on the exploration of innovative concepts for fusion … publications/2001/abdou...the presentation...

Fusion Engineering and Design 54 (2001) 181–247

On the exploration of innovative concepts for fusionchamber technology

M.A. Abdou a,*, The APEX TEAM, A. Ying a, N. Morley a, K. Gulec a,S. Smolentsev a, M. Kotschenreuther b, S. Malang c, S. Zinkle d, T. Rognlien e,

P. Fogarty d, B. Nelson d, R. Nygren f, K. McCarthy g, M.Z. Youssef a,N. Ghoniem a, D. Sze h, C. Wong i, M. Sawan j, H. Khater j, R. Woolley k,

R. Mattas h, R. Moir e, S. Sharafat a, J. Brooks h, A. Hassanein h, D. Petti g,M. Tillack l, M. Ulrickson f, T. Uchimoto m

a Mechanical and Aerospace Engineering Department, Uni6ersity of California-Los Angeles, 44–114 Engineering IV,420 Westwood Plaza, Los Angeles, CA 90095, USA

b Uni6ersity of Texas, Austin, TX, USAc Forschungszentrum Karlsruhe, Karlsruhe, Germany

d Oak Ridge National Laboratory, Oak Ridge, TN, USAe Lawrence Li6ermore National Laboratory, Li6ermore, CA, USA

f Sandia National Laboratory, Albuquerque, NM, USAg Idaho National Engineering and En6ironmental Laboratory, Idaho Falls, ID, USA

h Argonne National Laboratory, Argonne, IL, USAi General Atomics, San Diego, CA, USA

j Uni6ersity of Wisconsin, Madison, WI, USAk Princeton Plasma Physics Laboratory, Princeton, NJ, USA

l Uni6ersity of California-San Diego, La Jolla, CA, USAm Uni6ersity of Tokyo, Tokyo, Japan

Received 10 March 2000; accepted 7 June 2000

Abstract

This study, called APEX, is exploring novel concepts for fusion chamber technology that can substantially improvethe attractiveness of fusion energy systems. The emphasis of the study is on fundamental understanding andadvancing the underlying engineering sciences, integration of the physics and engineering requirements, andenhancing innovation for the chamber technology components surrounding the plasma. The chamber technologygoals in APEX include: (1) high power density capability with neutron wall load \10 MW/m2 and surface heat flux\2 MW/m2, (2) high power conversion efficiency (\40%), (3) high availability, and (4) simple technological andmaterial constraints. Two classes of innovative concepts have emerged that offer great promise and deserve further

www.elsevier.com/locate/fusengdes

* Corresponding author. Tel.: +1-310-2060501; fax: +1-310-8252599.E-mail address: [email protected] (M.A. Abdou).

0920-3796/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.

PII: S0920-3796(00)00433-6

M.A. Abdou et al. / Fusion Engineering and Design 54 (2001) 181–247182

research and development. The first class seeks to eliminate the solid ‘‘bare’’ first wall by flowing liquids facing theplasma. This liquid wall idea evolved during the APEX study into a number of concepts based on: (a) using liquidmetals (Li or Sn–Li) or a molten salt (Flibe) as the working liquid, (b) utilizing electromagnetic, inertial and/or othertypes of forces to restrain the liquid against a backing wall and control the hydrodynamic flow configurations, and(c) employing a thin (�2 cm) or thick (�40 cm) liquid layer to remove the surface heat flux and attenuate theneutrons. These liquid wall concepts have some common features but also have widely different issues and merits.Some of the attractive features of liquid walls include the potential for: (1) high power density capability; (2) higherplasma b and stable physics regimes if liquid metals are used; (3) increased disruption survivability; (4) reducedvolume of radioactive waste; (5) reduced radiation damage in structural materials; and (6) higher availability.Analyses show that not all of these potential advantages may be realized simultaneously in a single concept. However,the realization of only a subset of these advantages will result in remarkable progress toward attractive fusion energysystems. Of the many scientific and engineering issues for liquid walls, the most important are: (1) plasma–liquidinteractions including both plasma–liquid surface and liquid wall–bulk plasma interactions; (2) hydrodynamic flowconfiguration control in complex geometries including penetrations; and (3) heat transfer at free surface andtemperature control. The second class of concepts focuses on ideas for extending the capabilities, particularly thepower density and operating temperature limits, of solid first walls. The most promising idea, called EVOLVE, isbased on the use of a high-temperature refractory alloy (e.g. W–5% Re) with an innovative cooling scheme based onthe use of the heat of vaporization of lithium. Calculations show that an evaporative system with Li at �1 200°C canremove the goal heat loads and result in a high power conversion efficiency. The vapor operating pressure is low,resulting in a very low operating stress in the structure. In addition, the lithium flow rate is about a factor of ten lowerthan that required for traditional self-cooled first wall/blanket concepts. Therefore, insulator coatings are notrequired. Key issues for EVOLVE include: (1) two-phase heat transfer and transport including MHD effects; (2)feasibility of fabricating entire blanket segments of W alloys; and (3) the effect of neutron irradiation on W. © 2001Elsevier Science B.V. All rights reserved.

Keywords: Chamber technology; First wall; Blanket; Liquid walls; Free surface; Refractory alloys; Two-phase flow; Plasma–materialinteraction; MHD effects

1. Introduction

A study, called APEX, was initiated in early1998 as part of the US Fusion Energy SciencesProgram initiative to encourage innovation andscientific understanding. The primary objective ofAPEX is to identify and explore novel, possiblyrevolutionary, concepts for the chamber technol-ogy that can substantially improve the attractive-ness of fusion energy systems. The chambertechnology includes the components in the imme-diate exterior of the plasma (i.e. first wall, blan-ket, divertor, and vacuum vessel) and has atremendous impact on the economic, safety andenvironmental attractiveness of fusion energysystems.

The APEX study is being carried out by amulti-disciplinary, multi-institution integratedteam. The emphasis of the study has been on

fundamental understanding and advancing theunderlying engineering sciences, integration of thephysics and engineering requirements, and en-hancing innovation for the chamber technology.

This paper presents a summary of the technicalresults from the first phase of the APEX study. Anumber of promising ideas for new innovativeconcepts have already emerged from this firstphase. While these ideas need extensive researchbefore they can be formulated into mature designconcepts, some of them offer great promise forfundamental improvements in the vision for anattractive fusion energy system.

These ideas fall into two categories. The firstcategory seeks to totally eliminate the solid ‘bare’first wall. The most promising idea in this cate-gory is a flowing liquid wall. The liquid wall ideais ‘concept rich’. These concepts vary from ‘liquidfirst walls’, where a thin layer of liquid (B2 cm)

M.A. Abdou et al. / Fusion Engineering and Design 54 (2001) 181–247 183

flows on the plasma-side of the first wall, to ‘thickliquid walls’, where an all-flowing thick liquid(\40 cm) serves as the liquid wall/liquid blanket.Other variations in the liquid wall concepts in-clude the type of ‘restraining force’ utilized to‘control’ the movement and geometry of the liq-uid. Candidate liquids range from high conductiv-ity, low Prandtl number liquid metals to lowconductivity, high Prandtl number liquids such asthe molten salt Flibe. While all concepts in theliquid wall category share some common advan-tages and issues, each concept has its own uniqueset of incentives and issues.

The second category of ideas focuses on extend-ing the capabilities, particularly the power densityand temperature limits, of solid first walls. Apromising example is the use of high temperaturerefractory alloys (e.g. tungsten) in the first walltogether with an innovative heat transfer and heattransport scheme based on vaporization oflithium.

This paper is organized as follows. Section 2summarizes the study approach. Section 3 pro-vides an introduction to the basic scientific princi-ples of liquid walls. Sections 4 and 5 present theanalysis of free-surface hydrodynamics, heattransfer, and MHD effects, as well as tritiumbreeding, activation, and other considerations forthick and thin liquid walls, respectively. A liquidwall concept based on the utilization of electro-magentic forces to restrain the liquid flow move-ment and geometry is introduced in Section 6.Plasma–liquid interactions are addressed in Sec-tions 7 and 8. Concepts based on using high-tem-perature refractory solid first walls are analyzed inSection 9 for evaporative lithium cooling and inSection 10 for helium cooling. Section 11 high-lights a concept for flowing Li2O particulates.Structural material and safety considerations forall concepts are investigated in Sections 12 and 13,respectively. A summary of the study is providedin Section 14. It should be noted that because ofthe depth and breadth of the study, the scope ofthe presentation in this paper has been limited tothe key technical points. Considerable additionaldetails are provided in the study ‘report’ of ref.[1].

Table 1Highlights of the APEX study approach

Emphasize innovation(1)(2) Understand and advance the underlying engineeringsciences(3) Utilize a multidisciplinary, multi-institution integratedTEAM(4) Provide for open competitive solicitations

Close coupling to the plasma community(5)(6) Direct participation of material scientists and systemdesign groups

Direct coupling to IFE Chamber Technology community(7)(8) Encourage international collaboration

2. Study approach

The APEX objective, ‘to identify and explorenovel, possibly revolutionary, concepts for theChamber Technology that can substantially im-prove the attractiveness of future fusion energysystems’, represents a challenge that was clearlyrecognized from the onset of the study. Therefore,careful attention was paid to the study approach.Some of the key elements of the APEX approachare highlighted in Table 1.

Chamber Technology includes the componentsin the immediate exterior of the plasma (e.g. firstwall, blanket, divertor, and vacuum vessel). Con-cepts for Chamber Technology must satisfy thebasic functional requirements shown in Table 2,which include providing a vacuum environment,plasma exhaust, heat removal, and tritiumbreeding.

A set of primary goals for Chamber Technol-ogy was adopted to guide the APEX study. These

Table 2Functional requirements of chamber technology

Provision of vacuum environmentExhaust of plasma burn productsHeat removal and power extraction of surface heat loads

(from plasma particles and radiation)Heat removal and power extraction of bulk heating (from

energy deposition of neutrons and secondary gammarays)

Tritium breeding at the rate required to satisfyself-sufficiency

Radiation protection


Table 3Primary goals for chamber technologya

High power density capability (main driver)1.Neutron wall load \10 MW/m2

Surface heat flux \2 MW/m2

High power conversion efficiency (\40%)2.3. High availability

Lower failure rate (MTBF\43 MTTR)Faster maintenance (MTTRB0.023 MTBF)

4. Simpler technological and material constraints

a Goals used to calibrate new ideas and measure progress.

