $\varphi _{\Delta t/2,V}$ does not matter. |
988 |
> |
|
989 |
> |
Furthermore, kinetic potential can be separated to translational |
990 |
> |
kinetic term, $T^t (p)$, and rotational kinetic term, $T^r (\pi )$, |
991 |
> |
\begin{equation} |
992 |
> |
T(p,\pi ) =T^t (p) + T^r (\pi ). |
993 |
> |
\end{equation} |
994 |
> |
where $ T^t (p) = \frac{1}{2}v^T m v $ and $T^r (\pi )$ is defined |
995 |
> |
by \ref{introEquation:rotationalKineticRB}. Therefore, the |
996 |
> |
corresponding flow maps are given by |
997 |
> |
\[ |
998 |
> |
\varphi _{\Delta t,T} = \varphi _{\Delta t,T^t } \circ \varphi |
999 |
> |
_{\Delta t,T^r }. |
1000 |
> |
\] |
1001 |
> |
The free rigid body is an example of Lie-Poisson system with |
1002 |
> |
Hamiltonian function |
1003 |
> |
\begin{equation} |
1004 |
> |
T^r (\pi ) = T_1 ^r (\pi _1 ) + T_2^r (\pi _2 ) + T_3^r (\pi _3 ) |
1005 |
> |
\label{introEquation:rotationalKineticRB} |
1006 |
> |
\end{equation} |
1007 |
> |
where $T_i^r (\pi _i ) = \frac{{\pi _i ^2 }}{{2I_i }}$ and |
1008 |
> |
Lie-Poisson structure matrix, |
1009 |
> |
\begin{equation} |
1010 |
> |
J(\pi ) = \left( {\begin{array}{*{20}c} |
1011 |
> |
0 & {\pi _3 } & { - \pi _2 } \\ |
1012 |
> |
{ - \pi _3 } & 0 & {\pi _1 } \\ |
1013 |
> |
{\pi _2 } & { - \pi _1 } & 0 \\ |
1014 |
> |
\end{array}} \right) |
1015 |
> |
\end{equation} |
1016 |
> |
Thus, the dynamics of free rigid body is governed by |
1017 |
> |
\begin{equation} |
1018 |
> |
\frac{d}{{dt}}\pi = J(\pi )\nabla _\pi T^r (\pi ) |
1019 |
> |
\end{equation} |
1020 |
> |
One may notice that each $T_i^r$ in Equation |
1021 |
> |
\ref{introEquation:rotationalKineticRB} can be solved exactly. For |
1022 |
> |
instance, the equations of motion due to $T_1^r$ are given by |
1023 |
> |
\begin{equation} |
1024 |
> |
\frac{d}{{dt}}\pi = R_1 \pi ,\frac{d}{{dt}}A = AR_1 |
1025 |
> |
\label{introEqaution:RBMotionSingleTerm} |
1026 |
> |
\end{equation} |
1027 |
> |
where |
1028 |
> |
\[ R_1 = \left( {\begin{array}{*{20}c} |
1029 |
> |
0 & 0 & 0 \\ |
1030 |
> |
0 & 0 & {\pi _1 } \\ |
1031 |
> |
0 & { - \pi _1 } & 0 \\ |
1032 |
> |
\end{array}} \right). |
1033 |
> |
\] |
1034 |
> |
The solutions of Equation \ref{introEqaution:RBMotionSingleTerm} is |
1035 |
> |
\[ |
1036 |
> |
\pi (\Delta t) = e^{\Delta tR_1 } \pi (0),A(\Delta t) = |
1037 |
> |
A(0)e^{\Delta tR_1 } |
1038 |
> |
\] |
1039 |
> |
with |
1040 |
> |
\[ |
1041 |
> |
e^{\Delta tR_1 } = \left( {\begin{array}{*{20}c} |
1042 |
> |
0 & 0 & 0 \\ |
1043 |
> |
0 & {\cos \theta _1 } & {\sin \theta _1 } \\ |
1044 |
> |
0 & { - \sin \theta _1 } & {\cos \theta _1 } \\ |
1045 |
> |
\end{array}} \right),\theta _1 = \frac{{\pi _1 }}{{I_1 }}\Delta t. |
1046 |
> |
\] |
1047 |
> |
To reduce the cost of computing expensive functions in $e^{\Delta |
1048 |
> |
tR_1 }$, we can use Cayley transformation, |
1049 |
> |
\[ |
1050 |
> |
e^{\Delta tR_1 } \approx (1 - \Delta tR_1 )^{ - 1} (1 + \Delta tR_1 |
1051 |
> |
) |
1052 |
> |
\] |
1053 |
> |
The flow maps for $T_2^r$ and $T_3^r$ can be found in the same |
1054 |
> |
manner. |
1055 |
> |
|
1056 |
> |
In order to construct a second-order symplectic method, we split the |
