NEUTRAL BEAM HEATING OF A REVERSED-FIELD PINCH IN THE MADISON SYMMETRIC TORUS
by
Jeff Waksman
A dissertation submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
(Physics)
at the
UNIVERSITY OF WISCONSIN–MADISON
2013
Date of final oral examination: June 18, 2013
The dissertation is approved by the following members of the Final Oral Committee: Cary Forest, Professor, Physics Jay Anderson, Associate Scientist, Physics Chris Hegna, Professor, Engineering Physics Paul Terry, Professor, Physics David Anderson, Professor, Electrical And Computer Engineering c Copyright by Jeff Waksman 2013 All Rights Reserved i
A star is drawing on some vast reservoir of energy by means unknown to us. This reservoir can scarcely be other than the sub-atomic energy which, it is known, exists abundantly in all matter; we sometimes dream that man will one day learn how to release it and use it for his service. The store is well-nigh inexhaustible, if only it could be tapped.
— Sir Arthur Stanley Eddington, 1920 ii
Abstract
Neutral beam injector (NBI) heating of electrons has been observed in the Madison Symmetric Torus (MST). This heating is consistent with a simple 1D model which includes the effect of fast ion diffusion, neutralizationlosses and classical collisional processes. This heating was achieved with a 1 MW NBI (25 kV, 40 A). Auxiliary heating of MST has been observed with Thomson scattering to be 100 50 eV in the core of 200 kA PPCD (pulsed ± poloidal current drive) plasmas. Auxiliary heating has been measured in other PPCD plasmas, but not in standard confinement plasmas. This data represents the first confirmed auxiliary heating of a reversed-field pinch (RFP) plasma. Enhanced confinement plasmas are conducive to auxiliary heating because of the improved thermal confinement. Ion temperatures were measured with a Rutherford scattering device, but investigation deter- mined that NBI fast ions are able to get into the Rutherford analyzer at a large enough rate to add a significant non-Gaussian component to the observed data. Controlling for this additional neutral hydrogen signal eliminates false detection of auxiliary heating with NBI in standard MST plasmas. Further analysis suggests that fast ions from magnetic reconnection events could also be polluting MST sawtooth Rutherford data. A 1-D model of NBI heating in MST was developed. This model takes measured and calculated inputs such as plasma temperature, plasma density and ohmic heating profiles and solves for heat diffusion coefficient profiles. The assumption underlying this calculation, that the heat diffusion coefficients do not change when the NBI is firing, is justified because the NBI does not significantly suppress mid-radius magnetic fluctuations. The base 1-D model includes fast ion deposition, fast ion diffusion and fast ion slowing down. Fast ions are lost to the plasma via thermalization and contact with the MST wall. Neutral losses iii are added to the model using NENE,` a Monte Carlo code. Finally, non-classical (resonant) fast ion diffusion is observed and modeled. This diffusion acts to limit core fast ion density, and appears as rapid bursts that occur approximately once or twice every millisecond.
The model output show that the 1-D classical model is consistent with measured Te in PPCD plasmas. Fast ion diffusion is crucial in driving a flatter heating profile to limit heat conduction- losses. Measured core Te is only possible with significant mid-radius heating. Neutralization losses are modeled, which provide a loss mechanism for fast particles near the wall. Low core neutral densities in PPCD are crucial for the measured auxiliary heating. Neutral- ization losses are significant in standard plasmas, and explain the lack of significant Te. Finally, resonant diffusion is found not to have a significant effect on auxiliary heating of MST as long as a loss mechanism (neutralization) exists for fast ions near the MST wall. The 1-D model demonstrates that auxiliary heating with NBI in PPCD can be modeled classi- cally, where fast ion diffusion is a crucial physical process. Non-PPCD plasmas require a signifi- cant loss mechanism for fast ions near the MST wall, which can be provided by neutralization. iv
ACKNOWLEDGMENTS
When I get asked years from now how I managed to survive graduate school, my answer will be that the MST team is all anybody could possibly ask for. The collaborative and cooperative spirit of every single person working on MST is a model for how science should be done. I will very much miss working in the MST control room. There are individuals within the group that deserve special thanks. First and foremost are my faculty advisors past and present, Stewart Prager and John Sarff. I also need to thank Cary Forest for stepping in at a difficult time to fill John’s shoes, saving my thesis defense. Particular thanks go to Jay Anderson and Gennady Fiksel for working with me on a daily basis, and suffering literally thousands of door knocks as I asked question after question. Gennady is the godfather of the NBI group, and it is his vision and hard work that formed the initial backbone of all of this research. Jay has done a fabulous job leading the NBI/fast ion group the past few years. Without Jay’s guidance I absolutely never would have gotten my 1-D model to start working. As Jay spent more time with me than literally anybody else on MST, I’d appreciate it if you direct any complaints about this dissertation to him. The entire NBI/fast ion group has been great, but I want to thank several in particular who have really spent a lot of time helping me along the way. Mark Nornberg provided a ton of guidance with thesis writing and coding. Deyong Liu found time during his 23 hour/day work days to do all of the TRANSP runs you see here, as well as to do a lot of the initial fast ion confinement analysis. Jon Koliner, Scott Eilermann, Josh Reusch and the Thomson scattering team all went out of their way to take and analyze a bunch of data for this dissertation. I need to thank Abdul Almagri for countless hours working on the Rutherford beam with me. I think we have tested/cleaned/fixed/replaced every single component of that beam at this point. v
MST is blessed with an incredible engineering and technical staff. Future grad students take note: If you have a technical problem of any kind, discuss it loudly near the offices of MST’s technical staff and they will come out of their offices and they will solve your problem. While every single person on staff has provided assistance of some kind or another to me, I want to thank several in particular. Steve Limbach and Steve Oliva have both spent many, many hours working on various neutral beams. The Hackerium staff of Paul Wilhite, Alex Squitieri and Andrey Levochkin has been invaluable. The electronics shop staff of Mikhail Reyfman and Dave Deicher have done a great job deciphering electronics problems, even when I don’t have drawings. I also want to thank Paul Nonn for taking the lead on the bolometer calibration setup when I was a green graduate student who didn’t know the difference between a drill press and a lathe. And I can’t forget Peter Weix, Bill Zimmerman, John Laufenberg and everybody else who kept MST running the past 7 years. Even when I blew up the L-6 inductor we were only down for about two days. I want to thank my fellow graduate students who kept me sane by playing around 8 million games of pickup basketball. I want to give special thanks/sympathy to Meghan McGarry-Unks for being my office mate for nearly my entire time in graduate school. Last, but definitely not least, I want to thank my family for everything they’ve done to make it possible for me to finish this dissertation. I got to see you far more than a person living 1000 miles away probably should (although I have a sneaking suspicion that Dad would have kept coming regularly on his own, even if I wasn’t living here, once he discovered Camp Randall, the Kohl Center and fried cheese curds). Anything I achieve in my life will be because of you. But no, you can’t call me ”Doctor Waksman”. vi
TABLE OF CONTENTS
Page
Abstract ...... ii
List of figures ...... ix
1 Madison Symmetric Torus ...... 1
1.1 The Reversed Field Pinch ...... 1 1.2 Madison Symmetric Torus ...... 5 1.3 Standard Confinement In The RFP ...... 8 1.4 The Sawtooth Cycle ...... 10 1.5 Pulsed Poloidal Current Drive ...... 15 1.6 The RFP As A Fusion Device ...... 18 1.7 Thesis Overview ...... 23
2 Experiment Diagnostics ...... 26
