Home Page About CC-ATOMS Proposal ITR Proposal: Prior Research

Proposal Contents: Summary Introduction Need Goals Research Activities

PRIOR RESEARCH: THE MOLECULAR WORKBENCH

REC 9980620, Bob Tinker & Boris Berenfeld Co-PIs. Dates: 01/01/00 - 12/31/02, $1,364,944
The proposed project is a direct outgrowth of the Molecular Workbench project, which is studying the effectiveness of using highly manipulable molecular dynamics models to enhance student learning. The project facilitates student learning through the use of guided inquiry about the connections between microscopic and macroscopic properties of materials. The approach involves the coordination of an atomic-scale dynamics models with interactive Flash applications, and Pedagogica. In collaboration with teachers, instructional materials using atomic-scale modeling are being tested in a variety of middle and high school classrooms (Tinker, 2000a; 2000b; 2001a; 2001c).

For more details about the project than can be provided here, including the software, images and descriptions of activities, scripted hypermodels, and curricula, see: http://workbench.concord.org/download/

The Atomic-scale Engine
The Molecular Workbench2D and Molecular Workbench3D (formerly called Oslet) are molecular simulation engines developed for the project. MW2D is written in pure Java. MW3D is written in Java and Java3D. The models are based on molecular mechanics, an important part of contemporary computational chemistry and biology. The applications solve the equations of motion for ensembles of atoms and molecules, provide a variety of views of the evolving system in real time, and support a highly interactive environment for experimenting with these ensembles. These models represent ways of bringing current research into the classroom. For instance, the Murad-Powles model of osmosis illustrated in Figure 1 is drawn directly from recent scientific literature (Murad & Powles, 1993), as are the meso-scale models described below.

The Molecular Workbench 2D (MW2D)
MW2D develops an object-oriented framework (OOF) for interactive molecular simulations. The core of OOF provides software abstractions of basic objects for atoms, molecules, fields, boundaries, data sets, and the interactions among them. The supporting parts of the OOF architecture involve abstraction of the processes of virtual molecular simulation experiments, the algorithms for the engines, the interfaces through which the user can interact with objects and data sets of a model, and the general model-view-controller rules used in designing and observing a model. To support a high degree of interactivity and flexibility, over 50 types of basic user-object interactions associated with model building activities have been defined.

MW2D consists of four major functional modules: a builder, a viewer, a simulator, and a data analysis environment. The software abstraction model in the OOF provides the interfaces for these modules to interact and integrate.

The builder constructs the components of a model. The viewer offers a graphic device onto which the instantaneous image of a model is rendered. The simulator introduces physical time to a model and is the virtual machine to advance its time evolution. The data analysis environment provides facilities to get access to and analyze, in real time or post-simulation time, the raw data of a model produced in a simulation.

The builder supports a variety of user-object interactions for constructing arbitrary 2D atoms and molecules at the atomic and meso-scale. In the atomic-scale model, the properties of the atoms can be edited graphically, including their mass, size, color, and charge. Atoms interact through a Lennard-Jones potential that has an r-6 attractive part with a strong r-12 repulsive core. The amount of attraction and the size of the core can be determined by moving the minimum of a curve of the Lennard-Jones potential. Molecules consisting of atoms connected by bonds with radial and angular harmonic forces can be easily constructed through a graphical user interface. The user can change the boundary conditions, and customize the external field. To make the builder more user-friendly, many of these interactions are observed by an undo/redo manager and an error manager. The meso-scale model was developed to extend the computations to chemical and biological systems with molecules that are too large to model on an atom-by-atom basis.

The viewer of MW2D is capable of visualizing a model in a variety of different styles. These styles include not only the standard modes such as ball-and-stick, wire frame, vector display, and kinetic energy shading, but also extended mode such as contour plots of electron density and potentials. Beside this, the viewer provides visual interactive supports for the builder, simulator and data analysis environment.

The simulator is a virtual machine for computing the motion of elements in a collection of atoms, molecules, boundaries, fields, and barriers using molecular mechanics (Allen and Tildesley, 1987; Brooks, et al, 1990). Supporting standard molecular simulation protocols such as constant-volume-energy (NVE), constant-volume-temperature (NVT) and constant-pressure-temperature (NPT), the simulator allows the user to study different statistical mechanics ensembles. The simulator also provides five different kinds of boundaries for the user to emulate different situations. For example, reflecting boundary effect can be eliminated by introducing periodic boundary conditions in horizontal, vertical or both directions so that infinite system can be modeled.

We have developed three simulators, each specialized for a different modeling goal.