During the work on exploring novel ideas, theteam adopted a set of scientific evaluation criteriawhich are discussed in ref. [1]. These criteriaincluded:1. Satisfying functional requirements (see Table

2).2. Demonstrating potential for improved attrac-

tiveness, based on: (a) high power density ca-pability; (b) high conversion efficiency; (c) highavailability; (d) attractive safety and environ-mental attributes; (e) simple technological andmaterial constraints, and (f) low cost.

It is important to note that the process flowdiagram in Fig. 1 was not intended as a ‘rigid’sequence of events. Rather, it was only a ‘guide’to measure progress and a tool to focus resourceson ideas with better potential. Strong technicaljudgement by the scientists was the best guidancewhenever new, and often surprising, technical re-sults were obtained. Innovation was needed, andhas taken place as an ongoing process. For exam-ple, when technical results indicated that the tem-perature of a free-surface liquid wall may belimited by plasma impurity considerations, thefollowing innovative ideas were proposed by teamscientists:1. The use of Sn–Li because it has low vapor

pressure at elevated temperatures.2. Effective schemes to promote controlled sur-

face mixing and wave formation to eliminatethe surface thermal boundary layer.

3. Novel ideas for two-stream flows that keep thefree-surface temperature low enough for com-patibility with plasma operation while the bulkliquid temperature is sufficiently high for at-tractive energy conversion.

3. Introduction to liquid walls

The idea of flowing liquid walls has emerged asone of the most promising concepts explored sofar in APEX. The area of liquid walls appears tobe ‘concept rich’ with many ideas emerging in thepast 2 years that have widely different characteris-tics. Therefore, an introduction to liquid walls isnecessary before presenting the technical results ofthe next five sections.

goals, shown in Table 3, have been used as guide-lines to calibrate the potential attractiveness ofnew ideas and to measure progress. These goalsprovide quantitative targets for key parametersand features related to Chamber Technology thathave the highest impact on the attractiveness offusion energy systems. The rationale for thesegoals is provided in Ref. [2].

In the early stage of the APEX study, an assess-ment was conducted to understand the limitationsand constraints of traditional concepts (i.e. con-cepts that were developed over the past 20 years).The results of this assessment [2] are not repeatedhere. By understanding the limitations and con-straints of the traditional concepts, this assess-ment partially illuminated the path towardextending limits, overcoming constraints, andhelped stimulate ideas for potentially more attrac-tive novel concepts.

A diagram illustrating the APEX process forscreening and evaluating the scientific bases ofnew ideas is given in Fig. 1. These ideas wentthrough a ‘screening process’ which relied on ‘ex-pert judgement’ by the APEX team. The teamtolerated ‘high-risk’ ideas whenever there was aclear potential for high-payoff. The ideas thatpassed the screening test proceeded to the stage of‘design idea formulation and analysis using exist-ing tools’. The technical work on those ideas isreported in ref. [1]. However, it should be notedthat in the course of this work it became evidentthat existing models and data were not sufficient.Therefore, substantial effort was devoted to devel-oping new models and exploring new phenomenafor the more promising concepts such as liquidwalls and high-temperature refractory alloys.


Fig. 1. Flow diagram illustrating steps in the APEX process.


It must be clearly noted here that the conceptof liquid walls is an idea that, prior to APEX,has not been subjected to extensive analysis andevaluation. A brief history is in order. The ideaof using a liquid blanket in a fusion device wasfirst suggested by Christofilos in 1970 [3] for afield reversed concept (FRC). In this design, theplasma volume was surrounded by a 75 cmthick, free surface lithium blanket flowing at 30m/s−1. Subsequent uses of the liquid walls formagnetic fusion devices have been documented[4–7].

With regards to inertial fusion reactors, thefirst published reactor design concept was a liq-uid wall concept proposed by Fraas of ORNL[8]. This design, called BLASCON, features acavity formed by a vortex in a rotating liquid-lithium pool. Subsequent liquid wall design con-cepts include a liquid-lithium waterfall [9],HYLIFE [10], HYLIFE-I [11], and HYLIFE-II[12].

3.1. Liquid wall options

The liquid wall idea evolved during the APEXstudy into a number of concepts that have somecommon features but also have widely differentissues and merits. These concepts can be

classified (as shown in Table 4) according to: (a)thickness of the liquid; (b) type of liquid used;and (c) the type of restraining force used tocontrol the liquid flow (i.e. adhere to a backingwall).

The working liquid must be a lithium-contain-ing medium in order to provide adequate tri-tium. The only such practical candidates are theliquid metals lithium and Sn–Li, and the moltensalt Flibe. Lead–lithium was eliminated as acandidate early in the study [1,2]. Lithium andFlibe were considered for traditional conceptsfor many years. Sn–Li was introduced intoAPEX because it has very low vapor pressure atelevated temperature, which is an important ad-vantage in a plasma-facing flowing liquid. Thehydrodynamics and heat transfer related charac-teristics and issues of high-conductivity, lowPrandtl Number liquid metal flows are consider-ably different from those of the low-conductiv-ity, high Prandtl Number Flibe flows. Flowingliquid metals may require the use of electricalinsulators to overcome the MHD drag, whilefor Flibe free surface flows, MHD effects causedby the interaction with the mean flow are lesssignificant. The effects on plasma stability andconfinement also differ based on the electricalconductivity of the working liquid.

The thickness selected for the liquid wall layerflow directly facing the plasma and in front of asolid ‘backing wall’ leads to different conceptsthat have some common issues but many uniqueadvantages and challenges. Both thin and thickliquid walls can adequately remove high surfaceheat flux. A primary difference between thin andthick liquid walls is the magnitude of attenua-tion of neutrons in the liquid before they reachthe backing wall. As seen later, the ‘thin’ liquidwall concept is easier to attain, but ‘thick’ liquidwall concepts greatly reduce radiation damageand activation of the structure behind the liquid.

Widely different liquid wall concepts are ob-tained by applying various forces to drive theliquid flow and restrain it against a backingsolid wall. As shown in Table 4, at least fourconcepts can be considered based on the driv-ing/restraining force scheme. The first is calledgravity–momentum driven (GMD). In the

Table 4Liquid wall options

Thin (�2 cm)ThicknessModerately thick (�15 cm)Thick (\40 cm)

Working liquid LithiumSn–LiFlibe

Hydrodynamic Gravity–momentum driven(GMD)driving/restraining

forceGMD with swirl flowElectromagnetically restrainedMagnetic propulsionSingle, contiguous, streamLiquid structureTwo streams (fast flowing thinlayer on the plasma side andslowly flowing bulk stream)


GMD concept, illustrated in Fig. 2, the liquid isinjected at the top of the chamber with an angletangential to the curved backing wall. The fluidadheres to the backing wall by means of cen-trifugal force and is collected and drained at thebottom of the chamber. The criterion for thecontinuous attachment of the liquid layer is sim-ply that the centrifugal force pushing the liquidlayer towards the wall is greater than the gravi-tational force.

A GMD with the swirl flow concept is ob-tained by giving the fluid an azimuthal velocityto produce rotation. The ‘swirl flow’ results in asubstantial increase in the centrifugal accelera-tion towards the back wall and better adherenceto the wall, when the backing wall curvature inthe poloidal direction is large and the toroidalcurvature is comparable to poloidal curvature.While swirl flow may not be needed for moder-ate aspect ratio tokamaks, it appears to be nec-essary in the highly elongated, very low aspectratio spherical torus (ST).

In APEX, the GMD has been explored fortokamaks. The GMD with the swirl flow con-cept has been investigated for both the ST andFRC tokamaks. Other plasma confinementschemes will be investigated in the future. Theanalysis of the GMD and GMD with swirl flowconcepts are described in more detail in Sections4 and 5.

The electromagnetically restrained (EMR)concept, illustrated schematically in Fig. 3, isapplicable only to liquid metals. The EMR con-cept has been explored only for lithium in toka-mak configurations. In the EMR, a force fieldpushing the liquid against the backing wall isgenerated by injecting current to flow throughthe liquid lithium in the poloidal direction. Theinjected poloidal current interacts with thetoroidal magnetic field to generate an internal‘J×B’ body force causing the liquid layer toadhere to the back wall. The EMR concept isexplored in Section 6.

Another liquid wall concept not yet exploredin APEX is the magnetic propulsion idea pro-posed by Zakharov [13] and illustrated in Fig.

4. The idea is to create a pressure driving forcethrough the interaction of the toroidal magneticfield with an externally applied longitudinal elec-trical current in the liquid metal layer. The non-uniformity of the toroidal magnetic field thusgenerates a non-uniform Lorentz force. The re-sultant pressure gradient or propulsion effectcauses the flow to accelerate from the inboard,where the magnetic field is stronger, to the out-board region. In addition, the Lorentz forceprovides an active feedback mechanism for sta-bilizing the flow while its normal componentkeeps the layer adhered to the structural wall.

3.2. Moti6ation for liquid wall research

Liquid walls offer many potential advantagesthat represent an excellent opportunity to sub-stantially enhance the attractiveness of fusionenergy systems. Examples of advantages thatmay be realized if we can develop good liquidwall designs are listed in Table 5. The potentialimprovements in plasma stability and confi-nement are analyzed in Section 8. The other po-tential advantages are discussed in Sections 4–6.

As explained earlier, there are many optionsfor liquid wall concepts. It is not clear yet thatall these advantages can be realized simulta-neously in a single concept. However, the real-ization of only a subset of these advantages willresult in remarkable progress toward the attrac-tiveness of fusion energy systems.

3.3. Key issues for liquid walls

The scientific and engineering issues for liquidwalls are addressed in Sections 4–8. Of all theseissues, a number of scientific issues stand out asthe highest priority for near-term liquid wall re-search, which are summarized in Table 6. Theseinclude:1. Plasma–liquid interactions including both

plasma–liquid surface and liquid wall–bulkplasma interactions. Plasma stability andtransport may be seriously affected and poten-tially improved through various mechanismsincluding control field penetration, H/Hepumping, passive stabilization, etc.


Fig. 2. Illustration of principles of gravity–momentum driven (GMD) liquid wall concept (Vb =fluid velocity, g� =gravitationalacceleration, Rc=radius of curvature). Liquid adherence to back wall by centrifugal force. Applicable to liquid metals or moltensalts.Fig. 3. Illustration of principles in electromagnetically restrained (EMR) liquid metal wall concept (Bb =magnetic field, Jb = inducedcurrent, Fb =electromagnetic force, Vb =fluid velocity). Externally driven current (J) through the liquid stream. Liquid adheres to thewall by EM force Fb =Jb ×Bb .Fig. 4. Illustration of basic principles of magnetic propulsion liquid metal wall concept (Bb =magnetic field, Jb = induced current,Fb =electromagnetic force, Vb =fluid velocity). Adheres to the wall by Fb =Jb ×Bb . Utilizes 1/R variation in Fb =Jb ×Bb to drive theliquid metal from inboard to the outboard.


