A realistic open-cycle gas-core nuclear rocket simulation model must be capable of a self-consistent nozzle calculation in conjunction with coupled radiation and neutron transport in three spatial dimensions. As part of the development effort for such a model, five hydrodynamic codes were used to compare with a converging-diverging nozzle experiment. The codes used in the comparison are CHAD, FLUENT, KIVA2, RAMPANT, and VNAP2. Solution accuracy as a function of mesh size is important because, in the near term, a practical three-dimensional simulation model will require rather coarse zoning across the nozzle throat. In the study, four different grids were considered. (1) coarse, radially uniform grid, (2) coarse, radially nonuniform grid, (3) fine, radially uniform grid, and (4) fine, radially nonuniform grid. The study involves code verification, not prediction. In other words, the authors know the solution they want to match, so they can change methods and/or modify an algorithm to best match this class of problem. In this context, it was necessary to use the higher-order methods in both FLUENT and RAMPANT. In addition, KIVA2 required a modification that allows significantly more accurate solutions for a converging-diverging nozzle. From a predictive point of view, code accuracy with no tuning is an important result. The most accurate codes on a coarse grid, CHAD and VNAP2, did not require any tuning. Their main comparison among the codes was the radial dependence of the Mach number across the nozzle throat. All five codes yielded a very similar solution with fine, radially uniform and radially nonuniform grids. However, the codes yielded significantly different solutions with coarse, radially uniform and radially nonuniform grids. For all the codes, radially nonuniform zoning across the throat significantly increased solution accuracy with a coarse mesh. None of the codes agrees in detail with the weak shock located downstream of the nozzle throat, but all the codes indicated the presence of a weak downstream shock.

The next giant leap for mankind will be the human exploration of Mars. Almost certainly within the next thirty years, a human crew will brave the isolation, the radiation, and the lack of gravity to walk on and explore the Red planet. However, because the mission distances and duration will be hundreds of times greater than the lunar missions, a human crew will face much greater obstacles and a higher risk than those experienced during the Apollo program. A single solution to many of these obstacles is to dramatically decrease the mission duration by developing a high performance propulsion system. The gas-core nuclear rocket (GCNR) has the potential to be such a system, We have completed a comparative study of the potential impact that a GCNR could have on a manned Mars mission, The total IMLEO, transit times, and accumulated radiation dose to the crew will be compared with the NASA Design Reference Missions.

This is the final report of a three-year, Laboratory Directed Research and Development (LDRD) project at Los Alamos National Laboratory. We have developed a prototype plug-and-play (PCUBED) environment based upon a C++ class called a fragment. A fragment is a universal object that can represent any data type. Fragments provide an excellent intuitive approach to the development of an efficient architecture, as well as providing a common data implementation within and between codes. As a result, the PCUBED environment allows for the generation of many different codes within a common framework. At this time, there are seven major codes implemented within the PCUBED environment. Input, output, restart, setup, and graphics are programmed using a high-level approach to insure human efficiency. In contrast, computationally intensive algorithms are programmed using a lowlevel approach to insure computational efficiency. Fragments provide a straightforward approach to switch between high-level and low-level programming. PCUBED has been tested on a Macintosh PowerPC; on IBM, SUN, HP, and SGI workstations; and on the CRAY YMP and Cray T3D. Using this environment, we have incorporated a drift diffusion, energy balance, hydrodynamic, and Monte Carlo model for metal-oxide semiconductor field-effect transistors (MOSFETs) into a single architecture. With all the models in a common framework, we have investigated the noise characteristics of hybrid and delta-f models. Although hybrid and delta-f models appear viable in one dimension, the noise level of higher order transport coefficients in two and three dimensions makes the utility of such combined methods questionable. Background and Research Objectives One of the major problems with the development of new algorithms is the inability to rapidly test and evaluate against existing algorithms in an unbiased fashion. Primarily, this problem arises because existing algorithms are buried within large codes that are not amenable to controlled testing because materials specification, setup, and execution of a specific algorithm is not independent of the rest of the code. Moreover, it is often very difficult to incorporate new algorithms into existing codes because the existing data written in different languages, such as C, C++, f90, or f77. As a result, it is difficult to determine if differences are due to the algorithm, the implementation, or the language. Principal Investigator, email: les @lanl.gov

Realistic interplanetary space exploration depends critically upon the development of a high-speciŽ cimpulse propulsion system. Previous studies indicate that the speciŽ c impulse of an open-cycle gas-core nuclear rocket (OCGCNR) might approach 3000 s. Although the OCGCNR is deceptively simple in concept, it will be difŽ cult to develop in practice because the core is a uranium plasma that must be nearly totally conŽ ned. Before constructing a more comprehensive model for this engine, there is a requirement to understand the limits of present full-scale simulation models and recent scaled experiments. In this scoping study we have used a two-dimensional, axisymmetric, Ž nite difference code to investigate the formation and stability of a recirculation region observed in a scaled experiment. It has been proposed that such a recirculation region, or vortex, might provide improved conŽ nement of the uranium fuel. Our simulation results indicate that a more comprehensive model must treat the rocket nozzle in a selfconsistent fashion to properly calculate the conŽ nement of the uranium plasma. Under conditions that lead to vortex formation, the position of the vortex depends upon the inlet geometry and injection velocity, the nozzle position and subsonic convergence angle, the base-bleed injection rate, and turbulence. With a large base-bleed injection rate, a vortex forms but is then swept away through the nozzle, a result that resolves an inconsistency between a full-scale engine simulation model and recent scaled experiments.

