Simulation with computers

Figure 15.12. Droplet shapes as simulated with computational fluid dynamics during XME with a cross-flow velocity of 1 m/s. Numbers in the figures are the times elapsed after the start of the droplet formation in gs. After Kelder et al. (2006).

Simulation can explore innovative solutions difficult to investigate experimentally. For example, the integration of process simulation with Computer Fluid Dynamics can replace costly prototypes. The flowsheeting of virtual plants involving rigorous hydraulic modelling of units is very likely in the next years. 

Computational methods used to extend the time scale of atomically detailed simulations have improved in the last 15 years. Accurate MTS simulations, with computational gains up to a factor of 10, have extended the applicability of molecular dynamics simulations and refinements in the computation of medivun-range forces could provide stable results with increasing speedups. It is apparent that stability limitations will prevent the extension of these algorithms to the range of time scales that are needed to study many processes of interest, however. On the other hand, reaction path approaches can be used... 

At 700°C and 1 atm this leads to a diffusion constant of 0.81 cm /s The flow field around a superheater tube is very complex involving both laminar and turbulent boundary layers and the estimation of the local boundary layer thicknesses (velocity, diffusion and thermal boundary layers) around the tube requires computer simulations with computational fluid dynamic (CFD) software packages. However, for this rough analysis an average value of the thermal boundary layer thickness around the tube is enough and can be estimated if the average Nusselt number around the tube is known... 

It has been shown before that a number of plant situations can only be simulated with computer codes. 

The themiodynamic properties calculated by different routes are different, since the MS solution is an approximation. The osmotic coefficient from the virial pressure, compressibility and energy equations are not the same. Of these, the energy equation is the most accurate by comparison with computer simulations of Card and Valleau [ ]. The osmotic coefficients from the virial and compressibility equations are... 

Despite all the shortcomings listed above, full particle classical MD can be considered mature [84]. Even when all shortcomings will be overcome, we can now clearly delineate the limits for application. These are mainly in the size of the system and the length of the possible simulation. With the rapidly growing cheap computer memory shear size by itself is hardly a limitation several tens of thousands of particles can be handled routinely (for example, we report a simulation of a porin trimer protein embedded in a phospholipid membrane in aqueous environment with almost 70,000 particles [85] see also the contribution of K. Schulten in this symposium) and a million particles could be handled should that be desired. 

In principle, mesoscale methods can provide a means for connecting one type of simulation to another. For example, a molecular simulation can be used to describe a lipid. One can then derive the parameters for a lipid-lipid potential. These parameters can then be used in a simulation that combines lipids to form a membrane, which, in turn, can be used to compute parameters describing a membrane as a flexible sheet. Such parameters could be used for a simulation with many cells in order to obtain parameters that describe an organ, which could be used for a whole-body biological simulation. Each step, in theory, could be modeled in a different way using parameters derived not from experiment but from a more low-level form of simulation. This situation has not yet been realized, but it is representative of one trend in computational technique development. 

These calculations can incorporate various types of constraints. It is most common to run simulations with a hxed number of atoms and a hxed volume. In this case, the temperature can be computed from the average kinetic energy of the atoms. It is also possible to adjust the volume to maintain a constant pressure or to scale the velocities to maintain a constant temperature. 

It is also possible to simulate nonequilibrium systems. For example, a bulk liquid can be simulated with periodic boundary conditions that have shifting boundaries. This results in simulating a flowing liquid with laminar flow. This makes it possible to compute properties not measurable in a static fluid, such as the viscosity. Nonequilibrium simulations give rise to additional technical difficulties. Readers of this book are advised to leave nonequilibrium simulations to researchers specializing in this type of work. 

Force field calculations often truncate the non bonded potential energy of a molecular system at some finite distance. Truncation (nonbonded cutoff) saves computing resources. Also, periodic boxes and boundary conditions require it. However, this approximation is too crude for some calculations. For example, a molecular dynamic simulation with an abruptly truncated potential produces anomalous and nonphysical behavior. One symptom is that the solute (for example, a protein) cools and the solvent (water) heats rapidly. The temperatures of system components then slowly converge until the system appears to be in equilibrium, but it is not. 

Computational methods have played an exceedingly important role in understanding the fundamental aspects of shock compression and in solving complex shock-wave problems. Major advances in the numerical algorithms used for solving dynamic problems, coupled with the tremendous increase in computational capabilities, have made many problems tractable that only a few years ago could not have been solved. It is now possible to perform two-dimensional molecular dynamics simulations with a high degree of accuracy, and three-dimensional problems can also be solved with moderate accuracy. 