Table 5Motivation for liquid wall research

What may be realized if we can de6elop good liquid walldesigns:

� Improvements in plasma stability and confinement-Enable high b, stable physics regimes if liquid metals

are used

� High power density capability-Eliminate thermal stress and wall erosion as limiting

factors-Smaller and lower cost components (chambers, shield,

vacuum vessel, magnets)

� Increased potential for disruption survivability

� Reduced volume of radioactive waste

� Reduced radiation damage in structural materials-Makes difficult structural material problems more

tractable� Potential for higher availability

-Increased lifetime and reduced failure rates-Faster maintenance (design-dependent)

surfaces in contact with the LM layer. Eddycurrent forces perpendicular to the surface canpull the LM off the surface, even when completeaxi-symmetry is assumed in the toroidal direc-tion. Additionally, gradients in toroidal fieldcan exert a significant drag on the free surfaceflow. For thick liquid walls, the main issueconcerns the formation and removal of theliquid flow in the plasma chamber, and theaccommodation of penetrations.

3. Heat transfer at free surface and temperaturecontrol. Liquid surface temperature and vapor-ization is a critical, tightly coupled problembetween plasma edge and liquid free surfaceconditions including radiation spectrum, sur-face deformation, velocity and turbulent char-acteristics. Being a low thermally conductingmedium, the Flibe surface temperature highlydepends on the extent of the turbulent convec-tion. However, the normal velocity at the freesurface as well as the turbulent eddy near thesurface can be greatly suppressed. A greaterdegradation in heat transfer (up to 50%) wouldbe expected for the Flibe thick liquid concepts.The heat transfer at free surface issue is an evenmore serious concern, as the current limit onsurface temperature for Flibe, as estimated bythe plasma interface group, is significantly low.

The effects of liquid walls on the plasma core aswell as edge plasma–liquid surface interactionsrequire modeling and experiments in plasmadevices. Free-surface fluid flow and heat transfer,with and without magnetic field and hydrodynamiccontrol of free-surface flow in complex geometriesrequire modeling and laboratory experiments. It isworth noting that a number of such importantmodelling and experimental R&D activities havealready started as part of APEX and as part of theUS Liquid Wall Research Program.

4. Thick liquid wall concepts

4.1. Introduction

The replacement of the first wall with a flowingthick liquid offers the potential advantages of highpower density, high reliability and availability (due

2. Hydrodynamics flow feasibility in the complexgeometry including penetrations needed forplasma maintenance. The issue of establishinga viable hydrodynamic configuration threatensfeasibility, while it differs significantly for thickversus thin and for molten salts versus liquidmetals. The main issue facing liquid metals is ofcourse that of MHD interaction. Withouttoroidal axi-symmetry of the flow and field,reliable insulator coatings will be required on all

Table 6Key scientific issues for liquid walls

� Effects of liquid walls on core plasma including:-Discharge evolution (start-up, fueling, transport,

beneficial effects of low recycling)-Plasma stability including beneficial effects of

conducting shell and flow

� Edge Plasma–liquid surface interactions

� Free-surface heat transfer and turbulence modificationsat and near free-surfaces

� MHD effects on free-surface flow for low- andhigh-conductivity fluids

� Hydrodynamic control of free-surface flow in complexgeometries, including penetrations, submerged walls,inverted surfaces, etc.


to simplicity and low failure rates), reduced vol-umes of radioactive waste, and increased structurelifetime. All these advantages make the thick liquidwall approach a strong candidate in the APEXstudy. Specifically, neutronics analyses showed thatwith �42 cm layer thickness, about two orders ofmagnitude reduction in helium and hydrogen pro-duction is achieved with either Flibe or Sn–Li.With this thickness, and a 200 DPA damage limitfor structure replacement, the use of Flibe or Sn–Lican make the structure behind it a lifetime compo-nent. Furthermore, the volumes of radioactivewaste from the FW/blanket system, as well as fromthe entire system, are substantially reduced. Theemphases of the Phase I study included the explo-ration of design ideas, quantification of their highpower density capabilities, and identification andanalyses of the key feasibility issues of thick liquidwall configurations for various confinementschemes. The initial goal is to establish a viable freesurface flow configuration. This involves: (1) aninlet nozzle and penetrations that pass flow withoutdripping or splashing; (2) a free surface flow sectionthat allows liquid to cross temporally and spatiallyvariable magnetic fields and provides full wallcoverage; and (3) a head recovery nozzle systemthat accepts the flow and converts it from a freesurface flow to a channel (pipe) flow withoutcomplete loss of kinetic energy.

There are three lithium-containing candidateliquids for the walls: (1) Flibe — a good neutronabsorber and low electrically conducting mediumof molten salt; (2) lithium — a low Z material thatis more likely compatible with the plasma opera-tion; and (3) tin–lithium — an extremely low vaporpressure fluid at elevated temperatures. Bothlithium and tin–lithium are good electric conduc-tors. Utilization of these two liquid metals willrequire an understanding of MHD effects, not justin the surface flows, but in supply lines and feedsystems, and it also would likely require electricallyinsulating coatings.

4.2. Design options

Design ideas for establishing thick liquid wallswere addressed for the tokamak (such as ARIES-RS) [14], spherical torus (ST), and field reverse

configuration (FRC). The fact that topologies aredifferent in different confinement schemes requiresdifferent liquid wall design approaches. For exam-ple, as compared to the ARIES-RS, the ST confi-nement scheme tends to be highly elongated (largerback wall radius of curvature). The centrifugalforce acting on the liquid layer due to its poloidalmotion is less than in the ARIES-RS. However, ituses a smaller toroidal radius as illustrated in Fig.5. The centrifugal force acting on the flowing liquidlayer in the ST configuration can be increased byutilizing swirl motion in the azimuthal (toroidal)direction. Thus, one may expect a more stablehydrodynamics condition in the ST liquid blanket.However, being highly elongated, the fluid takesmore time to travel through the reactor if only onecoolant stream is used. This implies that the freesurface side may be overheated from a long plasmaexposure time. A typical FRC reactor can beviewed as a long cylinder in which a football shapedvolume of plasma lies at the center of the reactorchamber (see Fig. 6). The FRC confinement schemeappears more amenable to thick liquid walls due toits geometrical simplicity and lower strength mag-netic fields.

4.3. Hydrodynamics and MHD effects

One of the most fundamental issues for the thickliquid blanket is how to form, establish, andmaintain a thick liquid flow in a fusion reactor suchas the ARIES-RS (as shown in Fig. 5). The simplestapproach that can be conceived for a thick liquidblanket is free-flowing liquid under the effect ofgravitational and inertial forces. As illustrated inFig. 7, the thick liquid layer is injected at the topof the reactor chamber with an angle tangential tothe backing structural wall. As it flows along thecurved wall the fluid adheres to the structural wallby means of centrifugal and inertial forces. It thenis collected and drained at the bottom of thereactor.

This Flibe approach has been modeled with athree-dimensional, time-dependent Navier–StokesSolver that uses Reynolds Averaged Navier Stokes(RANS) equations for turbulence modeling and thevolume of fluid (VOF) free surface tracking al-gorithm for free surface incompressible fluid flows.


Fig

.5.

Hyd

rody

nam

icco

nfigu

rati

ons

ofth

ick

liqui

dw

alls

may

beve

rydi

ffer

ent

for

AR

IES-

RS

and

AR

IES-

ST.

Bot

hco

nfigu

rati

ons

are

conv

erte

dto

sing

lenu

llat

the

bott

omof

the

plas

ma,

inor

der

tobe

com

pati

ble

wit

hth

eliq

uid

wal

lco

nfigu

rati

on.


Fig. 6. General layout of a FRC power plant design.Fig. 7. A thick FW/blanket design incorporated into the ARIES-RS configuration.


Example solutions, as shown in Fig. 8, demon-strate that stable, thick fluid configurations can beestablished and maintained throughout a toka-mak reactor configuration. Nevertheless, the flowcontinuity requires that some amount of thinningresult from the gravitational acceleration and flowarea expansions as the flow proceeds downstream.As shown, the fluid thickness is reduced by abouta factor of 2 at the reactor midplane for an initialvelocity of 8 m/s. This thinning reduces the liq-uid’s potential for radiation protection of solidwalls behind the liquid and creates an unfavorablesituation for shielding. The jet thinning effect canbe overcome by increasing the initial jet velocity;and a fairly uniform thick liquid film can beobtained throughout the plasma core if the jet isinjected at velocities of 15 m/s or above (as shownin part b of Fig. 8).

The thinning effect resulting from the gravita-tional acceleration can be minimized by the MHDdrag from the Hartmann velocity profile in aliquid metal flow. Numerical analyses were per-formed to determine whether or not an insulatoris needed for free surface MHD flows, and todefine lithium’s initial velocity that enables a uni-form thickness to be maintained throughout theplasma chamber in the presence of the toroidalmagnetic field. The preliminary analysis showsthat the MHD drag effect significantly increasesthe layer thickness and causes the associated re-duction in the velocity. Thus, there is a need ofinsulators for a free liquid metal flow if a seg-mented toroidal liquid metal flow configuration isconsidered. As shown in Fig. 9, for an insulatedopen channel, calculations indicate that a uniform40 cm-thick lithium layer can be maintained alongthe poloidal path at a velocity of 10 m/s. At thisvelocity, the total pressure (dynamic and static)exerted on the backplate is about 4800 N/m2. Thismagnitude of pressure can be easily managed. In

contrast, if the side-walls of the channel are sub-merged under the fluid, a fast surface layer canform naturally. The liquid near the surface, abovethe submerged insulated side-walls, will be unfet-tered by MHD drag except in areas close topenetrations, and a thin, fast layer at the surfacewill result (see Fig. 10).

The influence of the conducting backplate onthe liquid metal flow characteristics is negligible inthe presence of a purely toroidal magnetic field.However, the MHD drag can be significant ifthere is a radial magnetic field component — onenormal to the free surface. Analyses indicate thata metallic backplate is acceptable with insulatedtoroidal breaks if the radial magnetic field is nomore than 0.1–0.15 T. This magnitude woulddrop to 0.015 T for the case of toroidally continu-ous flow. Other MHD issues such as flow acrossfield gradients (1/R dependence of the toroidalfield for example), temporal fluctuations duringstart-up and plasma control, have yet to be ad-dressed adequately — ongoing work is in pro-gress to address these important concerns.