2.1 Neutron Detection ...... 26 2.2 Thomson Scattering ...... 27 2.3 Far-Infrared (FIR) Interferometer-Polarimeter System ...... 28 2.4 CHERS ...... 31 2.5 ANPA ...... 34 2.6 Neutral Density Profiles Via D↵ ...... 36
3 Neutral Beam Injection ...... 40
3.1 A Brief History Of Neutral Beams On Fusion Devices ...... 40 3.2 MST’s 1 MW Neutral Beam ...... 42 3.3 Classical Fast Ion Dynamics ...... 45 3.4 Measuring NBI Shine-Thru ...... 47 3.5 Fast Ion Confinement ...... 53 3.5.1 Calculating Fast Ion Confinement ...... 53 3.5.2 Classical Fast Ion Confinement Theory ...... 56 3.5.3 Fast Ion Confinement Data ...... 58 3.5.4 Resonant Fast Ion Transport ...... 63 3.6 NBI Momentum Drive ...... 68 vii
Page
3.7 NBI Current Drive ...... 73 3.7.1 Current Drive Theory ...... 73 3.7.2 Current Drive Data ...... 76
4 Rutherford Scattering ...... 80
4.1 History of Rutherford Scattering ...... 80 4.2 MST’s Rutherford Scattering Diagnostic ...... 82 4.3 Rutherford Scattering Theory: Single Particle ...... 84 4.4 Rutherford Scattering Theory For An Ion Velocity Distribution ...... 87 4.5 Rutherford Scattering Theory: Generalized For A Beam With Finite Divergence . . 88 4.6 Improved Rutherford Processing ...... 90 4.7 Rutherford Analyzer As Neutral Hydrogen Analyzer ...... 92 4.8 Rutherford Analyzer Sawtooth Analysis ...... 97
5 NBI Heating Of MST Plasmas ...... 101
5.1 Measured Auxiliary Heating Of MST Plasma ...... 103 5.2 Auxiliary Heating Data In A Variety Of Plasma Conditions ...... 107 5.3 Impact Of Current & Density On Auxiliary Heating ...... 113
6 Modeling NBI Heating ...... 116
6.1 A 1-D Classical Heating Model ...... 117 6.2 Justifying A Classical Heating Model ...... 123 6.3 Modeling With TRANSP/NUBEAM ...... 128 6.4 Anomalous Ion Heating ...... 131 6.5 Modeling The NBI ...... 133 6.5.1 Modeling The NBI Beam Path ...... 134 6.5.2 Modeling NBI Beam Deposition ...... 137 6.5.3 Modeling Fast Ion Diffusion And Slowing Down ...... 142 6.6 Model Output Comparisons To Data ...... 149 6.7 Impact Of Zeff On 1-D Model ...... 158 6.8 Impact Of Fast Ion Diffusion On Heating ...... 161 6.9 Modeling Fast Ion Charge-Exchange Losses ...... 167 6.10 Modeling Resonant Fast Ion Transport ...... 177
7 Conclusions And Future Work ...... 183
7.1 Neutral Beam Heating Of MST ...... 183 viii
Appendix Page
7.2 1-D Auxiliary Heating Model ...... 184 7.3 Future Work ...... 185
References ...... 187
APPENDIX Bolometers And XUV Detectors ...... 202 ix
LIST OF FIGURES
Figure Page
1.1 Illustration of the RFP magnetic field configution ...... 4
1.2 Photograph of MST ...... 5
1.3 Cartoon of a toroidal pinch experiment ...... 6
1.4 Directions in MST ...... 7
1.5 Typical magnetic field profiles in MST ...... 9
1.6 Modeling of stochastic magnetic fields in MST ...... 10
1.7 Example data from a standard MST discharge ...... 11
1.8 Plasma relaxation during a sawtooth crash ...... 13
1.9 Anomalous Ion Heating During A Reconnection Event ...... 14
1.10 Reduced magnetic fluctuations during PPCD ...... 16
1.11 Changes in core Te and Ti During PPCD ...... 17
1.12 Electron temperature profile changes during enhanced confinement ...... 17
1.13 MST confinement compared to scaled tokamak confinement ...... 20
1.14 MST confinement compared to modified Connor-Taylor scaling ...... 20
1.15 RFP confinement scaling in RFX-mod ...... 20
1.16 Cutaway drawing of TITAN ...... 22
2.1 A cartoon of the observation of Thomson scattering ...... 28 x
Figure Page
2.2 A drawing of MST’s FIR system ...... 29
2.3 Impact of temperature, velocity and density on CHERS measurements ...... 32
2.4 Comparison of CHERS and Rutherford PPCD data ...... 34
2.5 Design of the MST ANPA ...... 35
2.6 Photo of the MST ANPA ...... 36
2.7 NENE` calculation of neutral density profile, 200 kA PPCD ...... 37
2.8 NENE` calculation of core neutral density, 400 kA standard ...... 38
3.1 Evidence for NBI heating of TPE-RX ...... 42
3.2 CAD drawing of MST’s 1 MW NBI ...... 43
3.3 Photograph of MST’s 1 MW NBI ...... 44
3.4 A cartoon of neutral beam charge exchange ...... 45
3.5 Photograph of the NBI shine-thru detector ...... 47
3.6 Drawing of the NBI shine-thru design ...... 48
3.7 Plot of electron-impact cross section vs Te ...... 50
3.8 Plotting shine-thru measured at different NBI alignments ...... 50
3.9 Photo of the NBI smashing into the beam dump ...... 51
3.10 Shine-thru data for 200 kA low density PPCD ...... 52
3.11 Shine-thru data for 200 kA high density PPCD ...... 52
3.12 Neutron signal vs early gas ...... 53
3.13 Neutron decay vs fast ion loss time ...... 55
3.14 Measured fast ion loss times in various plasmas ...... 55
3.15 Fast ion guiding center safety factor profile ...... 57 xi
Appendix Figure Page
3.16 Neutron data for 300 kA, F=0 ...... 59
3.17 Neutron data for 200 kA, low density PPCD ...... 59
3.18 Neutron data for 200 kA, low density PPCD ...... 59
3.19 Neutron data for 400 kA, low density PPCD ...... 59
3.20 ANPA data from a 400 kA standard plasma ...... 62
3.21 Energetic particle-driven instabilities in MST plasmas ...... 64
3.22 Resonant transport of fast ions in DIII-D ...... 65
3.23 Energetic particle-driven instabilities in 200 kA low density PPCD ...... 67
3.24 ANPA data during fast-ion driven modes in 200 kA PPCD ...... 68
3.25 Impact of NBI on magnetic fluctuation magnitude and velocity ...... 69
3.26 Co-current vs Counter-current NBI data ...... 71
3.27 Neutron flux, Co- vs Counter- NBI ...... 72
3.28 Simulated fast ion density, Co- vs Counter- ...... 72
3.29 Data from DIII-D showing NBI current drive ...... 74
3.30 Data from MAST showing NBI suppression of loop voltage ...... 74
3.31 Impact of tearing modes on NBI-driven current in DIII-D ...... 76
3.32 NBI current drive data on MST ...... 77
3.33 FIR polarimetry data during NBI ...... 78
3.34 FIR polarimetry data during NBI, zoomed in on the plasma core ...... 78
4.1 Schematic of MST’s Rutherford beam and analyzer ...... 83
4.2 Photograph of MST’s Rutherford analyzer ...... 84
4.3 Cartoon of single particle Rutherford scattering ...... 85 xii
Appendix Figure Page
4.4 Small angle approximation diagram ...... 86
4.5 Geometry of a finite-width beam Rutherford scattering ...... 89
4.6 Modeled Rutherford scattering energy distribution ...... 90
4.7 Example of adjusted Rutherford gas-only data ...... 93
4.8 Example of adjusted Rutherford plasma data ...... 93
4.9 Rutherford NBI “heating” data ...... 94
4.10 Rutherford as an NPA: Raw Data ...... 95
4.11 Rutherford as an NPA: Processed Data ...... 96
4.12 Fast Ion Tail Following A Sawtooth Event ...... 97
4.13 Rutherford Temperature Output, 400 kA Standard ...... 98
4.14 Raw Rutherford Data, 400 kA Standard ...... 99
4.15 Raw Rutherford Data vs Time, 400 kA Standard ...... 99
5.1 Example plasma parameters for 200 kA low density PPCD data ...... 102
5.2 Core electron temperature with and without NBI in low density 200 kA PPCD . . . . 104
5.3 Change in core electron temperature with NBI in 200 kA low density PPCD ...... 104
5.4 Electron temperature profiles with and without NBI for low density 200 kA PPCD . . 105
5.5 CHERS data, 200 kA low density PPCD ...... 106
5.6 Core electron temperature with and without NBI in high density 200 kA PPCD . . . . 108
5.7 Change in core Te with NBI in 200 kA high density PPCD ...... 108
5.8 CHERS data, 200 kA high density PPCD ...... 108
5.9 Core electron temperature with and without NBI in low density 400 kA PPCD . . . . 109
5.10 Change in core Te with NBI in 400 kA low density PPCD ...... 109 xiii
Appendix Figure Page
5.11 Core electron temperature with and without NBI in low density 400 kA PPCD . . . . 110
5.12 Change in core Te with NBI in 400 kA low density PPCD ...... 110
5.13 Core electron temperature with and without NBI in 300 kA F=0 plasmas ...... 111
5.14 Change in core Te with NBI in 300 kA F=0 ...... 111
5.15 CHERS data, NBI On vs Off, 300 kA PPCD ...... 111
5.16 Core electron temperature with and without NBI in 400 kA standard plasmas . . . . . 112
5.17 Change in core Te with NBI in 300 kA F=0 ...... 112
5.18 Stored electron energy vs time, 200 kA low density PPCD ...... 114
5.19 Stored electron energy vs time, 200 kA high density PPCD ...... 114
5.20 Change in stored electron energy, 200 kA low density PPCD ...... 114
5.21 Change in stored electron energy, 200 kA high density PPCD ...... 114
6.1 A cartoon of the “ith” plasma volume element ...... 118
6.2 Example calculated e profiles ...... 119
6.3 Cartoon of the two-dimensional array used in the Crank-Nicolson method ...... 121
6.4 Matrix form of the Crank-Nicolson solution ...... 122
6.5 Relationship between B˜ suppression and Te(0) ...... 124
6.6 Supression of the core magnetic modes by NBI ...... 125