The next giant leap for mankind will be the human exploration of Mars. Almost certainly within the next thirty years, a human crew will brave the isolation, the radiation, and the lack of gravity to walk on and explore the Red planet. However, because the mission distances and duration will be hundreds of times greater than the lunar missions, a human crew will face much greater obstacles and a higher risk than those experienced during the Apollo program. A single solution to many of these obstacles is to dramatically decrease the mission duration by developing a high performance propulsion system. The gas core nuclear rocket (GCNR) has the potential to be such a system. The gas core concept relies on the use of fluid dynamic forces to create and maintain a vortex. The vortex is composed of a fissile material which will achieve criticality and produce high power levels. By radiatively coupling to the surrounding fluids, extremely high temperatures in the propellant and, thus, high specific impulses can be generated. The ship velocities enabled by such performance may allow a 9 month round trip, manned Mars mission to be considered. Alternatively, one might consider slightly longer missions in ships that are heavily shielded against the intense Galactic Cosmic Ray flux to further reduce the radiation dose to the crew. The current status of the research program at the Los Alamos National Laboratory into the gas core nuclear rocket feasibility will be discussed.

Two models have been developed to simulate a vertical-cavity surface emitting laser. The first model is a 2D time-dependent bulk dielectric and absorption coefficients. These bulk coefficients depend upon the material, lattice temperature, and carrier concentration. This field model is coupled with a frequency-dependent gain model that describes the quantum well regions in the time domain. Treatment of frequency-dependent media in a finite-difference time-domain code is computationally intensive. On the other hand, because the volume of the active region is small relative to the volume of the distributed laser cavity, the computational overhead is reasonable. A semi-empirical transport model is used to describe the bult transport, which drives the quantum well transport. In addition, the semi-empirical model provides a spatial distribution for the lattice temperature and carrier concentrations. The second model is a 3D solution of Maxwell's equations. The 3D model can be used for cold cavity calculations. The 2D code generates the dielectric and absorption coefficients assuming azimuthal symmetry, providing the initial conditions for the 3D calculation.

To simulate vertical cavity surface emitting lasers (VCSELs), the authors are developing a three-dimensional, time-dependent field-gain model with absorption in bulk dielectric regions and gain in quantum well regions. Since the laser linewidth is narrow, the bulk absorption coefficient is assumed to be independent of frequency with a value determined by the material and the lattice temperature. In contrast, the frequency-dependent gain regions must be solved consistently in the time domain. Treatment of frequency-dependent media in a finite-difference time-domain code is computationally intensive. However, because the volume of the quantum well regions is small relative to the volume of the multilayer dielectric (MLD) mirror regions, the computational overhead is reasonable. A key issue is the calculation of the fields in the MLD mirror regions. Although computationally intensive, good agreement has been obtained between simulation results and matrix equation solutions for the reflection coefficient, transmission coefficient, and bandwidth of MLD mirrors. The authors discuss the development and testing of the two-dimensional field-gain model. This field-gain model will be integrated with a carrier transport model to form the self-consistent laser code, VCSEL.

To simulate vertical cavity surface emitting lasers (VCSELs), we are developing a 3D, time-dependent field-gain model with absorption in bulk dielectric regions and gain in quantum-well regions. Since the laser linewidth is narrow, the bulk absorption coefficient is assumed to be independent of frequency with a value determined by the material and the lattice temperature. In contrast, the frequency-dependent gain regions must be solved consistently in the time domain. Treatment of frequency-dependent media in a finite-difference time-domain code is computationally intensive. However, because the volume of the quantum-well regions is small relative to the volume of the multilayer dielectric (MLD) mirror regions, the computational overhead is reasonable. A key issue is the calculation of the fields in the MLD mirror regions. Although computationally intensive, good agreement has been obtained between simulation results and matrix equation solutions for the reflection coefficient, transmission coefficient, and bandwidth of MLD mirrors. We discuss the development and testing of the 2D field-gain model. This field- gain model will be integrated with a carrier transport model to form the self-consistent laser code, VCSEL.