An alternate way to form a thick liquid flow isto utilize a rotating swirl flow, where again cen-trifugal forces keep the liquid against the wall.The scheme appears attractive for the FRC devicebecause of its geometric simplicity. To create arotational flow, the liquid carries both longitudi-nal and azimuthal velocity components. The flowspirals while it proceeds along the flow axis asillustrated in Fig. 11. Three-dimensional numeri-cal hydrodynamic analyses with Flibe as theworking fluid show that a 0.6 m thick liquid firstwall/blanket can be formed in a circular vacuumchamber of 2 m radius at an axial velocity of 7m/s and an azimuthal rotational velocity of 10m/s. In this design, the fluid enters the mainchamber zone through a convergent nozzle and isdischarged to a divergent outlet after one rota-tion. The velocity field along the axial direction is

Fig. 8. Some amount of thinning in liquid layer thickness was observed along the poloidal path due to gravitational acceleration andtoroidal area expansion. (Z-velocity component from three-dimensional hydrodynamic computation increases due to the gravita-tional acceleration and it has its maximum value at the mid-plane since velocity vector has only Z component at that location.) (a)initial velocity=8 m/s, (b) initial velocity =15 m/s.Fig. 9. Relative change in liquid lithium thickness as a function of distance along the flow direction for insulated open channel. Thethinning effect due to gravitational acceleration can be minimized by the drag from the M-shape velocity profile (A uniformthickness of 40 cm can be established for 10 m/s lithium flow).


Fig. 10. Two-velocity liquid metal streams can be established by having submerged sidewalls.Fig. 11. (a) Illustration of the swirl flow mechanism in the main FRC section with converging inlet and diverging outlet sections.(b) The three-dimensional fluid distribution of FRC swirl flow. (Result of CFD simulation).


Fig. 12. Two-dimensional (r−z) velocity distribution and liquid layer height distribution in the axial direction at an angle of 0°(where the effect of gravity on hydrodynamic behavior of the liquid layer is maximum). Inlet operating velocity: Vu=10 m/s,Vz=10 m/s.Fig. 13. (a) The modeling of ST structural geometry including the modification in the outboard topology. (b) The liquidtwo-dimensional velocity magnitude contour at r−z plane at the outboard and liquid layer height distribution in the z-direction atan arbitrary azimuthal angle.


shown in Fig. 12. The fluid encounters differentnet forces along the circumferential direction dueto different relative orientations of gravity, whichresult in non-uniform hydrodynamics characteris-tics along the flow axis. Calculations indicate thatan acceptable variation (B10%) in liquid layerthickness in azimuthal and axial directions can bemaintained for flowing Flibe with axial and az-imuthal inlet velocities of 11 and 13 m/s, respec-tively, in a cylindrical chamber with a 2 m radiusand 12 m length.

The swirl flow concept can also be applied tothe ST outboard FW/blanket region. In this case,a thick liquid carrying both vertical and toroidalvelocity components is injected at the top of thereactor. The centrifugal acceleration (\35 m/s2)pushes the fluid outward and prevents the flowfrom deflecting into the plasma core. The axialvelocity increases as the flow proceeds down-stream due to the gravitational acceleration andthis leads to flow thinning. The thinning effect isfurther manifested in the ST because of thetoroidal area expansion along the flow direction(the flow area increases by a factor of 2 as theflow approaches the mid-plane). Various numeri-cal simulations were performed to identify ‘ways’to slow down the velocity and to reduce thethinning effect. Preliminary results, based on com-putational fluid dynamics (CFD), three-dimen-sional calculations, indicate that the thinningeffect can be mitigated by tailoring the back wallcontour and by incorporating a step along theflow direction. The calculated ‘step’ of about 0.2-m high, located at the reactor mid-plane, helps tomaintain a liquid layer thickness greater than 0.3m. In contrast to the rotational flow for theoutboard blanket, a fast annular liquid flowsalong (no rotation) the central post forming theinboard FW/blanket zone, as shown in Fig. 13.

These hydrodynamics calculations (with orwithout the effect of magnetic field) indicate thata fairly uniform thick liquid wall can be formed inthe aforementioned fusion configurations as longas the injected fluid carries an adequate inertialmomentum (e.g. corresponding to a velocity of 10m/s). Moreover, the pumping power requirementbecomes less of a concern for higher power den-sity confinement concepts (such as FRCs) as

shown in Fig. 14. However, for a thick liquid wallconcept to work, there remain many design issuesparticularly in the areas of moving liquid in andout of the reactor, of spatial and temporal MHDeffects, and of accommodation of penetrations.Certainly, designs and analyses of the inlet nozzleand exit head recovery systems are needed for allthe aforementioned concepts. Regarding the ac-commodation of penetrations (e.g. for plasmaheating), simulations are performed to understandthe underlying scientific phenomena and toprovide design guidance. The challenges for ac-commodating penetrations include: flow stagna-tion at the front point of the penetration resultingin discharge of the fluid towards the plasma, riseof the fluid level surrounding the penetration dueto the obstruction of flow path, and wake forma-tion that persists downstream of the penetration.Different means have been proposed to avoidfluid splashing and to minimize disturbance.These involve modification of penetration shapes,introduction of guiding grooves and fins, and

Fig. 14. The pumping power as a fraction of fusion power forvarious cases of liquid wall thickness and confinement configu-rations. Note that the pumping power is scaled from that ofthe HYLIFE-II design, taking into account the 50% headrecovery. The effect of MHD drag will increase the calculatedpumping power of a Sn–Li flowing liquid wall.


Fig. 15. (a) Reference penetration case operating conditions and dimensions. (b) Tailoring of the back-wall topology surroundingthe reference penetration and related dimensions.


Fig. 16. Three-dimensional analysis of Flibe flowing around a penetration surrounded by a tailored backwall for the case shown inFig. 15.Fig. 17. Temperature of free-surface liquid lithium as a function of distance along the flow path shown for various cases of treatingthe surface heat flux (i.e. penetration of Bremsstrahlung radiation. Also shown is the lithium bulk temperature.Fig. 18. Sn–Li free-surface temperature increase due to surface heating as the flow proceeds downstream for different velocitymagnitudes.Fig. 19. Effect of different heat transfer mechanisms on Flibe free surface temperature (initial velocity=10 m/s and filmthickness=2 cm; final velocity=13.3 m/s and film thickness =1.5 cm).

alteration of the back wall topology, such asshown in Fig. 15. Preliminary analyses for CLiFFconcepts (2 cm thick convective liquid layer)confirm the effectiveness of the proposed schemes.As illustrated in Fig. 16, three-dimensional nu-merical simulations show a much reduced distur-

bance level, no splash occurring at the stagnationpoint, and no separation in the flow field. Theseresults are encouraging and provide ‘mechanisms’for solving penetration issues in thick liquid wallconcepts (where we are dealing with a much largervolume of fluid). The problems associated with


accommodating penetrations are high priority is-sues for further study.

4.4. Heat transfer and free surface temperature

The power emanating from the burning plasmastriking the liquid will cause its surface tempera-ture to rise as it flows downward in the chamber.The temperature of the free liquid surface facingthe plasma is the crucial parameter which governsthe amount of liquid that evaporates into theplasma chamber and, therefore, is a potentialsource of plasma impurity. However, it can onlybe determined accurately if the heat transfer at thefree surface/vacuum boundary is well understoodas analyzed below. The allowable temperaturefrom a plasma impurity standpoint is analyzed inthe plasma edge modeling reported in Section 8.

Since a liquid metal wall will be ‘laminarized’by the magnetic field, the heat transfer at the freesurface wall is determined by the laminar convec-tion and conduction. Furthermore, the lithiumsurface temperature is reduced because the surfaceheat load is deposited into the bulk due to X-raypenetration. Calculations show that, under a sur-face heat flux of 2 MW/m2, the lithium freesurface temperature drop can be kept below100°C at a velocity of 20 m/s throughout theARIES-RS reactor, taking account of X-ray pene-tration (see Fig. 17). This film temperature dropincreases to 140°C if the lithium velocity is de-creased to 10 m/s. Nevertheless, it appears thatthe lithium free surface temperature can be main-tained below 400°C, which may be acceptable tothe plasma operation. The surface temperaturerise for Sn–Li as it reaches the bottom of thereactor is higher than that of lithium flowing atthe same velocity. This is due to a lower thermalconductivity and a higher z. However, because ofits low vapor pressure, Sn–Li can flow at lowervelocities for higher surface temperatures yet notjeopardize plasma operations. However, the po-tential of using Sn–Li for a non-structure-thickliquid wall design is limited by its high density. Asshown in Fig. 18, a velocity magnitude of about7.5 m/s or more is needed to maintain the liquidadherence to the wall as well as the surface tem-perature to remain below 827°C (vapor pressure

corresponding to that of lithium at 400°C). Thecorresponding pumping power is about 6% of afusion power of 5480 MW for a 45 cm thickSn–Li FW/blanket.

The molten salt Flibe, a low-conductivity, high-Prandtl fluid, is not fully laminarized by the pres-ence of the magnetic field. The heat transfer at theFlibe free surface wall is dominated by the rapidsurface renewal by the turbulent eddies generatedeither near the back wall or nozzle surfaces byfrictional shear stress, or near the free surface dueto temperature driven viscosity variations. Accu-rate calculations of Flibe free surface temperaturerequire the knowledge of the turbulent structures,eddy generation and dissipation, and the degree ofturbulence suppression by the magnetic field. Inan attempt to calculate the Flibe free surfacetemperature and to examine the effects of themagnetic fields on turbulence suppressions, theso-called k–o model of turbulence [1] was devel-oped. Preliminary results based on the k–o modelof the turbulence, including MHD effects andvarious boundary conditions, predict a range oftemperatures that may in some cases be beyondthe plasma compatible temperatures. As shown inFig. 19, if the Flibe flow is laminarized, the Flibefree surface can be overheated. The film tempera-ture drop can reach 700°C at the bottom ofARIES-RS under APEX 2 MW/m2 surface heatload (curve 1) while turbulent heat transfer con-siderably reduces Flibe free-surface temperaturedrop (curve 2). On the other hand, accounting forBremsstrahlung radiation penetration further re-duces surface temperature by about 90°C (curve3). Furthermore, heat transfer at the vacuum/freesurface interface can be significantly enhanced bythe existence of surface turbulence (curve 4), whileturbulence suppression due to MHD can be ne-glected at the current parameters of interest (curve4). If the Flibe surface temperature is high relativeto the plasma operation limit, further design mod-ifications such as using two coolant streams maybe required to accommodate this difficulty.