6.7 NBI suppression levels for all magnetic modes ...... 126
6.8 Supression of the core magnetic modes by NBI in PPCD ...... 127
6.9 NBI suppression levels for all magnetic modes in PPCD ...... 127
6.10 Typical toroidal flux profile in MST PPCD ...... 128
6.11 TRANSP fast ion deposition shape ...... 129 xiv
Appendix Figure Page
6.12 TRANSP/NUBEAM model core heating ...... 130
6.13 Impact of ion temperature assumptions on TRANSP/NUBEAM model ...... 132
6.14 Overhead view of simulated NBI path ...... 135
6.15 Distance of the NBI path from the geometric axis ...... 135
6.16 “Slice” view of simulated NBI beam ...... 136
6.17 Assumed NBI Beam Width ...... 138
6.18 NBI beamlet path ...... 138
6.19 Magnetic field “focuses” deposited fast ion ...... 140
6.20 Fast neutral density along NBI path ...... 141
6.21 Modeled fast ion density profiles ...... 143
6.22 Modeled fast ion energy deposition profile, 1-D model ...... 144
6.23 Modeled fast ion energy deposition, TRANSP/NUBEAM ...... 144
6.24 Modeled fast ion energy deposition profile, hybrid model with diffusion ...... 146
6.25 Modeled fast ion energy deposition, hybrid model without diffusion ...... 146
6.26 Comparing the modeled heat deposition profiles ...... 147
6.27 Modeled fast ion energy loss mechanisms, 1-D model ...... 148
6.28 Modeled fast ion energy loss mechanisms, TRANSP/NUBEAM ...... 148
6.29 Modeled change in core electron temperature ...... 150
6.30 Modeled electron temperature profiles ...... 150
6.31 Modeled heat diffusion losses, 1-D model ...... 152
6.32 Modeled ohmic power density, 1-D model ...... 153
6.33 Comparison of core heating components, NBI On vs NBI Off ...... 154 xv
Appendix Figure Page
6.34 Comparison of core heating components, 1-D model vs TRANSP ...... 154
6.35 Modeled core electron temperature gradient ...... 155
6.36 Impact of ion temperature assumptions on 1-D heating+deposition model ...... 156
6.37 Impact of supressed magnetic fluctuations on core Te ...... 158
6.38 Change in core Te for varied Zeff profiles in 200 kA PPCD ...... 160
6.39 Change in core Te for varied Zeff profiles in 400 kA PPCD ...... 161
6.40 1-D model output for varying TRANSP inputs ...... 162
6.41 Hybrid model core heating for varying diffusion coefficients ...... 164
6.42 1-D heating+deposition model output for varying diffusion coefficients ...... 165
6.43 Fast ion density profiles: Diffusion On vs Off ...... 166
6.44 1-D model output, varying diffusion and neutral density ...... 168
6.45 An example “reduced” neutral density profile ...... 168
6.46 Fast ion density profiles with charge-exchange ...... 170
6.47 1-D model output for 400 kA PPCD, varying neutral density ...... 171
6.48 Core neutral density vs time, 400 kA PPCD ...... 171
6.49 Fast ion density profiles with varying neutral densities, 400 kA PPCD ...... 172
6.50 1-D heating+deposition model output for 400 kA standard ...... 174
6.51 NENE` output, ensembled vs single-shot data ...... 175
6.52 Fast ion density profiles for 400 kA standard ...... 176
6.53 Simulated fast ion profiles with resonant diffusion ...... 178
6.54 Simulated core fast ion density vs time with resonant diffusion ...... 178
6.55 Simulated core Te with resonant diffusion ...... 179 xvi
Figure Page
6.56 Simualted core Te for various values of DRES ...... 179
6.57 Fast ion density profiles with and without resonant diffusion ...... 180
6.58 Calculated Te, 400 kA standard, including non-classical effects ...... 181
6.59 Simulated core fast ion density, 400 kA standard with resonant diffusion ...... 181
Appendix Figure
A.1 Cartoon of pyrobolometer calibration setup ...... 205
A.2 Bolometer calibration: optimizing the bias voltage ...... 207
A.3 AXUV detector sensitivity vs photon energy ...... 209
A.4 Photo of a bolometer/XUV pair on MST ...... 210
A.5 Typical bolometer/XUV data ...... 212
A.6 Typical bolometer/XUV data (zoom) ...... 212
A.7 Bolometer/XUV data (Standard), NBI On vs Off ...... 214
A.8 Bolometer/XUV data (PPCD), NBI On vs Off ...... 214 1
Chapter 1
Madison Symmetric Torus
1.1 The Reversed Field Pinch
Reversed Field Pinch (RFP) research is, for the most part, inspired by and funded because of the desire for fusion power. Fusion power, if harnessed and controlled on Earth, presents a nearly infinite source of clean, renewable power. The work discussed throughout this dissertation should be understood within that context. The discovery of nuclear fusion actually goes back nearly 100 years. The catalyst was the development of the mass spectrometer by Francis W. Aston in 1919, which separated out isotopes of the elements of the periodic table, allowing him to accurately measure their respective masses (a discovery for which he won the Nobel Prize in Chemistry in 1922). In 1920, Aston was able to measure the mass of hydrogen (1.008 amu) and the mass of helium (4.003 amu). Arthur Stanley Eddington, one of Aston’s colleagues at Cambridge University, speculated that same year that the fact that helium was a little less than four times the mass of hydrogen meant that combining four hydrogen together would create a helium atom along with a release of energy. Einstein’s famous formula for mass-energy equivalance (E = mc2) meant that even that tiny bit of lost mass would mean a large release of energy. At the time, the dominant theory of the Sun was Lord Kelvin’s 19th century determination that gravitational potential energy was released as heat, leading him to calculate an age for both the Sun and Earth of tens of millions of years. Eddington argued instead that the Sun was releasing 2 heat from these fusion reactions, and calculated that the Sun had 15 billion years of fuel. While Eddington’s conception of the fusion reaction being four hydrogens combining into one helium was wrong (in his defense, the neutron wouldn’t even be discovered for another decade), he was correct that fusion fueled the sun, and his estimation of 15 billion years of fuel was remarkably close to accurate. It was at a speech presenting this research, on August 24th, 1920, that Eddington gave the powerful and prescient quote that opens this dissertation. The basic theory by which hydrogen isotopes could fuse to create helium+energy was devel- oped over the next few decades, giving scientists the confidence that we could engineer fusion on Earth. The first man-made release of fusion energy was in 1952, when the hydrogen bomb was tested for the first time. Controlling fusion power in the lab proved to be a more complicated en- deavor, since the conditions for significant fusion require extremely hot (hundreds of millions of
20 3 degrees) plasmas confined at relatively large densities ( 10 m ). Research was quickly begun in magnetic confinement of high temperature plasmas, which remains the dominant fusion research in the world today. The earliest designs were mostly Z-pinches, followed by Lyman Spitzer’s fa- mous stellarator in 1951, and then the magnetic mirror. The spark that began the modern era of fusion research came in 1968, at an IAEA meeting in Novosibirsk, where Soviet researchers announceed that their new T-3 tokamak design had achieved central electron temperatures reaching 1 keV (more than ten million degrees kelvin, nearly 2,000 ⇡ times hotter than the surface of the sun). A team from Culham Laboratory in England traveled to the Soviet Union to verify the results with their own Thomson scattering device (see more on Thomson scattering in Sec. (2.2)). Their verification of 1 keV temperatures immediately made the tokamak the most popular fusion design, which is the status it still holds today. The largest magnetically-confined fusion devices today in terms of plasma current, temperature and fusion power achieved are almost all tokamaks, including JET, EAST, NSTX and DIII-D, among many others. The next step is ITER, an international tokamak of staggering size that is already under construction in Cadarache, France, with first plasma projected for 2019. ITER is designed to be the transition step to a full-scale electricity-producing fusion power plant (DEMO), which would then lead to commercial fusion power as part of the electrical grid. 3