1057 |
> |
angular kinetic Hamiltonian function into five terms |
1058 |
> |
\[ |
1059 |
> |
T^r (\pi ) = \frac{1}{2}T_1 ^r (\pi _1 ) + \frac{1}{2}T_2^r (\pi _2 |
1060 |
> |
) + T_3^r (\pi _3 ) + \frac{1}{2}T_2^r (\pi _2 ) + \frac{1}{2}T_1 ^r |
1061 |
> |
(\pi _1 ) |
1062 |
> |
\]. |
1063 |
> |
Concatenating flows corresponding to these five terms, we can obtain |
1064 |
> |
the flow map for free rigid body, |
1065 |
> |
\[ |
1066 |
> |
\varphi _{\Delta t,T^r } = \varphi _{\Delta t/2,\pi _1 } \circ |
1067 |
> |
\varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t,\pi _3 } |
1068 |
> |
\circ \varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t/2,\pi |
1069 |
> |
_1 }. |
1070 |
> |
\] |
1071 |
> |
|
1072 |
> |
The equations of motion corresponding to potential energy and |
1073 |
> |
kinetic energy are listed in the below table, |
1074 |
> |
\begin{center} |
1075 |
> |
\begin{tabular}{|l|l|} |
1076 |
> |
\hline |
1077 |
> |
% after \\: \hline or \cline{col1-col2} \cline{col3-col4} ... |
1078 |
> |
Potential & Kinetic \\ |
1079 |
> |
$\frac{{dq}}{{dt}} = \frac{p}{m}$ & $\frac{d}{{dt}}q = p$ \\ |
1080 |
> |
$\frac{d}{{dt}}p = - \frac{{\partial V}}{{\partial q}}$ & $ \frac{d}{{dt}}p = 0$ \\ |
1081 |
> |
$\frac{d}{{dt}}Q = 0$ & $ \frac{d}{{dt}}Q = Qskew(I^{ - 1} j)$ \\ |
1082 |
> |
$ \frac{d}{{dt}}\pi = \sum\limits_i {\left( {Q^T F_i (r,Q)} \right) \times X_i }$ & $\frac{d}{{dt}}\pi = \pi \times I^{ - 1} \pi$\\ |
1083 |
> |
\hline |
1084 |
> |
\end{tabular} |
1085 |
> |
\end{center} |
1086 |
> |
|
1087 |
> |
Finally, we obtain the overall symplectic flow maps for free moving |
1088 |
> |
rigid body |
1089 |
> |
\begin{align*} |
1090 |
> |
\varphi _{\Delta t} = &\varphi _{\Delta t/2,F} \circ \varphi _{\Delta t/2,\tau } \circ \\ |
1091 |
> |
&\varphi _{\Delta t,T^t } \circ \varphi _{\Delta t/2,\pi _1 } \circ \varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t,\pi _3 } \circ \varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t/2,\pi _1 } \circ \\ |
1092 |
> |
&\varphi _{\Delta t/2,\tau } \circ \varphi _{\Delta t/2,F} .\\ |
1093 |
> |
\label{introEquation:overallRBFlowMaps} |
1094 |
> |
\end{align*} |
1095 |
> |
|
1096 |
> |
\subsection{Langevin dynamics for rigid particles of arbitrary shape} |
1097 |
> |
|
1098 |
> |
Consider a Langevin equation of motions in generalized coordinates |
1099 |
> |
\begin{equation} |
1100 |
> |
M_i \dot V_i (t) = F_{s,i} (t) + F_{f,i(t)} + F_{r,i} (t) |
1101 |
> |
\label{LDGeneralizedForm} |
1102 |
> |
\end{equation} |
1103 |
> |
where $M_i$ is a $6\times6$ generalized diagonal mass (include mass |
1104 |
> |
and moment of inertial) matrix and $V_i$ is a generalized velocity, |
1105 |
> |
$V_i = V_i(v_i,\omega _i)$. The right side of Eq. |
1106 |
> |
(\ref{LDGeneralizedForm}) consists of three generalized forces in |
1107 |
> |
lab-fixed frame, systematic force $F_{s,i}$, dissipative force |
1108 |
> |
$F_{f,i}$ and stochastic force $F_{r,i}$. While the evolution of the |
1109 |
> |
system in Newtownian mechanics typically refers to lab-fixed frame, |
1110 |
> |