4.5. Power con6ersion and two stream flows

The challenges of the liquid wall designs gobeyond achieving a low enough surface tempera-


Fig. 20. Thermal efficiency as a function of steam temperature in various systems. Power conversion efficiency determines materialchoice and bulk exit temperature.Fig. 21. The DPA rate in the backing solid wall.Fig. 22. The helium production rate in the backing solid wall.


ture compatible with the plasma operations, butalso to maintain a mean bulk temperature ofgreater than 600°C for high thermal efficiency (seeFig. 20). This temperature can be higher than themaximum allowable free surface temperature.One approach to overcome this difficulty in athick liquid wall design is to use two differentcoolant streams: one for surface heat removal andthe other for neutronics heat deposition in orderto simultaneously achieve these two conflictingtemperature requirements. A power conversionsystem would then include two cycles: one for theconversion of the thermal power fast plasma-fac-ing stream, and the other for conversion of thethermal power in the thick liquid behind it, whichhas a much higher thermal conversion potential.Since thick liquid walls will reduce both the neu-tron damage rate and helium transmutation rate,the choice of the structural material should bedetermined by the high temperature capabilityand liquid/structure compatibility. It appears thatthe use of tungsten alloys would achieve thehighest thermal efficiency because of its high tem-perature operation capability. The oxide-dispersedferritic steel (ODFS) can operate up to a tempera-ture of 650°C, which provides a thermal efficiencyof about 41.2%.

4.6. Effects of liquid walls on reducing acti6ationand radiation damage

Reducing activation and radiation damage tostructural materials are among the important ad-vantages of liquid walls, particularly the ‘thick’liquid wall concepts. The magnitude of these ad-vantages is design dependent. Calculations wereperformed to quantify the benefits as a functionof key design parameters. The results are brieflysummarized in this subsection. These results wereutilized to guide the choices for concept explo-ration discussed earlier in this section. More de-tailed analysis is required in the future to addressflow support structures other than the backingwall that may be also needed in liquid wall de-signs (for example, inlet nozzles and flow di-viders). The thickness of the liquid in front ofthese elements may be less than that protectingthe backing wall. Note also that these elements

are more accessible than the backing wall (or firstwall in traditional concepts) and therefore fastermaintenance may be possible. In particular, thekey factor is the extent to which liquid wallsattenuate the neutrons before they reach thestructural materials. The main structural elementin liquid wall designs is the backing wall. So, thethickness of the liquid wall is important in deter-mining the reduction in activation and radiationdamage in the back wall relative to the solid firstwall in traditional concepts.

4.6.1. Radiation damage parametersThe effects of the liquid layer thickness on

radiation damage parameters such as atomic dis-placement and helium production rates were stud-ied for lithium, Flibe, Sn–Li, and Pb–Li. Figs. 21and 22 show the rates (per full power year, FPY)of atomic displacement (DPA) and helium pro-duction (appm), respectively, in the back wallstructural material as a function of the liquidlayer thickness (L) protecting the back wall.

Without the liquid layer, which corresponds toa ‘bare wall’ case, the DPA and helium produc-tion rates in the backing solid wall are compara-ble in the four breeders. However, because Pb–Liexhibits larger reflection, the low-energy neutronflux is larger at the solid wall which results inlarger DPA rate (occurs at all energies). This alsogives smaller He-4/DPA ratio in the case of Pb–Li breeder (�8.7) as compared to the value withthe other breeders (10–11).

As the thickness of the liquid layer increases,the reduction in these damage parameters variesamong the four breeders. Lithium is the weakestmaterial in moderating neutrons as compared tothe other breeders. The reduction in DPA rate isless than an order of magnitude for L=42 cm,while the reduction in helium and hydrogen pro-duction is about an order of magnitude. Theattenuation characteristic of the Pb–Li breederfor the DPA rate is similar to lithium. However,the Pb–Li is superior to the other breeders inattenuating the helium and hydrogen productionrate in the solid wall. This is due to its largerattenuation power to high-energy neutrons(through (n, 2n) and (n, inelastic) reactions) whichis basically the main contributor to the high-


threshold helium and hydrogen reactions in thesolid wall. Because of the smallest He-4 produc-tion and the largest DPA rate with the Pb–Libreeder, the ratio He-4/DPA is the smallest (�0.3) at L=42 cm as compared to the values withthe other breeders (Li: �7, Flibe: �6, Sn–Li: 2).The attenuation characteristics of the Flibe andSn–Li are similar for the helium and hydrogenproduction. However, the Flibe gives the bestattenuation to the DPA rate since it is capable ofattenuating both the high- and low-energy com-ponent of the neutrons reaching the backing solidwall.

Using the damage values at the bare wall (L=0cm) and at the wall with various L thicknesses,one can estimate the tenfold thickness, L10, foreach breeder defined as the required thickness ofthe layer to reduce a particular response, R (dam-age parameter), by an order of magnitude. Thisthickness is given in Table 7 for the variousdamage parameters and breeders. For helium andhydrogen production rate, �22 cm is required toachieve an order of magnitude reduction withFlibe and Sn–Li and smaller thickness (�18 cm)is required in the Pb–Li liquid layer. Twice asmuch thickness is required in the Li case becauseof its poor attenuation characteristics for heliumand hydrogen production. As for the DPA rate,larger thickness is required. It is �26 and �36cm for the Flibe and Sn–Li, respectively, butmuch larger thickness (�58 cm) is required in theLi and Pb–Li to reduce the DPA rate by an orderof magnitude.

The DPA rate in backing solid wall is �26DPA/FPY, 3.6 DPA/FPY, 9.5 DPA/FPY, and 30DPA/FPY, with the Li, Flibe, Sn–Li, and Pb–Liliquid layer, respectively. If the 200 DPA is con-sidered as the limit at which the wall and shieldzone require replacement, the lifetime of these

components would be 7.7, 56, 21, and 7 yr, re-spectively. Clearly the presence of the liquid layermade these components last the lifetime of theplant (30 years) when Flibe is used as the breeder.In the case of Sn–Li, one replacement may berequired after �20 years. But three to four re-placements may be needed in the case of Li andPb–Li breeders.

4.6.2. Hazard and 6olume of radioacti6e wasteAnother clear advantage of deploying a thick

liquid wall/blanket concept is the substantial re-duction in the hazard and volume of the wastegenerated from the activation of solid materials(including solid walls, vacuum vessels, shield andmagnets themselves). It has been found [1] incomparing the liquid FW/blanket (42 cm-thick) toa conventional blanket (2 cm-thick ferritic steel,FS, FW — 40 cm-thick blanket made of 10% FSand 90% Flibe), while keeping the radial build thesame (ARIES-RS radial build and materials areassumed), the specific activity (curies/cc) at shut-down in the bare FW of the conventional blanketis two orders of magnitude higher than the spe-cific activity in the back wall of the liquid FW/blanket case. The specific biological hazardpotential (BHP) has the same features. The twoorders of magnitude difference continues duringthe first year and starts to narrow down after thefirst year following shutdown. The next step is tofind out how this may translate into advantagesfrom both the waste generation (dominated bylong-lived nuclides) and safety hazard (dominatedby short-lived and intermediate-lived nuclides)viewpoints.

An analysis comparing the waste disposal rat-ings and volume of waste generated in a powerplant based on the two concepts was conducted.

Table 7The tenfold thickness (in cm) of the liquid layera

Li/FSParameter Pb–Li/FSSn–Li/FSFlibe/FS

�56�36�26�58DPA (dpa/FPY)�22�46 �18Helium production (appm/FPY) �21�22 �22�44 �19Hydrogen production (appm/FPY)

a The thickness required to reduce a response by an order of magnitude.


Table 8Comparison of class C waste disposal ratings using Fetter limits

Zone Liquid blanket conceptFPY Conventional blanket concept

–3 1.37Inboard FW and blanket0.81Inboard shield 0.73300.14130 0.1Inboard VV

–3 1.34Outboard FW and blanket0.795 0.71Outboard shield 300.087 0.0630Outboard VV

The waste disposal ratings for the Fetter [15] and10CFR61 [16] limits are shown in Tables 8 and 9,respectively. Results in the tables are given forcompacted wastes after 1 year following shut-down. As shown in Table 8, according to Fetterlimits, all components of the liquid blanket con-cept would qualify for disposal as class C wasteafter 30 FPY. All components of the conventionalblanket concept, except for the first wall andblanket, also would qualify for disposal as class Cwaste after 30 FPY. The first wall and blanketwould not qualify for disposal as class C LLWunless they were replaced every 2 FPY instead ofevery 3 FPY. On the other hand, the 10% steelstructure in the conventional blanket provided theshield and vacuum vessel behind it with bettershielding, resulting in lower waste disposal ratingsin comparison to the waste disposal ratings of theshield and vacuum vessel behind the liquid blan-ket. Results in Table 9 show that, according tothe 10CFR61 limits, all components of both blan-ket concepts would qualify for disposal as class Cwaste. The absence of contribution from 192mIr tothe waste disposal ratings according to the10CFR61 limits (10CFR61 has no limits for192mIr) resulted in allowing for the disposal of thefirst wall and blanket of the conventional blanketconcept as LLW after 3 FPY.

A power plant based on the conventional blan-ket concept will produce the equivalent of aboutten blankets of additional waste during its life-time. However, a power plant based on either theliquid or conventional blanket concepts will gen-erate a comparable amount of waste from theshield, vacuum vessel, and magnets, whose vol-umes far exceed the volume of the blanket. As

shown in Fig. 23, the volume of the waste gener-ated during the life-time of a power plant (30FPY) based on the liquid blanket concept couldbe a factor of six lower than the volume of wastegenerated during the same life-time if the plantwas based on the conventional blanket concept.The factor of six is based on the assumption thatthe waste is non-compacted and the waste doesnot include the magnets. If the waste is compactedto 100% of its theoretical density, the reductionfactor drops from six to two. If the waste iscompacted and the magnet waste is included, apower plant based on the conventional blanketconcept will generate about 35% more waste dur-ing its life-time (30 FPY) than a similar powerplant based on the liquid blanket concept.

4.7. Key issues and R&D

The present state of understanding of thickliquid wall concepts does not reveal any basicflaws in the underlying scientific and technicalarguments for the concepts. Yet, there remainmany issues for the implementation of this con-cept in any magnetic fusion configuration. Nearterm R&D should focus on continued conceptexploration as well as modeling and experimentsfor key feasibility issues. These include:1. Edge-plasma and core-plasma modeling and

analysis as well as experimental research invarious confinement devices for plasma–liquidwall interactions.

