Adk44: Created page with "The AMBER 12 interface is based on ''pmemd'' rather than ''sander''. There are many limitations on ''pmemd'', predominantly that it lacks: * Analytical second derivatives; *..."

2019-05-13T09:23:48Z

Created page with "The AMBER 12 interface is based on ''pmemd'' rather than ''sander''. There are many limitations on ''pmemd'', predominantly that it lacks: * Analytical second derivatives; *..."

New page

The AMBER 12 interface is based on ''pmemd'' rather than ''sander''.

There are many limitations on ''pmemd'', predominantly that it lacks:

* Analytical second derivatives;
* Smooth cutoffs;
* Non-periodic vacuum simulations.

However, it has built-in support for CUDA acceleration and its code is much, much simpler than the ''sander'' code.

== Changes from pmemd code ==

=== amber_interface.F90 ===

Based upon pmemd.F90, with the following changes:

* Split the code into 3 subroutines: <code>amber_setup</code>, <code>energy_and_gradient</code>, <code>amber_finish</code>

=== smooth_cutoff.F90 ===

The module is named <code>smooth_cutoff_mod</code> and keeps track of <code>ifswitch</code>.

Implementing force switching cut-offs. Important variables to keep track of:

<code>ifswitch</code> determines which style of cut-offs is used.

==== ifswitch = 1 ====

<code>rc = sqrt(cut)</code>, i.e. <math>\text{cut} = r_c^2</math>

In generalised Born simulations, <code>cut</code> is a private variable. The publicly accessible variable is <code>gb_cutoff</code>. Note that this is different from <code>rgbmax</code>, which determines the cut-off for calculating gb radii.

We switch on the additional terms at <math>r_\text{on}</math>. This is arbitrarily set to 0.9 of the cut-off distance.

<math>r_\text{on}^2 = (0.9r_c)^2</math>

We have these other terms, <code>term1(1:natom), term2(1:natom), term3(1:natom)</code>, which I shall represent as <math>c_1, c_2, c_3</math> respectively.

<math>r^2 = (x_i-x_j)^2 + (y_i-y_j)^2 + (z_i-z_j)^2</math>

<math>
c_1 = \frac{c_3}{c_2} =
\begin{cases}
0.0, & \mbox{if }r \le r_\text{on} \\
\frac{-12(r_\text{on}^2-r^2)}{(r_c^2-r^2)(r_c^2+2r^2-3r_\text{on}^2)}, & \mbox{if }r > r_\text{on}
\end{cases}
</math>

<math>
c_2 =
\begin{cases}
1.0, & \mbox{if }r \le r_\text{on} \\
(r_c^2-r^2)^2(r_c^2+2r^2-3r_\text{on}^2)\frac{1}{r_c^6}\frac{1}{(1-0.9^2)^3}, & \mbox{if }r > r_\text{on}
\end{cases}</math>

<math>
c_3 =
\begin{cases}
0.0, & \mbox{if }r \le r_\text{on} \\
-12(r_c^2-r^2)(r_\text{on}^2-r^2)\frac{1}{r_c^6}\frac{1}{(1-0.9^2)^3}, & \mbox{if }r > r_\text{on}
\end{cases}</math>

The energies and their derivatives are given by:

<math>E_\text{elec} = \frac{q_iq_j}{\epsilon_\text{int}r}c_2 \mbox{ and } \frac{-1}{r}\frac{dE_\text{elec}}{dr} = \frac{E_\text{elec}}{r^2} + E_\text{elec}c_1</math>

<math>E_\text{gb} = E_\text{gb}c_2 \mbox{ and } \frac{dE_\text{gb}}{dr} = \frac{-1}{r}\frac{dE_\text{gb}}{dr}c_2 + E_\text{gb}c_1</math>

<math>E_\text{vdW} = \left ( \frac{A}{r^{12}} - \frac{B}{r^6} \right ) c_2 \mbox{ and } \frac{-1}{r}\frac{dE_\text{vdW}}{dr} = \left (\frac{12A}{r^{12}} - \frac{6B}{r^6} \right )\frac{1}{r^2}c_2 + \left ( \frac{A}{r^{12}} - \frac{B}{r^6} \right ) c_3</math>

=== mdin_ctrl_dat.F90 ===

* Line 31: <code>ifswitch</code> added to declaration
* Line 45: <code>ifswitch</code> added to <code>/ mdin_ctrl_int /</code> common block
* Line 49: <code>mdin_ctrl_int_cnt = 56</code> (rather than 55, since there is one more integer variable now)
* Line 216: <code>ifswitch</code> added <code>/cntrl/</code> namelist, for reading min.in files.
* Line 434: initialised <code>ifswitch = 0</code>

Notes on AMBER 12 interface - Revision history

Adk44: Created page with "The AMBER 12 interface is based on ''pmemd'' rather than ''sander''. There are many limitations on ''pmemd'', predominantly that it lacks: * Analytical second derivatives; *..."