While the tokamak is the most popular form of magnetic confinement of plasmas, there are many other confinement designs in use in large scales today. Stellarators and mirrors are still used, as are Spherical Tokamaks, a Levitated Dipole, the Reversed Field Pinch and others. The basis for the Reversed Field Pinch can be traced all the way back to the ZETA machine, which operated at Harwell Laboratory in the United Kingdom from 1957 through 1968. ZETA was a toroidal pinch device, and the largest fusion device of its era. ZETA is unfortunately famous for the wrong reasons now. It was at ZETA in 1958 that high measured neutron fluxes were mis- interpreted, and published in Nature with the blaring title: “Controlled Release Of Thermonuclear Energy” [Thonemann et al., 1958]. It was realized soon after that these high neutron signals were due to an ion acceleration process rather than fusion, and a retraction was published five months later [B.Rose et al., 1958]. Nevertheless, ZETA did play a pivotal role in the development of the RFP. It was several years after ZETA was decommissioned that John Bryan Taylor wrote his seminal paper [Taylor, 1974], based on data from ZETA operation. Taylor noted that after the initial start-up of the plasma, it would often relax into a “quiescent”, stable state. Taylor noted that for large pinch ratios, a reversed toroidal field would actually be generated near the edge. To understand what Taylor was observing we can talk about the helicity (or “knottedness”) of the magnetic field lines by defining [Moffatt, 1969]:
K = A BdV (1.1) · ZV Here, K is the magnetic helicity and A is the magnetic vector potential (B = A). Magnetic r⇥ helicity is a conserved quantity for a perfectly conducting plasma, though there will be a slight decay in helicity in a resistive plasma, as was measured in MST by Ji et al. [1995]. Over time, the plasma will “relax” into a state of minimum energy. Assuming conserved helic- ity, Taylor found the solution for this state in his 1974 paper:
B = B (1.2) r ⇥ The ideal MHD momentum balance equation is: dV ⇢ = JXB p (1.3) m dt r 4
J is the current density, p is the plasma pressure, V is the bulk plasma velocity and ⇢m is the plasma density. The equilibrium ( @ =0) form of Eq. (1.3) is JXB = p. Taken with Ampere’s` @t r Law ( B = µ J), we can say that the plasma is “force free”, since: r ⇥ o B p = J B = r ⇥ B = B B =0 (1.4) r ⇥ µo ⇥ µo ⇥
This relaxed state is known today as the “Taylor State”. Solving Eq. (1.2) for the case of a cylinder gives the magnetic field components in terms of Bessel functions [Bodin & Newton, 1980]:
B = BoJo( r) (1.5)
B✓ = BoJ1( r) (1.6)
This value is related to the pinch paramter (✓ = B (a)/B¯ 2⇡aI / ) by the expression ✓ ✓ ⇠ a/2 [Taylor, 1986]. The solution to Eq. (1.5) is a reversed toroidal field when r > 2.405. For the ZETA data Taylor was looking at, this correlated to a pinch parameter ✓ =1.202, remarkably similar to the measured ✓ 1.4 observed for the onset of the reversed field in ZETA. Thus, for ⇡ appropriate initial toroidal fields and toroidal currents, the plasma will relax into a state with a reversed toroidal field.
Figure 1.1 A cartoon of the magnetic field configuration of a typical RFP. 5
The first modern reversed-field pinch was ETA BETA II, which had its first plasma results in 1979. This was followed in the early 1980s by experiments such as TPE-1RM, ZT-40(M) and HBTX1A. The next stage of large-scale RFP experiments consisted of the Madison Symmetric Torus (MST), Reversed-Field eXperiment (RFX), RELAX, TPE-RX and EXTRAP-T2R, all of which had first plasmas around 1990. Of those five machines, all are still in operation except for TPE-RX. A new RFP, the Keda Torus eXperiment (KTX), is currently in the design and construc- tion phase in Hefei, China.
1.2 Madison Symmetric Torus
Figure 1.2 Photograph of the Madison Symmetric Torus
The Madison Symmetric Torus (MST) is a large toroidal reversed-field pinch with a major radius (R) of 1.5 m and a minor radius (a) of 0.52 m. The torus consists of a 5 cm thick aluminum shell that serves several purposes. It is a vacuum vessel, maintaining pressures 1 µTorr between ⇡ shots. It also serves as a conducting wall very close to the plasma (carbon limiters with a width ⇡ 1 cm keep the plasma off the wall) to help stabilize ideal instabilities. It also serves as a single-turn toroidal field coil to generate the start-up toroidal magnetic field. Because the torus is a conducting 6 shell, all port holes (for diagnostics, vacuum pumping, etc.) are kept small, to limit magnetic field errors.
Figure 1.3 A generic sketch of a toroidal pinch experiment [Anderson, 2001]
Typical plasma pulses in MST last for anywhere between 40 and 100 ms. Though MST can be run with several different fuels, all data used in this research were done with the standard deu-
19 3 terium. Electron density (n ) at the core is typically around 0.3 1.5 10 m , though densities e ⇥ 19 3 greater than 4 10 m have been reached during pellet injection [Wyman et al., 2008]. Core ⇥ electron temperature is typically between 0.2-2.0 keV. Toroidal plasma current (IP ) is typically between 200-600 kA. The basic operation of MST is described in Dexter et al. [1991]. A 1.3 cm insulated gap, known as the toroidal gap, is between the top and bottom half of MST, on the inboard side. A potential across this gap generates a poloidal electric current in the shell, which generates a starting BT . Next, MST behaves as a large transformer (as seen in a simplified cartoon in Fig. (1.3) [Anderson, 2001]), with the plasma (sparked by a pair of filaments) acting as a single-turn secondary. The transformer is pulsed, which generates a toroidal plasma current. This plasma current generates the poloidal field. 7
V pg + – 0° 360°
Bp v , v , Ip top view of t mode vDi MST edge p Er – p poloidal cut view core core Bt, z + V vDi tg vE x B vE x B during edge sawtooth J vmode edge Er Bp Bt, z
vt, Ip
Figure 1.4 A cartoon of the magnetic field, plasma current, pressure gradient and drift directions in MST. Note that the typical direction of the poloidal field is the opposite of the cartoon in Fig. (1.1).
The poloidal mean field MF drives current parallel to the magnetic field, which near the wall E is often nearly totally poloidal. This current drives a magnetic field in the reverse toroidal direction (see Fig. (1.4) for a drawing of the directions of these currents and fields). The reversal paramater (F ) in an RFP is defined as: B (a) F = T (1.7) B h T i This reversal parameter is a free experimental parameter for the MST operator. A larger startup
BT will mean less reversal (an F>0 mode is possible) while a smaller startup BT with a large driven poloidal current will mean deep reversal. A “standard” MST discharge has F=-0.2. A parameter that describes the amount of field line twist is the “safety factor,” q:
rB (r) q(r)= T (1.8) RBT (r)
By definition, q=0 at the reversal surface. In typical MST plasmas, q 0.2 in the core. ⇡ 8
1.3 Standard Confinement In The RFP
In any magnetic configuration, perturbations to the magnetic field can be described by a Fourier series. Working in a spherical coordinate system with minor radial component r and major ra- dius R, and assuming that the equilibrium (unperturbed) magnetic field can be described as B =
(0,B✓,B ) (B✓ and B are the general variables used for the poloidal and toroidal magnetic fields, respectively): i(k r !t) B˜(r, t)= B˜k(r)e · (1.9) Xk The wave vector k is defined as k =(kr,m/r,n/R), where m and n are the poloidal and toroidal mode numbers respectively. These magnetic perturbations are resonant when the wave vector is perpendicular to the equilibrium magnetic field:
m n 0=k B = B + B (1.10) · r ✓ R
Eq. (1.10) can solved for m/n, and for the safety factor:
m r B (r) = = q(r) (1.11) n R B✓(r)
The modes are thus resonant for any rational values of q, which represent the locations where the magnetic perturbation makes an integer number of poloidal rotations for an integral number of toroidal rotation. For example, the m=1, n=6 mode is a magnetic mode which makes exactly six poloidal rotations for every one toroidal rotation. All of the m=0 modes are resonant at q=0, which is the reversal surface. If these resonant modes are unstable then they will grow in amplitude, and the magnetic field lines can break into chains of magnetic islands at mode-rational surfaces [Biewer et al., 2003], where magnetic islands will be created by magnetic reconnection (see Rechester & Stix [1976] and Callen [1977]). The width of the magnetic islands is given as [Fiksel et al., 2005; Lichtenberg & Lieberman, 1983]:
˜bmn rmn wmn =4 (1.12) s B✓ nq mn 0 9
Standard MST Magnetic Profiles 1
0.75 B 0.5 0.25 B 0
Normalized B-Fields -0.25 0.2
0.15
0.1 /RB
0.05 q=rB 0
-0.05 0 0.2 0.4 0.6 0.8 1 r/a
Figure 1.5 Typical magnetic field profiles in MST. The top plot has normalized toroidal and poloidal magnetic fields for a typical MST plasma. The bottom plot has a typical q profile.