2. Experimental data on the achievable minimumliquid surface temperatures without MHD ef-fects for turbulent Flibe and MHD lami-narized lithium/tin–lithium flow under highpower density conditions.


3. Identification of practical heat transfer en-hancement schemes necessary for minimizingliquid surface temperatures.

4. Experimental characteristics of small-scale liq-uid metal flow hydrodynamics configurationsapplicable to MFE confinement schemes suchas 1/R toroidal field variation, and effects offinite radial, poloidal, and vertical fieldcomponents.

5. Computer simulation of MFE relevant three-dimensional free surface liquid wall thermaland hydrodynamics performance with MHDeffects. In particular, hydrodynamics charac-teristics near the penetrations and supply andreturn lines.

6. Identification of the most promising hydrody-namics configurations with respect to differentMFE confinement schemes.

In addition, engineering innovations and analysesare required for the following numerous mechani-cal design issues including:� How to move mass quantities of liquid metal

or salt in and out of the machine reliably.� How to provide sufficient access for supply

piping and return ducts.� How to design the piping and nozzles for reli-

able operation at high fluid velocity.� How to start and stop the system safely.� How to keep the stream attached to the in-

board wall (must prevent toroidal rotation ofinboard stream).

� How to provide sufficient penetrations forheating and diagnostics.

� How to account for image current effects frommoving plasma.

Issues related to effects of liquid metals on theplasma core and edge-plasma liquid–surface in-teractions are discussed in Sections 7 and 8.

5. Thin liquid wall concepts

The thin liquid wall concept was explored inAPEX for liquid metals and for Flibe. Initialdesigns of thin liquid walls were developed andthe associated advantages and disadvantages wereanalyzed. Thin liquid wall concepts are calledCLiFF.

The idea behind CLiFF (the Convective LiquidFlow First-Wall concept) is to eliminate the pres-ence of a solid FW facing the plasma throughwhich the surface heat load must conduct. Thisgoal is accomplished by means of a fast moving(convective), thin liquid layer flowing on the FWsurface (see Fig. 24). This thin layer is easier tocontrol than a thick liquid FW/blanket, but stillprovides a renewable liquid surface immune toradiation damage and sputtering concerns, andlargely eliminates thermal stresses and their asso-ciated problems in the first structural wall. Theattractiveness potential and key issues for theCLiFF design are summarized in Table 10. TheCLiFF class of liquid wall concepts is viewed as amore near-term application of liquid walls.

Details of the preliminary design, heat transfer,power balance, thermal-hydraulics, neutronics,activation and safety are included in this section.It is noted that the first several centimeters ofvarious thick liquid FW/blanket concepts dis-cussed in the preceding section will behave in asimilar fashion to the CLiFF concept discussed

Table 9Comparison of class C waste disposal ratings using 10CFR61 limits

FPYZone Liquid blanket concept Conventional blanket concept

–Inboard FW and blanket 0.49530.2130Inboard shield 0.25

30Inboard VV 4.22×10−3 2.82×10−3

3 0.473Outboard FW and blanket –30 0.21Outboard shield 0.25

1.69×10−32.54×10−3Outboard VV 30


Fig. 23. Comparison of total volume of waste generated in power plants based on thick liquid metal blanket and conventionalblanket concepts.Fig. 24. Conceptual sector schematic of CLiFF implementation in ARIES-RS reactor.

here, and significant overlap with those analyses isseen in what follows.

5.1. Design description

The majority of the work reported here wascarried out for the tokamak. Specifically, theARIES-RS geometry was utilized whenever possi-

ble, with modifications for the unique structuresand high flowrates required for CLiFF. Thismeans, however, that the ARIES-RS fusionpower needs to be scaled-up to 4500 MW to givethe 10 MW/m2 peak neutron wall load and 2MW/m2 peak surface heat flux goals of the APEXstudy. Tokamaks present a difficult challenge forliquid walls due to the fact that the plasma cham-


ber is relatively closed with short scrape-offlengths, and so, vaporized liquid wall materialmust be screened by the edge plasma to keep itfrom penetrating to the core.

The general CLiFF design, as seen in Fig. 24, isconceptually simple in its implementation. A thinfast liquid layer is injected near the top of theplasma chamber. The layer flows down the reac-tor walls without excessive slowing or thinning,and is removed in some fashion from the bottomof the chamber. Layer thickness h on the order of0.5–2 cm, and velocity U on the order 10 m/s, areconsidered. The curved back wall fits the plasmashape and provides an adhesion force due to theliquid’s centrifugal acceleration. The criterion forcontinuous attachment of the liquid layer is sim-ply U2/Rc\g cos a, where g is the acceleration ofgravity, Rc is the radius of curvature of the firstwall section and a refers to the angle of theoutward surface normal to gravity vector (so 0° iscompletely inverted).

The velocity range is chosen quite high both toensure adhesion to the back-wall, but also to keepthe exposure time to the plasma short, and thuskeep the surface temperature low. If one desiresan inlet temperature that is \300°C (for powerconversion reasons), it turns out that it is thissecond restriction that is the more severe, basedon the maximum surface temperature estimatesprovided by the preliminary plasma edge analysis.The high velocity requirements and the large cov-erage area result in volumetric flow rates in excessof 10 m3/s compared to ARIES-RS in the 3 m3/srange.

The conceptual CLiFF design shown in Fig. 24has an integrated droplet-type divertor. Somemeans (mechanical or electrical) is used to stimu-late the breakup of the FW flow into a dropletscreen. It is hoped that the droplet screen willhave a higher heat removal capability due to therapid rotation and internal circulation in thedroplets, but this fact remains to be proven. In

Table 10Potential advantages and issues of CLiFF concept for APEX

Potential Issue

Removal of surface heat loads (greater than 2 MW/m2 possible). Hydrodynamics and heat transfer involve complicatedLocal peaking and transients can be tolerated MHD interaction between flow, geometry, and the

magnetic field:Suppression of turbulence and waves

FW surface protected from sputtering erosion and possibly LM-MHD drag thickens flow and inhibits drainage fromchamberdisruption damageEffects of varying fields on LM surface stability and drag

Beneficial effects on confinement and stability from conductingshell and DT gettering effects

Evaporating liquid can pollute plasma, surface temperaturelimits unknown

Elimination of high thermal stresses in solid FW components,having a positive impact on failure rates

High flowrate requirement can result in low coolant DT ortwo coolant streams

Possible reduction of structure-to-breeder ratio in FW area, withbreeder material facing virgin neutron flux

Effect of liquid choice on edge plasma gettering, tritiumthrough-put, and tritium breeding

Integrated divertor surface possible where CLiFF flow removesall a heat

Neutron damage in structure is only slightly reducedcompared to standard blankets, blanket change-outrequired for high power density operation

Complex tokamak D-shape and ports can likely beaccommodated


addition, for LMs, the droplet screen will beelectrically isolated from the main FW flow andplasma currents will not be able to close. Theliquid film can be removed from the vacuumchamber by gravity drainage or by an EM pumpif the working liquid is an electrical conductor.

Supply nozzles will form the desired liquid flowat the top of the reactor. These nozzles can bedesigned to be protected from surface heat flux bythe flowing liquids.

Note also that since these nozzles are at the topof the reactor chamber, the surface heat load andnuclear heat will be lower than the peak mid-plane values. Liquid removal from the plasmachamber is accomplished through a combinedvacuum pumping and liquid drain port. It isenvisioned that the liquid flow itself will pump aportion of the implanted plasma particles into thepumping ducts by convection, thus aiding in im-purity removal.

The working liquid should be a tritium breed-ing material like lithium, tin–lithium or Flibe.Thus the liquid removed from the reactor can berecirculated to the blanket as the main tritiumbreeder and coolant. The bulk nuclear heat isadded on top of the FW/divertor heat before theliquid is sent to the power conversion system. Inthis manner, the FW and divertor power is con-verted at relatively high thermal efficiency.

Penetrations for various heating, fueling anddiagnostics functions will be provided as much aspossible in the lower half of the outboard FW.Flow can be guided by means of submergedgrooves around the penetration, and close againdownstream to form a continuous surface protec-tion as discussed in the previous section. Coolingof the penetration structures themselves will beaided by the CLiFF flow. It is likely that for LMs,the penetrations will have to be electrically iso-lated from the flow by means of an insulatorcoating. This will be true in supply lines andnozzles as well.

Off-normal plasma events like disruption canpossibly induce large currents in LM CLiFF flowsand cause the layer to be splashed or torn off thewall altogether. For poorly conducting Flibe, theeffect of the disruption is not as clear. It is hopedthat, in any case, splashing will turn out to be anallowable response, and that the liquid wall will

just be restarted following the disruption. For anall-liquid wall system, this seems a reasonableassumption, except for possible damage to anten-nae and sensitive diagnostics. It is hoped that‘liquid tolerant’ antennae could be designed thatcould accept the occasional splashing of liquidmetal, but this certainly remains to bedemonstrated.

5.2. Hydrodynamic and heat transfer analysis

Aside from plasma compatibility, one of thekey issues for CLiFF is related to finding a feasi-ble hydrodynamic configuration. A significantamount of design analysis has been done so far onCLiFF in order to answer the three basic ques-tions: How do you form it? How do you drain it?How do you maintain it? It is noted that liquidmetals and Flibe behave very differently in themagnetic environment of a tokamak. The lowthermal and electrical conductivity of Flibe leadsto a FW flow that will still be turbulent, and haveheat transport at the free surface and flow drag atthe back-wall that depend heavily upon this tur-bulence. For LMs the converse case occurs, whereit is expected that the MHD effects will dominatethe drag, and the thermal conduction dominatesheat transfer.

5.2.1. Turbulent Flibe flowSeveral models have been applied to predicting

the flow profiles for Flibe, ranging from simplehydraulic models for the steady state equilibriumflow profile, to more complex two- and three-di-mensional non-steady codes for studying phenom-ena like surface waves and penetrations. The 1.5D hydraulic calculations indicate that flow depthequilibria in the range of 2 cm can be achieved forFlibe flows in the 10 m/s range (see Fig. 25). Amore sophisticated, low-Reynolds number k–o

model of turbulence was also applied to theCLiFF flow in order to study the effect of MHDturbulence on the flow profile. In comparison tothe ordinary k–o model, the present one wasextended to the MHD case by means of addi-tional terms in the closure equations. Due toturbulent viscous friction, the layer thickness in-creases rapidly over the initial flow section (see


Fig. 25. Surface height predictions for Flibe with various models: Line 1: k–o, 2: Darcy Weisbach friction factor=0.025, and 3:laminar.

again, Fig. 25). This is in contrast to the resultspresented earlier where the simple friction factorformulation predicts nearly constant flow heightand velocity profiles for CLiFF. This contradic-tory result is cause for concern because if the layerslows down significantly, the transit time throughthe plasma chamber will go up, as well as thesurface temperature. Attempts to benchmark thek–o and friction factor against available data fromthe UCLA Mega-Loop Experiment [17] are incon-clusive — the data splits the difference betweenthe k–o and friction factor model.