it is also convenient to handle the rotation of rigid body in |
1111 |
> |
body-fixed frame. Thus the friction and random forces are calculated |
1112 |
> |
in body-fixed frame and converted back to lab-fixed frame by: |
1113 |
> |
\[ |
1114 |
> |
\begin{array}{l} |
1115 |
> |
F_{f,i}^l (t) = A^T F_{f,i}^b (t), \\ |
1116 |
> |
F_{r,i}^l (t) = A^T F_{r,i}^b (t) \\ |
1117 |
> |
\end{array}. |
1118 |
> |
\] |
1119 |
> |
Here, the body-fixed friction force $F_{r,i}^b$ is proportional to |
1120 |
> |
the body-fixed velocity at center of resistance $v_{R,i}^b$ and |
1121 |
> |
angular velocity $\omega _i$, |
1122 |
> |
\begin{equation} |
1123 |
> |
F_{r,i}^b (t) = \left( \begin{array}{l} |
1124 |
> |
f_{r,i}^b (t) \\ |
1125 |
> |
\tau _{r,i}^b (t) \\ |
1126 |
> |
\end{array} \right) = - \left( {\begin{array}{*{20}c} |
1127 |
> |
{\Xi _{R,t} } & {\Xi _{R,c}^T } \\ |
1128 |
> |
{\Xi _{R,c} } & {\Xi _{R,r} } \\ |
1129 |
> |
\end{array}} \right)\left( \begin{array}{l} |
1130 |
> |
v_{R,i}^b (t) \\ |
1131 |
> |
\omega _i (t) \\ |
1132 |
> |
\end{array} \right), |
1133 |
> |
\end{equation} |
1134 |
> |
while the random force $F_{r,i}^l$ is a Gaussian stochastic variable |
1135 |
> |
with zero mean and variance |
1136 |
> |
\begin{equation} |
1137 |
> |
\left\langle {F_{r,i}^l (t)(F_{r,i}^l (t'))^T } \right\rangle = |
1138 |
> |
\left\langle {F_{r,i}^b (t)(F_{r,i}^b (t'))^T } \right\rangle = |
1139 |
> |
2k_B T\Xi _R \delta (t - t'). |
1140 |
> |
\end{equation} |
1141 |
> |
The equation of motion for $v_i$ can be written as |
1142 |
> |
\begin{equation} |
1143 |
> |
m\dot v_i (t) = f_{t,i} (t) = f_{s,i} (t) + f_{f,i}^l (t) + |
1144 |
> |
f_{r,i}^l (t) |
1145 |
> |
\end{equation} |
1146 |
> |
Since the frictional force is applied at the center of resistance |
1147 |
> |
which generally does not coincide with the center of mass, an extra |
1148 |
> |
torque is exerted at the center of mass. Thus, the net body-fixed |
1149 |
> |
frictional torque at the center of mass, $\tau _{n,i}^b (t)$, is |
1150 |
> |
given by |
1151 |
> |
\begin{equation} |
1152 |
> |
\tau _{r,i}^b = \tau _{r,i}^b +r_{MR} \times f_{r,i}^b |
1153 |
> |
\end{equation} |
1154 |
> |
where $r_{MR}$ is the vector from the center of mass to the center |
1155 |
> |
of the resistance. Instead of integrating angular velocity in |
1156 |
> |
lab-fixed frame, we consider the equation of motion of angular |
1157 |
> |
momentum in body-fixed frame |
1158 |
> |
\begin{equation} |
1159 |
> |
\dot \pi _i (t) = \tau _{t,i} (t) = \tau _{s,i} (t) + \tau _{f,i}^b |
1160 |
> |
(t) + \tau _{r,i}^b(t) |
1161 |
> |
\end{equation} |
1162 |
> |
|
1163 |
> |
Embedding the friction terms into force and torque, one can |
1164 |
> |
integrate the langevin equations of motion for rigid body of |
1165 |
> |
arbitrary shape in a velocity-Verlet style 2-part algorithm, where |
1166 |
> |
$h= \delta t$: |
1167 |
> |
|
1168 |
> |
{\tt part one:} |
1169 |
> |
\begin{align*} |
1170 |
> |
v_i (t + h/2) &\leftarrow v_i (t) + \frac{{hf_{t,i}^l (t)}}{{2m_i }} \\ |
1171 |
> |
\pi _i (t + h/2) &\leftarrow \pi _i (t) + \frac{{h\tau _{t,i}^b (t)}}{2} \\ |
1172 |