In Eq. (1.12), n and m are the toroidal and poloidal mode numbers, and q is the gradient of 0mn the q-profile of the m,n rational surface. The lower-n resonant surfaces (n=5,6,7) are the core-most m=1 modes in standard MST discharges, and they also occur at the the flattest part of the q-profile (see Fig. (1.5)). This means that the magnetic islands nearest the core will be largest, causing island overlap. The magnetic field lines become tangled and stochastic. Modeling of what this looks like is shown in Fig. (1.6). Overlapping magnetic islands create conditions of rapid radial transport, as the free streaming orbits of particles parallel to the magnetic field result in radial energy and particle diffusion [Biewer et al., 2003]. In standard MST plasmas this results in energy confinement times of 1 ms [Chap- ⇡ man et al., 2002]. This poor confinement is a significant problem for any potential fusion reactor, and this led to the development of an improved confinement scheme, pulsed poloidal current drive (PPCD), which is described in Sec. (1.5). 10
Figure 1.6 Modeling from Biewer et al. [2003]. The top plot includes the calculated magnetic island widths on top of the q-profile. The lower plot is a magnetic field line tracing applying the MAL code from a DEBS simulation.
1.4 The Sawtooth Cycle
The “Taylor State” described in Eq. (1.5) and Eq. (1.6) was solved for a perfectly conducting cylinder with a force-free plasma. Since MST is not a cylinder, and has finite resistivity and a non-zero pressure gradient, MST never truly reaches a relaxed state. Experimentally, the toroidal shape of MST and the magnetic shear leads to less parallel current density on the edge and greater parallel current density in the core [Terry et al., 2004]. Over time, the higher temperatures in the core lead to lower resistivities and an even larger accumulation of parallel plasma density. The existence of the plasma-wall boundary at the edge increases resistivity at the edge and also helps contribute to the peaking of the plasma current profile [Chapman, 1997]. This gradient is a source of free energy for MHD instabilities such as the tearing modes described in Sec. (1.3) is created [Biewer et al., 2003]. At some critical gradient, tearing instabilities rapidly increase, driving a rapid “relaxation” event, leading to a flattening of the profile and the pressure profile.
12
An RFP such as MST, with standard confinement, will go through this cycle repeatedly. The plasma will continue accumulating free energy, having a rapid relaxation event, and starting all over again. The data produced by this cyclical process visually looks like sawteeth, and so it is known as a “sawtooth cycle”. Each relaxation event itself is typically called a “sawtooth crash”. Data for a standard MST 400 kA plasma and sawtooth cycle can be seen in Fig. (1.7). Note that during each relaxation/sawtooth event, there is a rapid fluctuation in equilibrium toroidal current and a large increase in equilibrium magnetic field, in addition to the gigantic spike in magnetic fluctuations (on the order of 1-2% of ). The toroidal gap voltage always sees a very brief spike in signal at the precise moment of a sawtooth crash, and this is why sawtooth analysis on MST typically uses toroidal gap voltage as a time marker. During each relaxation event the plasma is tending toward the theoretical Taylor State which, as described in Eq. (1.4), has a flat current density profile and is “force-free”, meaning that the pressure gradient is zero. While MST never actually achieves a p =0state, the pressure gradient 5 does flatten out quite a bit during each relaxation event. Both the electron density and temperature profiles around a sawtooth crash have been measured carefully in MST. The electron density and temperature profiles around a relaxation event are plotted in Fig. (1.8). The density data were taken with the FIR interferometer (Sec. (2.3)) and are from Ding et al. [2009]. The electron temperature data were taken with Thomson scattering (Sec. (2.2)) and are from Reusch [2011]. In both cases the peaked profiles flatten out over a very quick ( 200 µs) timescale. ⇠ One other important physical result of relaxation events is anomalous energy transfer to ions. While electrons see a decrease in temperature during a sawtooth crash, ions see a dramatic temper- ature increase. The existence of so-called “anomalous” ion heating has long been studied in many different pinch machines [Fujita et al., 1991] going all the way back to ZETA [Jones & Wilson, 1962]. Even away from sawtooth events, the anomalous heating term is required to explain why electron and ion temperatures are approximately equal in equilibrium RFP plasmas, where the en- ergy confinement time (⌧✏) is shorter than the thermal relaxation time between ions and electrons (⌧ ), and so a classical collisional process gives a relation Ti (⌧ /⌧ )T , which implies that the ei ⇡ ✏ ei e ion temperature should be lower than the electron temperature. [Fujisawa et al., 1991]. 13
Figure 1.8 Data from an ensemble of relaxation events. The top plot is density data taken with the FIR interferometer described in Sec. (2.3). The vertical axis in this plot is electron density 19 3 (X10 m ). Data is from Ding et al. [2009]. The bottom two plots are electron temperature data taken with Thomson scattering (Sec. (2.2)), from Reusch [2011]. 14
Figure 1.9 Example of anomalous ion heating during a reconnection event on MST. The ion temperature data is core C+6 data from CHERS. This data is from Magee [2011].
The “anomalous” process is linked to magnetic reconnection. Magnetic field lines break apart and reconnect, changing the magnetic topology and converting magnetc energy into plasma en- ergy. This process plays an important role in the self-organization of fusion plasmas (in tokamaks, spheromaks and other devices, including the RFP), and also plays a key role in the dynamics of stars (including our sun), accretion disks and galaxies [Zweibel & Yamada, 2009]. It was noticed very early on that the level of anomalous ion heating in MST correlated with magnetic fluctuation activity [Scime et al., 1992]. Further information can be gathered by ob- serving sawtooth crashes, where ion temperature increases by 100% or more (see Fig. (1.9) for example data). Fiksel et al. [2009] determined that not only is the degree of ion heating dur- ing sawteeth a function of the magnitude of magnetic fluctuations, but it is also mass dependent ( E / E M 1/2). therm mag / i 15
1.5 Pulsed Poloidal Current Drive
The sawtooth events in MST degrade plasma confinement in two different ways [Chapman et al., 1996]. First, the increase in the strength of magnetic fluctuations drives radial transport. Second, the sawtooth crash itself drives a large injection of neutral particles, which leads to in- creased charge exchange losses. The low ( 1 ms) energy confinement times in standard MST plasmas led to a desire to suppress magnetic fluctuations to increase plasma confinement, driving the development of pulsed poloidal current drive (PPCD). Modeling by Ho [1991] and others suggested that an auxiliary parallel current driven at the plasma edge could flatten the profile and reduce the dominant magnetic fluctuations [Sovinec & Prager, 1996; Chapman, 1997]. This theory was put into practice on MST, where a fast current pulse is driven through the toroidal field winding to induce a poloidal electric field [Sarff et al., 1994]. This fluctuation reduction requires E (a) 0 [Chapman et al., 2002] where: k
E = E B/B =(E✓B✓ + E B )/B (1.13) k ·
This edge poloidal current drove magnetic fluctuation amplitudes down to record lows for MST, leading to a five-fold increase in energy confinement and record electron temperatures [Sarff et al., 1997]. With further optimization of PPCD in lower current plasmas, energy confinement of 10ms has been reached, a ten-fold improvement over standard plasmas [Chapman et al., 2002]. Representative data from a typical 400 kA PPCD shot is plotted in Fig. (1.10). The same signals from a standard (non-PPCD) 400 kA shot are plotted in comparison. In that particular PPCD shot, the enhanced confinement period begins at around 13 ms into the shot, and lasts until around 22 ms. During this period there is a dramatic reduction in magnetic fluctuations, with no relaxation events. An increase in electron density is also seen, due to the increase in particle confinement. The large decrease in magnetic fluctuations leads to a large decrease in energy diffusion (see Sec. (6.2) for data demonstrating this effect), driving an increase in core electron temperature. Core temperature data from a representative PPCD plot is shown in Fig. (1.11).