The effect of the magnetic field on the flowparameters is negligible if the Hartmann number isless than about 1000, and hence for CLiFF withHa=500, we conclude that there is no strongimpact of MHD on the Flibe flow hydraulics.

Heat transfer calculations using this same modelindicate that depending on surface turbulence as-sumptions, the temperature rise at the surface canbe quite low. For a 10 m/s, 2 cm thick Flibe flow,the surface temperature rise is in the range of

30–160°C depending on whether optimistic orpessimistic assumptions are used. The effect of themagnetic field again appears to be small. Whenconsidering the thermal hydraulics, it is seen thatthe temperature window for Flibe is limited (seeFlibe system diagram in Fig. 26), and so thesurface heat transfer is critical for feasibility.There are, however, no experimental data, and thisissue needs closer study and experimental valida-tion.

The surface stability for Flibe CLiFF flows wasalso analyzed using a linear stability analysis tech-nique for infinitesimal disturbances. For CLiFF,the results show that whenever the flow is adhered,it should be stable as well. The effect of finite sizeperturbations may alter this picture. The primarysource of large disturbances comes from the turbu-lence of the flow itself. The fluid dynamic behaviorof the first-wall flow system may be affected due tothese eddy generating mechanisms includingboundary layer relaxation near nozzles, Gortler-type instabilities, structural vibrations, etc.


Penetrations have also been analyzed for theFlibe case using a three-dimensional free surfacecode that allows the introduction of arbitrarily

formed structures. The penetrations consideredare elongated into ellipses in order to be morehydrodynamically streamlined. The specific case

Fig. 26. CLiFF — flow/temperature schematic-Flibe option.Fig. 27. Influence of the wall conductance ratio on the layer thickness increase (2b=1 m). Line 1 — cw=0; 2 — cw=1.0×10−6;3 — cw=2.0×10−6 (Lithium).


considered has dimensions 20 cm wide and 90 cmlong (in the flow direction). The back-wall in thevicinity of the penetration is tailored to guide theliquid around the penetration itself, and to aid inclosing the liquid again downstream of the pene-tration. Results presented in the previous sectionof thick liquid walls show that such a designsolution can successfully guide the flow aroundpenetrations, but additional work and optimiza-tion is needed for their design.

5.2.2. Magnetohydrodynamics for lithium andSn–Li flows

Mathematically these types of flows can bedescribed by a set of Navier–Stokes equations forincompressible fluids and Maxwell’s equations forelectromagnetic phenomena. The numerical toolsused to analyze this system of equations are basedon two-dimensional, simplified magnetohydrody-namic equations and can be performed in practicefor any values of governing parameters for ductsof various geometries. This is an extreme simplifi-cation of the physics and assumes that all currentsclose in their own cross-sectional plane. This typeof calculation is accurate for well-behaved, nearlyfully developed flows with simple geometries, butignores significant effects near field gradients anddeveloping regions.

It is well known that the presence of electricallyconducting walls can lead to larger electrical cur-rents in the flow domain and, as a result, to asignificant increase in the MHD drag effect. In thecase of free surface MHD flows, this effect mani-fests itself in the increase of the layer thicknesswith the accompanying reduction in the velocity.Ideally, if the liquid layer is assumed to be com-pletely axi-symmetric in the toroidal direction,flow along poloidal flux surfaces with no fieldgradients, no MHD drag will occur. This idealcase, though, is not possible in practice and welook at two variants to gauge the relative effectsof the MHD. One case is the presence of fins,side-walls, or penetrations breaking up the flowtoroidally, and the other is a slight deviation ofthe flow path from the flux surfaces resulting in asmall surface-normal field component. Figs. 27and 28 illustrate the results for these two cases forlithium, where we assume that a doubling of the

Fig. 28. Influence of the radial magnetic field and the wallconductance ratio on the maximum thickness of the lithiumlayer (for axi-symmetric case, Cw=�).

initial height represents an unacceptable result.Note that in Figs. 27 and 28, the thicknesses onthe vertical axes are scaled by the initial thickness.

For the case of side-walls, it was found thatelectrically insulated side-walls are acceptable onlyif they are no closer than 1 m toroidally, and thatlow conductivity walls like SiC (thickness=1 cm,assumed s=103 V−1m−1) are acceptable pro-vided they are no closer than 8 m. Bare metalwalls (thickness=2 mm, s=106 V−1m−1), evenif very thin, can be no closer than 110 m, and soare not feasible for CLiFF. For the case of asmall radial field it was found that if the back-wall is bare metal the allowable field is onlyBrB0.1 T. This value goes up to BrB0.5 T if thebacking wall is insulated. These calculations as-sume that there are insulated side-walls present atsome distance to break up the toroidal electricpath (but they are separated by enough distancethat they do not add appreciable drag). If com-plete axi-symmetry is assumed, where inducedcurrents close on themselves, the allowable radialfield is BrB0.015 T! These calculations indicatethat serious work is needed in the area of LM-MHD analysis and experiments to prove thatpassive flow schemes like CLiFF are possible.

Heat transfer at the surface is calculated for Liand Sn–Li using only conduction, but assumingsome penetration of X-ray photons in the case oflithium. The conclusion is that at 10 m/s thetemperature rise will be on the order of 150°C for


Li, and 300°C for Sn–Li. The thermal-hydrauliccalculations utilizing these numbers result in ablanket outlet temperature around 650°C for theSn–Li, but much lower for the lithium, possiblynecessitating a two-stream approach, where onlypart of the Li flow is sent to the blanket.

The results of stability computations are in agood agreement with the linear stability analysisconclusions. Long wavelength initial disturbancesgrow very rapidly on the inverted surface underthe effect of gravity and centrifugal accelerationand then propagate down with slowly decreasingamplitude. The growth rate and the maximumamplitude depend on the wavelength. The shortwaves (B20 cm) are suppressed rapidly by thesurface tension, while the long wave disturbances(1.5–2 m) are not suppressed over the whole flowlength. The most dangerous disturbances arethose having the long wavelength of about 2 m,for which the amplitude can reach 40–50% of theinitial flow depth, however, layer disintegration,flow separation, and/or excessive increase in thethickness do not accompany the wave propaga-tion. Therefore, special means to suppress surfaceinstability are not needed provided inlet fluctua-tions are at a level B5–10%.

Due to the complexity of the problem, no de-tailed work has yet been done in the area ofaccommodation of penetrations with liquidmetals. Such penetrations represent in MHD flowboth a disturbance to the hydrodynamic flow fieldvia the physical diversion of liquid from its initialcourse, and also, and more significantly, a distur-bance to electrical current paths that potentiallycan overwhelm the flow with local and globalMHD drag. Preliminary conclusions, gleanedfrom the discussion of side-walls above, is thatany penetration will require an insulator coatingto isolate the structure from the free surface flow.

5.3. Nuclear heat, tritium breeding, and acti6ation

The thin layer of liquid does not significantlyalter the radial build of ARIES-RS, however, thechoice of working liquid plays a big role in theneutronics. Analyses of the nuclear heating andactivation have been carried out using theARIES-RS radial build at higher power density

and with different coolants. The conclusions arethat waste and damage issues in the vacuumvessel, the shield and magnets are lower whenFlibe and Sn–Li are used, as compared tolithium. Solid walls damage parameters are re-duced by �10–15% with the 2 cm Li-layer and�20–30% with 2 cm Flibe or Sn–Li layers.Lithium coolant offers the best tritium breedingpotential at natural Li-6 enrichment. Lithium andFlibe coolants have maximum tritium breedingratio (TBR) at 25% Li-6 enrichment (local TBR�1.5 for Li and �1.2 for Flibe) whereas it keepsincreasing with Li-6 enrichment in the Sn–Licoolant (�TBR �1.3 at 90% Li-6). The inclu-sion of beryllium drastically enhances TBR in theFlibe and Sn–Li cases (local TBR�1.7 in Flibeat 25% Li-6 and �1.4 in Sn–Li at 90% Li-6enrichment) which indicates that the tritium self-sufficiency condition could be met with Flibe orSn–Li breeder. With regard to power depositionhowever, the Sn–Li offers the largest power mul-tiplication (PM) among the several breeders. PMis �1.4 for Sn–Li, �1.14 for Li and �1.02 forFlibe. The Sn–Li breeder therefore could offerbetter plant thermal output for the same fusionpower.


There are several dominant issues that go di-rectly to the feasibility of this concept, and manymore issues that weigh heavily on the ultimateattractiveness. The amount of allowable evapora-tion must be determined for all liquid candidates.This is both a feasibility issue and an attractive-ness issue. We recognize that a fully consistentanswer to this question will require a considerableamount of research in modeling and analysis ofplasmas with liquid wall boundaries, as well asexperimental research in various confinementdevices.

In addition to the plasma compatibility, theissue of establishing a viable hydrodynamicconfiguration threatens feasibility. The issues inthis category differ significantly for molten saltsversus liquid metals. For Flibe, the main issueconcerns the penetration of heat at the free sur-face and the availability of a robust operating


window. Other issues as to the formation andremoval of the liquid flow in the plasma chamber,and the accommodation of penetrations are alsoserious, but in our opinion solvable via numericalmodeling and scaled experiments with Flibe simu-lants (such as water). The heat transfer issue is amore serious unknown, as current limits on sur-face temperature for Flibe are estimated by theplasma interface group at about 560°C. Also aserious issue for Flibe, is the behavior in thedivertor region, where direct plasma contact oc-curs. The amount and nature of the materialsputtered and redeposited needs to be determinedbefore accurate plasma modeling of the regioncan take place.

The main issue facing liquid metals is of coursethat of MHD interaction. The CLiFF flow itself isvery sensitive to changes in drag since the onlyforces governing the flow are gravity and friction.Without toroidal axi-symmetry of the flow andfield, reliable insulator coatings will be requiredon all surfaces in contact with the LM layer.MHD forces from surface normal components ofmagnetic field can upset this balance, even whencomplete axi-symmetry is assumed in the toroidaldirection. Additionally, gradients in toroidal fieldcan exert a significant drag on the free surfaceflow. LMs however, offer the potential for activecontrol that is not present with the molten salt.By biasing and applying electric currents, the LMcan be pumped or pushed against the back-wallin-situ — offering the chance to ‘confine’ theliquid wall just as we confine the plasma. All theseeffects need to be analyzed in greater detail, withboth modeling and small-scale experimental ef-forts to see if a suitable flow is indeed possible inthe real fields of a tokamak or other plasmaconfinement devices.