Unlike a conventional microwave tube, a virtual-cathode device operates above the space-charge limit where the depth of the space-charge potential can cause electron reflection. The region associated with this electron reflection is referred to as a virtual cathode. Microwaves can be generated through oscillations in the position of the virtual cathode and through the bunching of electrons trapped in a potential well between the real and virtual cathodes. These two mechanisms are competitive. There are three basic classes of virtual cathode devices: (1) reflex triode; (2) reditron and side-shoot vircator; and (3) reflex diode or vircator. The reflex diode is the highest power virtual-cathode device. For the reflex diode the energy exchange between the beam and electromagnetic wave occurs in both the axial and radial directions. In some designs the oscillating-virtual-cathode frequency exceeds the reflexing-electron frequency exceeds the oscillating-virtual-cathode frequency. For the flex diode a periodic disruption in magnetic insulation can modulate the high- frequency microwave power. Overall, particle-in-cell simulation predictions and axial reflex diode experiments are in good agreement. Although frequency stability and phase locking of the reflex diode have been demonstrated, little progress has been made in efficiency enhancement. 58 refs., 11 figs.

Results from two 4 MV, 100 kA coaxial magnetically insulated transmission line virtual cathode oscillator experiments are presented. In both experiments, two distinct microwave pulses with vastly different frequencies were generated during the beam current pulse. The first, lower frequency pulse was found to be produced by an electron reflexing process, while the second, higher frequency pulse was due to the virtual cathode oscillation mechanism.

This paper reviews the basic characteristics of free-electron laser (FEL) oscillators and describes a group of codes that have been developed to analyze and design rf-linac-driven FELs. The optical performance of an FEL has been treated using 1-D time-dependent (finite pulse) codes that describe the characteristics of the optical temporal pulse shape and spectrum during the evolution of the oscillator from low-intensity small-signal gain conditions to high-intensity large-signal gain conditions. These codes can include frequency-dependent elements, such as narrow-band filters or grating rhombs. Diffraction effects, transverse optical and electron-beam mode properties, misalignments, as well as aberrations on optical elements are modeled with the 3-D code FELEX. This code has now been extended to include the effects of imperfections in the wiggler magnetic field and light emission at higher optical harmonics. A separate accelerator modeling capability allows the use of a numerically-generated electron pulse in FELEX for 3-D integrated numerical FEL simulations.

We have combined several technologies in a rather conservative fashion to perform feasibility experiments to power a vacuum, electron beam diode with an explosive power supply. The magnetic flux compression generator used in these experiments was the 13.2-cm-wide by 52.8-cm-long plate generator. The output of this device was switched into an air-core, foil-wound, step-up transformer with a turn ratio of 30:1. The transformer and diode were connected through a 9-m-long piece of Sieverts high-voltage cable. The diode structure consisted of a high-voltage terminal on a stacked-ring grading structure with a 5-cm-diameter stalk to support a 12.7-cm-diameter field emission cathode. The cathode-anode gap was 2.5 cm, and the anode was 6.35-micron--thick aluminized mylar. Currents of greater than or equal to14 kA and voltages of greater than or equal to540 kV were achieved with a pulse length of greater than or equal to100 ns. Theoretical predictions agree with experimental performance until the diode shorted. However, this event limited the ultimate performance of the experiments. We consider diode shorting to be a major problem to be dealt with in future experiments involving >1 ..mu..s power pulses. This approach to powering electron beams should prove useful for experiments requiring only a few events ormore » for very high-energy diode design and scaling tests. 11 figs.« less

Intense bursts of mm wave power have been observed in microwave generation experiments with a relativistic electron beam (REB) virtual cathode oscillator.(1) In this device an electron beam is injected into a drift space at a current above the space-charge-limit, and a potential develops downstream which is large enough to reflect electrons back to the source region. Two mechanisms can give rise to microwave oscillations in a virtual cathode device: electrons reflexing between the real and virtual cathodes, and oscillations in the amplitude and position of the virtual cathode. Typically both mechanisms are present, but in the present experiments reflexing has been shown to be dominant.

The generation of microwave radiation at gigahertz frequencies in high-voltage, high-current, pulsed-power diodes has been investigated both with electromagnetic particle-in-cell simulations and laboratory experiments. Pulsed power in the form of a 1.2-4.5-MV 60-ns TEM wave is fed to the diode by a coaxial line consisting of a 3-cm-diameter anode and a 2-cm-diameter cylindrical stalk terminated by...

The generation of microwave radiation at gigahertz frequencies in high-voltage pulsed-power diodes has been investigated with electromagnetic particle simulations. Pulsed power in the form of a 0.5–1.0MV, TEM wave is fed to the diode via a 5-cm diameter cylindrical stalk onto which a 30-cm diameter cathode has been mounted. Located some 3–5 cm in front of the cathode is a foil anode grid. As the TEM wave propagates between the stalk and an outer cylindrical cage (70-cm diameter) a stream of electrons is emitted off the stalk when the local electric field exceeds 200 kV/cm. This flow is then magnetically insulated and confined by the induced Be fields resulting from the 25–50-kA current flow within the stalk. Convex shaping of the cathode surface allows the emitted electrons to form a virtual cathode beyond the grid, then phase bunch to produce narrow bandwidth 1-GHz microwaves over an area equal to that of the cathode. The electromagnetic radiation from the dipole-like electron motions centered on the grid, which follow closely the classical Barkhausen-Kurz description, is reported.