The Quick Simulator is designed for speed to create a highly interactive learning environment. Electrostatic forces can be added. The simulator assumes that the model is in thermal equilibrium with a heat bath, and imposes kinetic energy conservation by periodic adjustments of the velocities of objects. This simulator can take large time steps and is therefore capable of visualizing the very fast motion of large numbers of objects, creating dramatic dynamic visualizations. The speed appears as fast as other comparable simulations. Phase transitions emerge from ensembles of atoms and can illustrate ideal and non-ideal gas behaviors, condensation, melting, and phase separation.
The Advanced Simulator models molecules made from Lennard-Jones atoms connected by harmonic radial and angular forces, and uses sufficiently small time-steps so that the total energy of the system can be conserved. The user can add or subtract heat energy from this system and observe the temperature change, thus supporting student explorations of energy conservation, heat capacity, and latent heats. The user can easily design and simulate molecular models, such as an off-lattice bead-model of polymer. Although there is a speed penalty for this generality, this simulator can illustrate a very wide range of chemical and biological properties that emerge from these interactions. The Hamiltonian used in the advanced simulator is:

where the first term combines the Lennard-Jones potentials, the second the electrostatic potentials, the third the Axilrod-Teller three-body potentials, the fourth the bond stretching potentials, the fifth the angle bending potentials, and the last are the torsional potentials which show up in dimensions. The above formula does not include the interaction terms with external fields, and the friction and mirror image potentials with boundaries.

The Meso-Scale Simulator has been developed to solve the problem of speed and crossing scale when modeling systems composed of many atoms, such as those commonly encountered in biology. In this model, entire molecules are represented by an elliptical generalization of the Lennard-Jones force field called the Gay-Berne model. The MSS is more efficient, because it ignores a molecule's internal degrees of freedom. Instead, it models the molecules as bodies that interact with each other through the Gay-Berne force field and, optionally, electrostatic fields. MSS can be used to simulate typical liquid crystal properties, including transformation between isotropic, nematic or scemetic phases. Some of these properties are seen in some biological systems, such as cell membranes and DNA molecules at high concentrations. See http://people.concord.org/~qxie/oslet/guide/potentials.html#uaff for the Hamiltonian.
At the current stage of development, the data analysis environment (DAE) is a loose combination of analysis and graph tools that jointly support mining, classifying, refining, analyzing, and visualizing the data produced by a simulation. The DAE supports basic computations on all the simulators, such as computing time and spatial correlation functions, obtaining an x-ray negative, and profiling energy distributions. The DAE can also compute and graph any statistical or running average over an arbitrary subset of the objects in a simulation in real time. For instance, in a model consisting of red atoms and green molecules, the application can compute the average kinetic energy of each species separately. To ensure that the simulation runs as quickly as possible, only the computations needed for a particular simulation are calculated by requests to a task server. The task server of a simulator is a concept in the OOF to mean a controller that dynamically manages computing sub-processes derived from basic simulation processes.

The basic elements of the DAE are property data sets. A property is a function of the information (such as coordinates, velocities and accelerations) of a selected set of atoms. Besides the seed properties­such as the position of each atom­ that the system provides, advanced properties can be generated using a "property factory". Properties can be registered with the task server through a check-in procedure, and then are computed. The results in the data store of the task server can be exported to the graph tools for data visualization and analysis.

MW2D can be deployed on all desktop platforms: Windows, Mac OS, and Linux. There is also an applet version of MW2D that runs on the Web. The software and documentation is at http://people.concord.org/~qxie/beta.html .

Molecular Workbench 3D (MW3D)

MW2D handles only two-dimensional models. Two-dimensional representations are preferred in many learning situations, because 2D visualizations are often more clear. The crystal structure of condensed matter is, for instance, much more obvious in 2D than 3D. However, three dimensions are essential to represent many crystal structures, much of chemistry, and most of biochemistry. It is also important for students to gain 3D visualization experience.

To determine the feasibility of using 3D in atomic-scale educational models, we have built a crystal lab, which allows learners to construct and explore a variety of crystal lattice structures based on the inputs of unit cell information and the assignment of space groups. The central part of the crystal lab is a 3D molecular graphics engine that we developed. It was necessary to create this from scratch in Java so that it could be controlled by Pedagogica and fit into our environment; none of the existing graphics engines fit our requirements. This engine supports standard 3D navigation interactivity such as translation, rotation, zooming and spinning. Many standard 3D graphics options are available such as changing background of the virtual universe and the ambient light color, adjusting directional illumination light, tuning graphics quality, and so on. Standard molecular graphics displaying styles such as wire-frame and ball-and-stick, as well as a polyhedral space-filling style with the option of semi-transparent effect, are provided. Although it might not be precisely correct, the polyhedral space-filling style offers a straightforward way to convert a microscopic discrete lattice to a macroscopic continuum solid. Combined with the molecular graphics engines we will use algorithms such as those used to search the coordination atoms of a selected atom, and an I/O mechanism capable of reading the Protein Data Bank (PDB) format for molecular structures and exporting a scene graph in the Virtual Reality Modeling Language (VRML).