17
Figure 1.11 Core Te and core Ti measurements during 500 kA PPCD. The electron temperature is measured with soft x-ray (SXR), while the ion temperature data is central C+6 data collected with CHERS (Sec. (2.4)). Note the increase in core electron temperature, while core ion temperature stays flat. The top plot is of surface parallel electric field, which it should be noted is positive throughout the enhanced confinement period [Chapman et al., 2002].
Figure 1.12 Thomson Scattering (Sec. (2.2)) measurement of the electron temperature profile during enhanced confinement. Data is from a 380 kA PPCD discharge where the enhanced con- finement period ends at approximately 22.5 ms [Reusch, 2011].The plot on the right shows the time evolution of the electron temperature at four different radial points, with “Point: 0” being the one closest to the plasma core, and “Point: 19” being the closest to the plasma edge. 18
In Fig. (1.11), core electron temperature data was collected with the soft x-ray (SXR) diag- nostic. In this figure, CHERS data (Sec. (2.4)) was used to represent core ion temperature. This particular CHERS data was of central C+6 data, which over these time scales is an acceptable proxy for majority core ion temperature in MST (please see Sec. (2.4) and Reardon et al. [2003] for understanding why this assumption is a fair one). Core electron temperature can be seen increasing rapidly throughout the enhanced confinement period. At the same time, ion temperature at the core is not changing at all. The E (a) profile is k also shown in Fig. (1.11), where the fact that E (a) 0 throughout the enhanced confinement k period can be seen. With Thomson scattering, the full electron temperature profile throughout a PPCD shot is mea- sured with data from a representative 380 kA PPCD discharge plotted in Fig. (1.12). The rapid increase in core electron temperature can be seen clearly. Edge temperatures do not rise noticeably during PPCD, but the large change in magnetic fluctuations near the reversal surface allow a much larger mid-radius temperature gradient (see Sec. (6.2) for more on this topic). In this particular case, the enhanced confinement period ends at approximately 22.5 ms, at which point there is a very quick return to a standard temperature profile.
1.6 The RFP As A Fusion Device
The primary efforts to generate significant D-T fusion power in the laboratory so far in history have all involved tokamaks. As discussed in Sec. (1.1), the first effort to build a burning plasma experiment is currently underway in Cadarache, France. That reactor, ITER, is a tokamak. The tokamak has both advantages and disadvantages versus the RFP for generating fusion power. The tokamak has been studied for a longer period of time and on more machines than the RFP, which means that many of the key reactor components are more tested and understood. Auxiliary current drive and plasma heating are well-established, as is testing for divertors, plasma-facing components, wall materials, et cetera. Many of these issues are engineering tasks that simply must be done before a full-scale reactor can be built. That said, the biggest advantage that the tokamak 19 is perceived to have over the RFP is superior particle/energy confinement. Standard energy con- finement in MST is around 1 ms, compared to several hundred milliseconds for DIII-D, close to a full second for JET, and to a projected value of 3.7 seconds for ITER [ITER Physics Basis Editors, 1999; Shimada et al., 2007]. The stochastic nature of the RFP makes energy confinement very difficult. That said, the confinement gap between the RFP and the tokamak is not as large as it might appear at first glance. Energy confinement in tokamaks scales with basic plasma parameters. Toka- mak data scales very well with the IPB98(y,2) ELMy H-mode tokamak confinement scaling, which is [ITER Physics Basis Editors, 1999]:
0.93 0.15 0.69 0.41 0.19 1.97 0.58 0.78 ⌧E =(0.0562)I B P n M R ✏ (1.14)
The variables on the right side of Eq. (1.14) are, from left to right, toroidal plasma current, average toroidal magnetic field, loss power, line-averaged density, average ion mass, major radius, inverse aspect ratio and elongation. They are in units, where applicable, of seconds, MA, T, MW,
19 3 10 m , AMU and meters. Plugging in approximate values for MST to match the plasma current and magnetic field strength yields ”tokamak-like” energy confinement of only around 10 ms, which has been achieved in enhanced-confinement (PPCD) plasmas (Sec. (1.5)). In Fig. (1.13), MST standard and ”Improved” (i.e. PPCD) energy confinement times are com- pared to that standard tokamak scaling. It is clear in this chart both that ”tokamak-like” confine- ment has been achieved in MST, but also just how far the modern RFP is from the size and scale of the modern tokamak. Before getting to the attempts to scale confinement in the RFP, it’s necessary to introduce an important advantage of the RFP over tokamaks for generating fusion. The fact that so much of the magnetic field in the RFP is self-generated by the plasma means that the magnetic field strength needed at the magnetic coils is relatively small compared to the energy density in the plasma itself. This relationship is often described as the plasma :
Plasma Pressure nkBT = = 2 (1.15) Magnetic Pressure B /2µo 20
Figure 1.13 MST confinement compared with a scaled tokamak specified by the IPB98(y,2) ELMy-H mode empirical scaling [ITER Physics Basis Editors, 1999]. This specific figure, with the addition of MST data, is copied from Chapman et al. [2010].
Figure 1.14 MST confinement compared with Figure 1.15 Updated RFP confinement scaling modified Connor-Taylor scaling. This figure is using RFX-mod data. Plot is from Innocente et copied from Chapman et al. [2002]. The MST- al. [2009]. related dots are as follows: (a) MST 210 kA stan- dard, (b) MST 430 kA standard, (c) mid-1990s 340 kA PPCD, (d) circa-2000 390 kA PPCD, (e) circa-2000 210 kA PPCD. 21
In an engineering sense, magnetic pressure can be thought of as a proxy for machine cost, since the biggest operating cost of the reactor is the cost of generating the external magnetic fields. Plasma pressure is a proxy for fusion, since the rate of fusion reactions increases monotonically with increased plasma density and temperature. So can be seen as a proxy for fusions-per-dollar, or the financial efficiency of the machine. This makes the RFP very intriguing as a potential fusion reactor, should it be possible to keep sufficient plasma confinement. A of 26% has been measured on MST [Wyman et al., 2009], though no theoretical limit has yet been found and auxiliary heating could potentially increase further [Den Hartog et al., 2007]. 3/2 As a plasma heats up, its resistivity decreases (⌘ Te ). This means that a warmer plasma / 2 becomes increasingly difficult to heat ohmically (P⌦ ⌘j ). It has long been known that a / k tokamak must operate at an essentially ”collision-less” regime to generate sufficient fusion power, and so auxiliary heating is necessary [Golovin et al., 1970]. This conclusion led to the rapid development of the neutral beam injector (NBI) on a series of tokamaks in the early 1970s [Berry et al., 1975; Kelley et al., 1972; Menon, 1981]. The neutral beam has for decades now been the dominant source of auxiliary heating on tokamaks [Murakami et al., 1977; deGrassie et al., 2006]. A minimum of 33 MW of neutral beam power is being planned for ITER [Shimada et al., 2007]. For a more extended discussion of the history of neutral beams on tokamaks, see Sec. (3.1). The high achievable in the RFP means that a higher energy density is possible in the plasma core. This gives the RFP greater flexibility with its aspect ratio. It was found by Lawson [1977] that the RFP, if treated as a plasma column and if it had sufficiently small transport and impurity losses, could reach ignition with ohmic heating alone. Connor & Taylor [1984] theorized that confinement in the RFP was limited by resistive fluid turbulence near the reversal surface, and suggested that energy confinement could be scaled as ⌧ a2I3 /N 3/2, where N is the line- E / averaged density. Werley [1991] used that scaling to calculate that an RFP could reach ignition with ohmic heating alone with a plasma current as low as 8.1 MA. This scaling was further modified
2 3 3/2 with data existing in the mid-1990s to ⌧E =10.2(a I /N ) [Werley et al., 1996]. With this scaling in mind, an effort to design an RFP reactor (TITAN) began in 1986 [Na- jmabadi & the TITAN Research Group, 1988]. An introduction to the final report can be found 22 in Najmabadi et al. [1993]. A cutaway drawing of TITAN from that report is in Fig. (1.16). The final design for TITAN featured a major radius of 3.9 m and a minor radius of 0.6 m, as well as a plasma current of close to 18 MA. The design was theorized to produce approximately 2.3 GW of fusion power, and to generate a net electrical power of 0.97 GW. A ⌧E of approximately 150 ms was calculated. TITAN is massive compared to currently operating RFPs, but it is in fact quite small compard to ITER, which is already under construction. ITER has a major radius of 6.2 m, a minor radius of 2.0 m and a plasma current of 15-17 MA, yet will produce only 500-700 MW of fusion power [Shimada et al., 2007].
