| dc.description.abstract | The LTP (hydrogen) based negative ion source plays an important role inthe neutral beam injection system - one of the primary means of plasmaheating in magnetic fusion. In this thesis, we have performed PIC-MCCbased simulations of such plasmas wherein the ROBIN negative ion source(consisting of an LTP source with a magnetic filter) installed at IPR,Gandhinagar is taken as a testbed problem for the validation of the model.ROBIN has a driver, an expansion chamber, a magnetic filter, and extractionsystem consisting of 3 different grids. Plasma is generated in the RF driverregion, and that expands in the expansion chamber before encountering themagnetic filter field. A magnetic filter is a localized magnetic field (few tensof gauss) perpendicular to the plasma flow (diffusion flux or transport) andcontrols the plasma flux flowing from the expansion chamber to the extractionsystem. As a first step, we have performed 1D-3V PIC-MCC simulations, andwe observe a good qualitative match between the simulation and experimentalresults in terms of plasma density and electron temperature. The quantitativemismatch between the ROBIN experiment and 1D-simulation results is due tothe fact that the effect of drifts and instabilities (present in real experiments)are not captured properly in the 1D model. However, even with severallimitations, we find that 1D-3V PIC-MCC simulations can predict plasmabehavior in such LTP experiments with acceptable accuracy.As a second step in this direction, we have developed an in-house serial 2D-3VPIC-MCC code and also validated it with results available in the literature.However, stringent numerical constraints associated with a 2D PIC codemake it computationally prohibitive on CPUs in the case of real experimentalgeometry (total number of particles, number of grid points and simulationtime-scale). Therefore, we parallelized our 2D-3V PIC-MCC codes for sharedas well as distributed memory systems consisting of multi-core and many-corearchitectures (GPUs). We have also proposed a hybrid parallel scheme(OpenMP+MPI) which can be used to perform such expensive simulations onan HPC cluster with several nodes. One of the novel contribution towardsthe PIC-MCC code development has been made in terms of using differentparticle sorting strategies which significantly improved the memory accesstime leading to a remarkable enhancement in speedup compared to traditionalstrategies used for PIC-MCC implementation.The parallel 2D-3V PIC-MCC code have been used to simulate ROBINexperiment with real physical dimensions to understand the plasma transportacross magnetic filter. Most of the previous works in this area used ascaled geometry as well as relaxed the stringent numerical criteria for suchsimulations due to computational requirements, however we performedsimulations by satisfying all the strict numerical constraints such as time step,grid spacing and PPC required for kinetic modelling of such LTP experiments.Plasma density and electron temperature profiles from our 2D-3V PICsimulations follow similar trends (qualitative as well as quantitative) asseen in experimental results. This immensely helped us to understand therole of instabilities as well as different diffusion and collisional processes,and subsequently quantifying the plasma transport accurately. Even withcertain limitations present in our model, simulation results show a reasonablygood match with the phase-1 ROBIN experimental results. Particularly thesimulations are showing similar important patterns in plasma characteristicsas seen in the experiments. Comparison of the simulation and experimentalresults from ROBIN gives us sufficient confidence to do further case studiesfor future ROBIN experiments. Several case studies have been performedto understand the role of the magnetic filter profile on plasma transport,which will help in planning future experiments by using the magnetic filteras a switching mechanism to achieve the required density and electrontemperature profiles for efficient operation of negative ion source.Various collision dependent physical phenomena, having different time scalesand length scales are studied using 2D-3V PIC-MCC simulations. We havereported instabilities, observed near the filter field region. It is also observedthat the frequencies of those instabilities are close to some of the electronicand ionic collision frequencies which may create resonant phenomena in themagnetic filter region and influence the cross-field transport, and heating.From our investigations, we find that the application of a bias voltage (appliedto the extraction boundary) changes the potential profile and thereby plays animportant role in controlling the ion temperature near the extraction boundary.The nature of the instabilities also depend on the bias voltage. We areanticipating an ion heating due to instabilities originating in the filter fieldregion. 2D snapshots clearly shows discrete band structure which correspondsto drifts and instabilities, and the frequencies of the instabilities are identifiedusing Fast Fourier Transform (FFT) analysis. The instability correspondingto 105 Hz is identified as E B drift instability whereas, 106 Hz still requiresfurther investigation. In this study, we have shown these instabilities are oneof the causes for ion heating.Drifts and instabilities observed in our simulations may lead to double layer(DL) formation which has not been studied yet in the context of negativeion sources. This motivated us to perform detailed analysis with differentmagnetic field values and different bias voltages. Plasma profiles (such aspotential, electron and ion temperature, and ion velocities) are studied tounderstand the formation of DL and its effect on plasma transport. Ionacceleration is found near both source and extraction boundaries either due tosheath, instabilities, or DL.We observe DL formation under specific conditions(magnetic field and bias voltage). Two velocities components (one due to thefree ions and the other due to the trapped ions) are visible in our simulations.We found that DL depends on both the magnetic field and the differencebetween bias voltage and plasma potential. DL does not occur when the biasvoltage is more or equal to the plasma potential. When the bias voltage isgreater than plasma potential, electron sheath forms and reflects ions from theextraction boundary.A detailed investigation of Energy Distribution Functions (EDFs) helps ininterpreting the complex physics involved in such LTP problems. We havestudied the temporal and spatial evolution of EDFs using our PIC-MCCcode. We have observed that EEDF is Maxwellian in nature, but IEDF isnon-Maxwellian in nature. Our detailed Spatio-temporal analysis of EDFsrevealed that IEDF is more sensitive to changes in the filter field and biasvoltage compared to EEDF.All the past studies have focused on understandingelectron transport, however, our simulations suggest that to completelyunderstand the physics of plasma transport in such low-temperature sources,ion transport is equally important and needs to be investigated in more detail.Efficient negative ion generation in the negative ion source is a critical stepin the neutral beam injection (NBI) system of the future fusion reactor ITER.Achieving few tens of Amperes of H?? current in the negative ion sourceis technically challenging and needs more understanding of the physicsof the plasma transport in such sources. The important contributions ofthis thesis such as identification of instabilities, double layer formationand understanding of EDFs in the context of negative ion sources usingcomprehensive kinetic simulations will further improve our understanding ofphysics of plasma transport and help in enhancing the efficiency of negativeion generation process in such sources. The results are also relevant for similarkinds of different LTP based applications involving magnetic field such as Hallthrusters, ECR source, end-Hall source and magnetron discharge. | |