> |
r_i (t + h) &\leftarrow r_i (t) + hv_i (t + h/2) \\ |
1173 |
> |
A_i (t + h) &\leftarrow rotate\left( {h\pi _i (t + h/2)I^{ - 1} } \right) \\ |
1174 |
> |
\end{align*} |
1175 |
> |
In this context, the $\mathrm{rotate}$ function is the reversible |
1176 |
> |
product of five consecutive body-fixed rotations, |
1177 |
> |
\begin{equation} |
1178 |
> |
\mathrm{rotate}({\bf a}) = \mathsf{G}_x(a_x / 2) \cdot |
1179 |
> |
\mathsf{G}_y(a_y / 2) \cdot \mathsf{G}_z(a_z) \cdot \mathsf{G}_y(a_y |
1180 |
> |
/ 2) \cdot \mathsf{G}_x(a_x /2), |
1181 |
> |
\end{equation} |
1182 |
> |
where each rotational propagator, $\mathsf{G}_\alpha(\theta)$, |
1183 |
> |
rotates both the rotation matrix ($\mathsf{A}$) and the body-fixed |
1184 |
> |
angular momentum ($\pi$) by an angle $\theta$ around body-fixed axis |
1185 |
> |
$\alpha$, |
1186 |
> |
\begin{equation} |
1187 |
> |
\mathsf{G}_\alpha( \theta ) = \left\{ |
1188 |
> |
\begin{array}{lcl} |
1189 |
> |
\mathsf{A}(t) & \leftarrow & \mathsf{A}(0) \cdot \mathsf{R}_\alpha(\theta)^T, \\ |
1190 |
> |
{\bf j}(t) & \leftarrow & \mathsf{R}_\alpha(\theta) \cdot {\bf |
1191 |
> |
j}(0). |
1192 |
> |
\end{array} |
1193 |
> |
\right. |
1194 |
> |
\end{equation} |
1195 |
> |
$\mathsf{R}_\alpha$ is a quadratic approximation to the single-axis |
1196 |
> |
rotation matrix. For example, in the small-angle limit, the |
1197 |
> |
rotation matrix around the body-fixed x-axis can be approximated as |
1198 |
> |
\begin{equation} |
1199 |
> |
\mathsf{R}_x(\theta) \approx \left( |
1200 |
> |
\begin{array}{ccc} |
1201 |
> |
1 & 0 & 0 \\ |
1202 |
> |
0 & \frac{1-\theta^2 / 4}{1 + \theta^2 / 4} & -\frac{\theta}{1+ |
1203 |
> |
\theta^2 / 4} \\ |
1204 |
> |
0 & \frac{\theta}{1+ \theta^2 / 4} & \frac{1-\theta^2 / 4}{1 + |
1205 |
> |
\theta^2 / 4} |
1206 |
> |
\end{array} |
1207 |
> |
\right). |
1208 |
> |
\end{equation} |
1209 |
> |
All other rotations follow in a straightforward manner. |
1210 |
> |
|
1211 |
> |
After the first part of the propagation, the friction and random |
1212 |
> |
forces are generated at the center of resistance in body-fixed frame |
1213 |
> |
and converted back into lab-fixed frame |
1214 |
> |
\[ |
1215 |
> |
f_{t,i}^l (t + h) = - \left( {\frac{{\partial V}}{{\partial r_i }}} |
1216 |
> |
\right)_{r_i (t + h)} + A_i^T (t + h)[F_{f,i}^b (t + h) + F_{r,i}^b |
1217 |
> |
(t + h)], |
1218 |
> |
\] |
1219 |
> |
while the system torque in lab-fixed frame is transformed into |
1220 |
> |
body-fixed frame, |
1221 |
> |
\[ |
1222 |
> |
\tau _{t,i}^b (t + h) = A\tau _{s,i}^l (t) + \tau _{n,i}^b (t) + |
1223 |
> |
\tau _{r,i}^b (t). |
1224 |
> |
\] |
1225 |
> |
Once the forces and torques have been obtained at the new time step, |
1226 |
> |
the velocities can be advanced to the same time value. |
1227 |
> |
|
1228 |
> |
{\tt part two:} |
1229 |
> |
\begin{align*} |
1230 |
> |
v_i (t) &\leftarrow v_i (t + h/2) + \frac{{hf_{t,i}^l (t + h)}}{{2m_i }} \\ |
1231 |
> |
\pi _i (t) &\leftarrow \pi _i (t + h/2) + \frac{{h\tau _{t,i}^b (t + h)}}{2} \\ |
1232 |
> |
\end{align*} |
1233 |
> |
|
1234 |
> |
\subsection{Results and discussion} |