Figure 1.16 A cutaway drawing of the final TITAN design, from Najmabadi et al. [1993].
TITAN was just a design, of course, and construction has never been started on an RFP reactor. More importantly, modern RFP confinement data simply does not confirm this modified Connor- Taylor scaling. Fig. (1.14) is a plot from Chapman et al. [2002] showing confinement in MST versus this scaling. The improved confinement can be seen in the progress from standard MST plasmas to the early form of PPCD to the improved PPCD. With improved PPCD, MST actually has better confinement than is predicted by Connor-Taylor. However, this confinement does not scale as well with plasma current. More recent confinement scaling data, from RFX-mod, is plotted
sec 1 8 1,n 2 in Fig. (1.15), where br = 2 n= 15 bt (a) . ⇥ P ⇤ 23
Using modern plasma codes, improved confinement scaling for the RFP has been under- taken [Stoneking et al., 1998; Scheffel & Schnack, 2000; Scheffel & Dahlin, 2006]. None of these codes consistently match MST or RFX-mod data, however, which makes it particularly difficult to extrapolate to a reactor-sized device like TITAN. In addition to the improved , the RFP offers several other advantages for fusion over the tokamak. The fact that the field strength at the magnetic coils can be much smaller means that it is possible that superconducting coils will not be necessary. This, combined with the fact that the RFP aspect ratio is flexible and the possibility of ”single piece maintenance” (the entire first wall and blanket can be removed as a single piece), means that engineering would be much more simple for an RFP reactor than a tokamak. The concerns regarding an RFP reactor can be summed up in the word ”scale”. We simply cannot confirm at this point in time that energy confinement will scale to a reactor, and we do not know about the possibility of problems like disruptions (which are known to occur in high energy tokamaks and have been studied for many years) in a reactor-scale RFP. Because of the uncertainty with scale, it is high leverage research to press down the path previ- ously taken by tokamak researchers. Current drive, momentum drive and auxiliary heating in the RFP are very different than the tokamak. While these techniques are all well-studied in tokamaks, the surface has barely been scratched with the RFP. Auxiliary heating with an NBI is an example of this type of research. It is well-understood in tokamaks, but turns out to be quite different in the RFP. The study of auxiliary heating with an NBI on MST is the focus of this dissertation. The 25 kV fast ions studied in this dissertation are also specifically relevant for RFP fusion research. Assuming reasonable RFP reactor values of B 10 T,T 10kV, 0.1 [FES, T,CORE ⇡ e ⇡ ⇡ 2008], alpha particle Larmor orbits will be similar in radius to 25 kV hydrogen in MST (2-3 cm near the core, 10-15 cm near the edge). Further study of fast ions in MST could shed light on the dynamics of fusion products in a future RFP fusion device.
1.7 Thesis Overview
The key scientific advances covered in this thesis are: 24
Statistically significant auxiliary heating of an RFP has been measured for the first time. • A 1-D code has been developed to analyze the physics of this auxiliary heating in an RFP. • Output from the 1-D code demonstrates clearly that fast ion diffusion and mid-radius heating • are crucial to replicate the measured core Te in MST. Neutralization losses and resonant fast ion transport can be modeled. Neutralization losses • are crucial for non-PPCD plasmas. Resonant fast ion transport is found to not be significant for auxiliary heating. The remainder of this thesis is organized as follows. The measurement tools used to collect the data are discussed in Chp. (2). This chapter is a brief overview of MST diagnostics crucial to this research. The operation and physics of MST’s 1 MW NBI are discussed in Chp. (3). Classical fast ion dynamics and physics are discussed here, including both theory and data. Fast ion confinement (both classical and non-classical), slowing down, heating, momentum drive and current drive are all covered in this chapter. MST’s Rutherford scattering diagnostic is discussed in Chp. (4). Fast hydrogen ions from the NBI pollute the Rutherford analyzer enough to significantly impact the measured temperature of MST. The possibility that this physics is also affecting MST sawtooth Rutherford analysis is discussed. In Chp. (5), the heating data are presented. Statistically significant auxiliary core heating of MST is demonstrated in several different PPCD conditions. It is also shown that no significant auxiliary heating occurs in standard confinement plasmas (both F=0 and standard reversal). The latter result is due to the poor energy confinement discussed in Sec. (1.6). The 1-D model developed to model auxiliary heating in MST is discussed in Chp. (6). This is a classical 1-D model that has several layers of features. This 1-D model is a significant improvement over 0-D heating models that had been created for MST prior to the installation of the 1 MW NBI. The necessity of a 1-D model is seen in the discussion of fast ion diffusion and temperature gradients ( T ), which are found to be crucial for limiting heat losses from the MST core. This er e physics could not be detected with a 0-D model. Further, it will be shown why well-established 25 tokamak transport codes (such as TRANSP and NUBEAM) cannot currently be used to model these plasmas. The 1-D model includes fast ion deposition, classical slowing down and classical diffusion. This model is shown to be consistent with measured auxiliary heating in PPCD. To model non- PPCD MST discharges, neutralization losses are included and found to be crucial. In addition, resonant diffusion is modeled and found to be largely irrelevant to auxiliary heating. Chp. (7) is the concluding chapter of this thesis, which summarizes all of the results from the preceding chapters and also discusses possible future paths for fast ion research on MST. Finally, an appendix section (Appendix (A)) discusses the array of bolometers and XUV detec- tors on MST. This research is mostly incidental to the rest of the dissertation, but provides a useful reference to any future MST researchers that wish to use those diagnostics. 26
Chapter 2
Experiment Diagnostics
This chapter will discuss in brief the key diagnostics used for the research in this dissertation. The contributions of individual people to each diagnostic are cited as is appropriate.
2.1 Neutron Detection
Neutrons from MST are measured with a Bicron-408 plastic scintillator along with a photomul- tiplier tube (PMT). It is cylindrically-shaped, 127 mm in diameter and 127 mm in length, coupled with a 130 mm diamter photomultiplier tube [Fiksel et al., 2005]. It is a polymerized base of
22 3 polyvinyltouene doped with hydrocarbons to a density of n =5.23 10 cm [Magee, 2011]. H ⇥ The fusion that happens inside MST is predominantly D-D fusion, which has two branches that occur with approximately equal probabilities:
D + D T (1.01MeV)+p(3.01MeV) (2.1) ! He3(0.82MeV)+n(2.45MeV) (2.2) !