It should also be noted that a force field for a wide variety of small molecules, CHARMm (note the small m, indicating the commercial version of the program and parameters), is available [39] and has been applied to protein simulations with limited success. Efforts are currently under way to extend the CHARMm small molecule force field to make the nonbonded parameters consistent with those of the CHARMM force fields, thereby allowing for a variety of small molecules to be included in computational smdies of biological systems. 

Simple considerations show that the membrane potential cannot be treated with computer simulations, and continuum electrostatic methods may constimte the only practical approach to address such questions. The capacitance of a typical lipid membrane is on the order of 1 j.F/cm-, which corresponds to a thickness of approximately 25 A and a dielectric constant of 2 for the hydrophobic core of a bilayer. In the presence of a membrane potential the bulk solution remains electrically neutral and a small charge imbalance is distributed in the neighborhood of the interfaces. The membrane potential arises from... 

ME Davis, JD Madura, BA Luty, JA McCammon. Electrostatics and diffusion of molecules m solution Simulations with the University of Houston Brownian dynamics program. Comput Phys Commun 62 187-197, 1991. 

Figure 3 Comparison of the densities (in g/cm ) of model compounds for membrane lipids computed from constant-pressure MD simulations with the coiTespondmg experimental values. The model compounds include solid octane and tricosane, liquid butane, octane, tetradecane, and eico-sane, and the glycerylphosphorylcholme, cyclopentylphosphorylcholme monohydrate, dilauroly-glycerol, anhydrous cholesterol, cholesterol monohydrate, and cholesterol acetate crystals. (Models from Refs. 18, 42, and 43).

A simple, time-honoured illustration of the operation of the Monte Carlo approach is one curious way of estimating the constant n. Imagine a circle inscribed inside a square of side a, and use a table of random numbers to determine the cartesian coordinates of many points constrained to lie anywhere at random within the square. The ratio of the number of points that lies inside the circle to the total number of points within the square na l4a = nl4. The more random points have been put in place, the more accurate will be the value thus obtained. Of course, such a procedure would make no sense, since n can be obtained to any desired accuracy by the summation of a mathematical series... i.e., analytically. But once the simulator is faced with a eomplex series of particle movements, analytical methods quickly become impracticable and simulation, with time steps included, is literally the only possible approach. That is how computer simulation began. 

Another detailed method of determining pressures is computational fluid dynamics (CFD), which uses a numerical solution of simplified equations of motion over a dense grid of points around the building. Murakami et al. and Zhoy and Stathopoulos found less agreement with computational fluid dynamics methods using the k-e turbulence model typically used in current commercial codes. More advanced turbulence models such as large eddy simulation were more successful but much more costly. ... 

Time available It takes some time to set up a simulation case. After a basic configuration has run successfully, it is easy and quick to do additional computer simulations with different boundary conditions. Experiments, on the other hand, are generally time-consuming. 

Evans, L.B., 1989. Simulation with respect to solid-fluid systems. Computers and Chemical Engineering, 13, 343. 

As is seen in Fig. 2(b), the results of Eqs. (33) and (34) are in fair agreement with computer simulation results. Carnahan and Starling (CS) [18] have made the observation that the result... 

In Fig. 3, the PY g(r) and y(r) are compared with computer simulation results and with a fit [19-21] of the computer simulation results. The agreement is quite good except that the PY result is too small at contact. This... 

A comparison of the predictions for the cavity pair correlation function of some first-principles theories with computer simulation results [72] at a fairly high density has been presented by Stell (see Fig. 2 in Ref. 69). The... 

In Sec. 3 our presentation is focused on the most important results obtained by different authors in the framework of the rephca Ornstein-Zernike (ROZ) integral equations and by simulations of simple fluids in microporous matrices. For illustrative purposes, we discuss some original results obtained recently in our laboratory. Those allow us to show the application of the ROZ equations to the structure and thermodynamics of fluids adsorbed in disordered porous media. In particular, we present a solution of the ROZ equations for a hard sphere mixture that is highly asymmetric by size, adsorbed in a matrix of hard spheres. This example is relevant in describing the structure of colloidal dispersions in a disordered microporous medium. On the other hand, we present some of the results for the adsorption of a hard sphere fluid in a disordered medium of spherical permeable membranes. The theory developed for the description of this model agrees well with computer simulation data. Finally, in this section we demonstrate the applications of the ROZ theory and present simulation data for adsorption of a hard sphere fluid in a matrix of short chain molecules. This example serves to show the relevance of the theory of Wertheim to chemical association for a set of problems focused on adsorption of fluids and mixtures in disordered microporous matrices prepared by polymerization of species. 

---