The overall performance of the graphics engine based on Java3D for heavy-weighted rendering is satisfactorily for crystals with fewer than 100 units and also for larger crystals in the light-weighted display mode (such as wire-frame).

Pedagogica and CCClient

Pedagogica, which was developed by Paul Horwitz to control a genetics simulation called BioLogica, has now been interfaced with MW2D. Pedagogica simplifies the conversion of open-ended models and tools into learning activities that support guided inquiry and generate specific, useful assessment data. We call the combination of a model with Pedagogica a hypermodel (Horwitz & Tinker, 2001; Tinker, 2001d). Pedagogica is a client application that controls models and their appearance, displays what options are available, recieves input concerning the state of the model, controls the interactions with the user, and coordinates other resources that might be used with the model. MW2D or other applications can appear in a window in Pedagogica along with other windows that may contain text, buttons, and other Java or Flash applications, as illustrated in Figure 5. We have, for instance, used this to show macroscopic views illustrated with Flash that are coordinated with microscopic views generated by MW2D to model phase changes at the atomic level that are mirrored with an illustration that changes between gas, liquid, and solid.
We have elected to execute hypermodels on client computers. This is because the models require considerable computational resources and because good connectivity is not always accessible to students, either due to travel, poor home access, or unreliable school networks. We made the decision to support Flash in Pedagogica because it simplifies authoring animations. We use animations extensively to illustrate the macroscopic representation of a system in coordination with a microscopic representation generated by MW2D. Executing Flash within Pedagogica required creating many of the basic graphics capacities used by Flash in Java.

Pedagogica is controlled by a script that can be written either in EASL, our homegrown Educational Application Scripting Language, or JavaScript. Both are object-oriented. Some graphical programming aides have been developed to simplify screen layout and sequencing pages. Pedagogica monitors events generated by the model and can use these to collect implicit assessment data about student progress through the activity as well as explicit assessment data from student answers to questions. These data can be time-stamped and given XML tags. Student activities written for Pedagogica can be easily changed to fit our research and, as we learn more, to fit the needs of different students..

To simplify the delivery and maintenance of client-side software, we have developed a thin client called CCClient that can download, update, cache, and run all the components required including Pedagogica, scripts, and MW2D, as well as returning the assessment data with XML tags. This permits the entire environment to be implemented remotely on systems with intermittent connectivity.

Findings
Eight curriculum activities based on rudimentary atomic-scale models have undergone formative testing in a public middle school 8th grade class. These approaches include integrated hands-on lab, computer modeling, and kinesthetic modeling activities (Tinker, et al, 2001). Each activity was focused on a single fundamental concept.

Students were highly engaged, responsive in discussions, and made connections between the various types of activities conducted. For example, they performed a computer simulation and a kinesthetic simulation of kinetic energy exchange between gaseous atoms during collisions-in other words, heat transfer through direct mixing of atoms. Their written responses regarding these activities indicated that they found both the computer and kinesthetic experiences helpful in understanding the concept and that doing both of them was better than doing one or the other. Only two students felt that doing both activities did not add anything to understanding gained from computer simulation alone.

In addition to student reactions, we obtained feedback regarding the software interface, including how to simplify it to aid in activity understanding and execution, and how often students would find a similar kind of activity engaging. We have determined that there is a balance between a didactic and an open-ended approach that allows students to be most successful. Students' interest in similar activities could be extended if they were allowed to do things outside of what was originally intended in the lesson. The computer interaction seems to work best if students are lead through an initial "tutorial" stage and are then given open-ended challenges to solve. In responses to final assessment questions given to students, there were strong indications that in only eight class periods spread over one month, students were beginning to use the atomic-scale, particulate theory of matter to help explain the macro-scale world they experience with their senses. We found these limited initial results to be very encouraging and feel that exposure to a broader range of activities, accompanied by the full research evaluation, will extend the scope of macro-scale phenomena that students are able to comprehend at the atomic-scale level. Classroom research with completed versions of the software is underway.

This material is based upon work supported by the National Science Foundation under Grant No. EIA-0219345. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.