The cross section of a 2.45 MeV neutron with hydrogen is 2.5 barns, giving a mean free ⇡ path of =1/n 7.6 cm [Magee, 2011]. The scintillator dimensions were specifically mfp H ⇡ chosen so that most neutrons would undergo exactly one collision. The scintillator is sensitive to both neutrons and high energy photons, so it is surrounded in all dimensions by approximately 4 inches of lead, which is a good shield of x-rays while being nearly transparent to neutrons. The entire set-up is approximately 1 m from the outer wall of MST. 27
2.2 Thomson Scattering
The Thomson scattering diagnostic on MST measures electron temperature at multiple radial locations simultaneously, at multiple points in time. The physics of Thomson scattering was orig- inally discovered more than a century ago by JJ Thomson [Thomson, 1906], the same researcher who had discovered the electron just a few years prior. Thomson scattering consists of the absorp- tion and re-radiation of an electromagnetic photon by a free charged particle. This charged particle can be an ion or an electron, though for the case of the popular plasma diagnostic the charged particle measured is always the electron. The physical process is well described in the theses of previous MST graduate students who have focused on Thomson scattering, namely Den Hartog [1989]; Biewer [2002]; Stephens [2010] and Reusch [2011]. Those theses should be read for an in-depth description of the Thomson scattering process, which will just be reviewed in brief below. If we assume that the incident electron is non-relativistic (i.e. significantly slower than the photon), then the electron will only be significantly accelerated by the electric field of the photon (the acceleration due to the magnetic field will be small because the Lorentz force is a function of particle velocity). This electric field causes the electron to oscillate at the same frequency as the photon, and in the same direction as the photon. This generates dipole electromagnetic radiation with the same frequency as the absorbed photon in the reference frame of the electron. This radiation is scattered with a dipole radiation pattern. In Fig. (2.1), this is a sin( ) distribution, where is the angle between the direction of the incident photon and the direction of the scattered photon. On MST, the incoming radiation is polarized in the toroidal direction, and the collection lens is located at = 90o. The collected photon will be scattered at the same frequency of the incoming photon, in the reference frame of the electron. Assuming that the electron is in fact moving with some non- relativistic velocity (as is always the case in a laboratory plasma), the scattered photon’s frequency will be Doppler shifted. Assuming a Maxwellian distribution for the electrons in the plasma, a distribution of frequencies of scattered photons will be measured. The width of this spectrum correlates the electron temperature. 28
Figure 2.1 A cartoon of the observation of Thomson scattering, from Stephens [2010].
In practice on MST, the incoming photons come from two Spectron SL858 ND:YAG lasers that operate with a 1064 nm wavelength and a 9 ns FWHM pulse duration. The lasers can be operated in multiple modes, with each laser capable of firing 15 times during an MST shot. For most of the data in this dissertation, the lasers are fired 1 ms apart (1 kHz), giving 30 evenly-spaced time locations. Some of the data was taken when one of the two lasers was not operational, giving only 15 time points instead of 30. Scattered photons are measured at 22 radial locations. Because the Thomson scattering system is counting photons, the number of photons measured generates a statistical uncertainty. This uncertainty can be quite large for radial points near the edge of the plasma, where there is less scattering (P n ). Because this data is also often the measured / e least relevant for the research conducted in this thesis, those radial locations are often discarded (see Fig. (6.30) for an example). But since these uncertanties are purely statistical, they can be added up in quadrature. The effect of ensembling the data is to decrease the magnitude of these error bars.
2.3 Far-Infrared (FIR) Interferometer-Polarimeter System
Laser interferometry and polarimetry have long been used for a variety of physical purposes. The famous Michelson-Morley experiment, which failed in 1887 and paved the way for the Theory of Special Relatively, was an interferometry experiment. MST features a CO2 laser and 11 discrete 29 chords at approximately the same toroidal location (between 250o and 255o toroidal, on the typical MST coordinate system) [Brower et al., 2001]. The system features a time response of nearly 1 MHz, and a phase noise of only around 1o. A drawing of the system is in Fig. (2.2).
Figure 2.2 A cartoon drawing of MST’s FIR interferometer-polarimeter system, courtesy of Liang Lin.
This system can be run as an interferometer or a polarimeter in order to measure an array of equilibrium or fluctuation quantities, only two of which will be discussed here. First, as an inter- ferometer, the FIR system can produce an accurate, reliable measurement of the electron density profile. On MST, two parallel, linearly-polarized lasers can travel down two different paths: one through the plasma and a “reference” path that only passes through vacuum. At the end of those paths, the phase change in the laser across each path is then measured. Assuming that both paths have the same length, the change in phase is defined as [Ding et al., 2008]:
pa2 x2 '(x)=re ne(r)dz (2.3) pa2 x2 Z 15 In Eq. (2.3), r is the classical electron radius (2.82 10 m), is the laser wavelength, z e ⇥ is the distance along the chord, x is the impact parameter and x2 + z2 = r2. Because there are 11 30 chords, an Abel Inversion [Hutchinson, 2002] can be performed:
1 a @'(x) dx ne(r)= (2.4) 2 2 ⇡re r @x px r Z In practice, '(x) is measured for each of the 11 chords simultaenously. A numerical fit is made of the data. A spatial derivative is taken to calculate @'/@x, and then an inversion is performed to calculate an electron density profile. Throughout this thesis, electron density data will be presented in two ways. Line-averaged electron density is measured by a single-chord CO2 interferometer, while electron density profile data is measured with the FIR interferometer system. A second use of the FIR system for this dissertation is as a polarimeter to determine plasma current density measurements in the core. In this case, it is the polarization of the laser, rather than the phase, that is measured. In practice, the beam passing through the plasma and the reference beam are initially linearly polarized and orthogonal. Next, the beams are passed through a /4 wave plate, generating a counter-rotating circularly polarized beam [Brower et al., 2003]. Because the right-hand and left-hand circularly polarized waves pass through mediums with different refractive indices, the Faraday rotation angle ( ) measured will equal [Ding et al., 2003]:
2 3 2⇡ nR nL e = dz = ne(z)B (z)dz (2.5) 2 8⇡2c3✏ m2 k Z o e Z
In Eq. (2.5), B = B✓cos✓, where ✓ is the angle between the polarimeter chord and the poloidal k magnetic field. Near the magnetic axis, this term becomes localized to the plasma core since contributions to cos✓ along the chord only come from the plasma center. Brower et al. [2002] used Ampe`re’s Law to solve for the current density on axis:
2 d 1 J(0) = (2.6) µocF dx nef(r, ↵)dz
In Eq. (2.6), cF is a constant and f(r, ↵) representsR the current profile shape (IP = J(0) 2⇡rf(r, ↵)dr). This calculation of current density in the core was used to help determine whether the MSTR NBI is driving non-inductive plasma current. The data is presented in Sec. (3.7.2). 31
2.4 CHERS
Charge exchange recombination spectroscopy (CHERS) has been a popular diagnostic for mea- suring the velocity and tempurate of impurity ions in high temperature plasmas for decades [Isler & Murray, 1983; Fonck, 1985]. The basic physics of spectroscopy can be described by what hap- pens when two atoms exchange an electron. In the case of an MST plasma, imagine that a neutral hydrogen exchanges an electron with an impurity ion with atomic number A and a charge +q. The electron typically is absorbed at a high energy state. When it decays to a lower energy state the ion releases a photon:
0 +q + (+q 1) H + A H + A ⇤ (2.7) ! (+q 1) (+q 1) A ⇤ A + (2.8) !
This same physical process describes interactions between impurity ions and both background neutrals and electrons in a plasma, and can be used for a variety of different ions. Common impurities in high temperature laboratory plasmas like MST are nitrogen, oxygen, boron, carbon and aluminum. CHERS has been used on MST to measure a variety of different impurities, with the most common being C+6. The energy of the photon emitted is a function of the energy state change and the impurity type. For example, the n=7-6 transition for C+6 releases a photon with a wavelength of 343.6 nm, and with a de-excitation time of approximately 90 ns [Magee, 2011]. In a plasma, the magnitude and wavelength of the photons emitted are influenced by the ion density, flow velocity and temperature. The wavelength of the emitted photon is known for the reference frame of the ion. If the ion has a velocity relative to the observer then that photon will be Doppler shifted. And assuming that a population of ions can be described with a Maxwellian distribution, there will be a Maxwellian distribution of photons, with the width of that distribution correlating with the temperature of the ions. A greater density of impurity ions will increase the number of emitted photons as well. This effect can be seen in Fig. (2.3), where the signal for the n=7-6 transition for C+6 is plotted for varying temperatures, flow velocities and densities for a Maxwellian distribution of particles. 32
Figure 2.3 Simulation of the effect of ion temperature (left), flow velocity (center) and density (right) on the measured distribution of photons for the n=7-6 transition for C+6, from Magee [2011].
This process can be stimulated with a neutral beam in a process typically called “active-beam spectroscopy,” which offers several benefits over passive spectroscopy [Isler, 1994]. The neutral beam provides signal levels much larger than the background signal, and at wavelengths long enough for accurate measurement of the Doppler broadening and shift. In addition, the neutral beam allows for local measurements when used with an array of viewing chords. Each chord will see the signal from interactions only in the local volume encompassed by both the viewing chord and the neutral beam. The background signal can be measured with a passive chord (that sees a similar volume of plasma but does not intersect the beam), and then subtracted out in order to calculate the active signal accurately. On MST, the neutral beam used for CHERS is a 45 kV hydrogen beam. The voltage is chosen because the maximum charge exchange reaction rate (<