Apart from the free surface flow itself, MHDissues exist in the LM supply and drain lines andblanket flows as well. Insulator coatings areneeded for these structures. Additionally, due tothe large LM flowrates required for CLiFF, largepressure drops are expected in the entrance re-gions between toroidal field coil legs. These pres-sure drops can theoretically be overcome byin-situ LM pumping, but lead to very large pump-ing powers for the CLiFF designs with LMs. Aclever design of inlet piping may help reduce this

effect, as would a reduction in the LM flowrate aswell.

Impact of liquid wall implementation on otherreactor systems is another category of issues forthe CLiFF concept. In particular, it will be likelythat heating and diagnostic ports must be re-designed to allow flow to pass around the pene-tration. Pumping systems with a considerableamount of vapor from liquid evaporation willneed to be modified. Tritium recovery (especiallywith hydrogen getters like lithium) will be evenmore challenging, and material selection and com-patibility to help optimize liquid wall performancemust be addressed. Flibe and Sn–Li databaseissues must be addressed for all liquid wall andblanket options as well.

6. Electromagnetically restrained lithium blanket

This section focuses on another type of thickliquid in which electromagnetic forces are utilizedto restrain, or ‘confine’, the fluid. In this concept,called the electromagnetically restrained (EMR)lithium blanket, an approximately one meter thickshell of liquid lithium metal almost completelysurrounds the tokamak’s toroidal plasma dis-charge, absorbing plasma particles, neutrons andother radiation while breeding tritium and collect-ing high temperature heat for power generation.The �1 m thickness is chosen based on consider-ations of tritium breeding, of absorbing most ofthe fusion power, and of minimizing activationand damage to the solid chamber walls locatedbehind the liquid. Of the candidate liquid materi-als, pure lithium metal is chosen due to its highabundance, superior tritium breeding, low chemi-cal toxicity, almost zero neutron activation, andits high conductivity resulting in low power con-sumption for the EMR action.

The EMR concept converts MHD difficultiesintroduced by the liquid metal’s electrical conduc-tivity into MHD advantages by deliberately in-jecting controlled electrical currents to influenceliquid flow dynamics. As depicted in Fig. 29, twoaxi-symmetric liquid lithium streams enter thetoroidal chamber’s top. The two streams are elec-trically separated there, either by an electrical


insulator or by a non-insulating structure in whichsome electrical dissipation is wasted via leakage. Atthe top, the two streams are biased to differentvoltages via electrodes connected to an externalpower supply. Poloidal current injected via theseelectrodes is conducted through the streams whichmeet and join at the bottom of the chamber. Theresulting J×B electromagnetic forces push thestreams against the chamber walls and thus helphold them away from the plasma. The EMRlithium blanket concept makes use of these electro-magnetic forces in conjunction with the othernatural forces that exist, including centrifugal (iner-tial) forces, contact forces, viscosity, and surfacetension. The liquid’s transit time from the top tothe bottom of the chamber is determined by grav-ity, frictional losses and chamber geometry. Sincecentrifugal force does not act alone in producingthe liquid blanket structure, slower liquid velocitiesmay be tolerated for the bulk liquid. Optionalnon-axi-symmetric solid structures could bemounted on the chamber walls to slow the lithium’srate of descent via induced eddy currents.

Conducting liquids flowing through magneticfields can generate large MHD forces opposing

their motion, if a closed path exists for electriccurrent to flow in response to the motion-inducedelectric field. For flow through pipes, these MHDforces can be overcome by using high pumpingpressure, but for free-surface liquid blankets, whichinherently have a low pressure gradient, externalpumping is not effective. The use of injected electriccurrents provides the possibility of compensatingfor some of the MHD effects in free-surface sys-tems. However, the flow described above will needto be highly axi-symmetric to avoid large dragforces from Hartmann layers forming on non-axi-symmetric structures. In addition, the flow mustconform to the shape of the poloidal flux surfacesto a large degree so that surface normal fieldcomponents are avoided as well.

In a variation on the EMR concept, a two-passdesign using hot and cold liquid sublayers may bedesirable to simultaneously achieve high exit tem-perature of the heated lithium while keeping themaximum vapor pressure of the colder plasma-fac-ing liquid lithium surface low. That the flow ishighly laminarized by the magnetic field may be anadvantage here, suppressing the mixing between thetwo streams and allowing them to flow directly ontop of one another. Detailed analysis of this prob-lem is being carried out in conjunction with thetwo-stream GMD research.

6.1. Flow phenomena with injected electric current

Significant forces can be generated in liquidlithium metal without excessive electrical power.The threshold of significance is levitation. Lithi-um’s mass density is about half of water’s, so itsgravitational weight density on earth is about 5000N/m3. With the approximately 5 T toroidal fieldtypical of many tokamak reactor designs, to gener-ate a force field matching lithium’s weight densityrequires a current density of J=rg/B=1 kA/m2

in the lithium, which implies an electric field of 350mV/m and an electric power dissipation of 0.35W/m3. These are modest parameters. At this ‘one-gee’ force-field level, a lithium EMR blanket sur-rounding an ITER-sized plasma would require atotal current of 50 kA, implying a loop voltage of0.01 V, and a power of 500 W. Increasing powerto 1 MW would increase the lithium force field tothe equivalent of 45 times gravity!

Fig. 29. Electromagnetic restraint (EMR) lithium blanket con-cept.


These calculations show that a relatively smallcurrent can easily overcome the effect of gravity.However, there will also be stray currents pro-duced during operation (due to plasma motion)that could very well exceed the purposely gener-ated currents. This fact demonstrates the impor-tance of coupled analysis of the liquid wall andplasma MHD and the potential need for activecontrol of the applied wall current.

6.2. Axi-symmetric LMMHD analyses

If highly conductive liquid metal were flowingin non-axi-symmetric patterns beside a tokamakplasma, MHD effects would produce non-axi-symmetric currents in the liquid. In addition tothe potential to induce significant MHD drag, thiscould produce non-axi-symmetric magnetic fieldswhich would perturb the plasma. Tokamaks andseveral other plasma confinement schemes requireprecisely axi-symmetric magnetic fields to main-tain nested internal flux surfaces. They have verylittle tolerance for departures from axi-symmetryand develop ‘magnetic islands’ which deteriorateplasma confinement at very small levels of non-axi-symmetric magnetic field ‘ripple’. A reactorblanket must therefore avoid doing harm to theplasma equilibrium, so strict axi-symmetry is animportant requirement for the portions close tothe plasma of a highly conductive, fast moving,liquid blanket.

Although exact three-dimensional MHD equa-tions for an incompressible liquid are compli-cated, they can be simplified without anyapproximation for the EMR lithium blanket con-cept by this requirement for axi-symmetry. Inderiving exact axi-symmetric LMMHD equationswith independent variables (r,z,t), it is convenientto express magnetic field via the poloidal magneticflux stream function, C, and the total poloidalthreading current stream function, I (includingany toroidal field coil system current). Formu-lated in primitive hydrodynamic variables, theresult is six time-dependent scalar PDEs, and anODE-integral equation describing the effect of thepower supply voltage. The full derivation of thissystem and associated boundary conditions aregiven in ref. [1].

It is important to note that boundary condi-tions on the surfaces of the liquid and solidmetallic conductors will be required for C. Thesetime-varying boundary conditions depend on theplasma and poloidal field coil currents, whichdepend on the plasma scenario. For the case of noplasma or PF coil currents, the above equationsare closed and are ready to be solved for specificcases. For cases including a plasma and/or PF coilcurrent histories, additional data is needed toconduct an analysis. This magnetic coupling ofthe plasma/liquid wall/magnet coils set is an im-portant feature of this formulation, and in the endwill be required even for passive schemes like theGMD or CLiFF to fully described the liquid wallreaction to electromagnetic plasma events and tocontrol field variations. It should be noted thoughthat galvanic halo currents flowing between theplasma and the liquid conducting surface are notmodeled in this system.

Although greatly simplified from the three-di-mensional case, the above-described equations arenot amenable to direct analytical solutions unlessmany approximating assumptions are made. Thathas not been done, but might perhaps be useful.The equations are amenable to numerical solu-tion. No commercially available simulation codewas identified capable of such a simulation, so thedevelopment of one has been undertaken. How-ever, the code is not complete at the time of thisreport, so no detailed numerical studies of theEMR concept are yet available. Some importantobservations, though, can be made can be madedirectly about the equations.� The toroidal swirl motion should remain iden-

tically zero as long as the poloidal current inthe liquid metal is aligned to follow poloidalflux surfaces.

� If liquid velocity and injected current wereboth aligned to poloidal flux surfaces the veloc-ity along streamlines should be unaffected bythe variables magnetics (I and C).

6.3. Necessary departures from axi-symmetry andkey issues

It is not possible to design an entirely axi-sym-metric blanket system since the flowing liquid


must cross between structural supports at somelocation, and in most versions of the concept needto exit and reenter the TF coil region. Analyses ofthese non-axi-symmetric regions will be morecomplex. There may be significant MHD pressurelosses and pumping problems in the non-axi-sym-metric regions.

The key issues with the EMR lithium blanketconcept all are based on the difficulty of predict-ing its performance. At the present time, there areno computer tools or other methods to designsuch a system although several efforts have beeninitiated and continue this year.

7. Effects of liquid metal walls on plasmaperformance

The interaction of liquid walls with the plasmacore is a complex topic that requires future stud-ies. In this section, we address some potentiallyfavorable effects of flowing liquid metal walls ontokamak plasma performance and reactorattractiveness.

Liquid metal walls have been considered intokamaks primarily for heat flux and radiationprotection, and to modify particle recycling. Inaddition, it is clear a priori that liquid metal wallscould in principle act as a close fitting conductingshell, but the advantages of this have not beenexamined. Here, we describe how this can lead tohigher plasma b values and improvedconfinement.

The stabilizing effects of the liquid metal can beeither passive (merely due to the presence of anearby conductor), or active (due to the flow ofthe liquid metal). The passive effects are signifi-cant because liquid metals such as lithium can becloser to a reactor plasma, as well as thicker, andthus more stabilizing. The active effects are im-portant because they can prevent flux penetrationin steady state, preventing resistive wall modes bynaturally converting liquid metal kinetic energyinto magnetic flux to compensate for resistivelosses. It is widely recognized that resistive wallmodes strongly limit performance in advancedtokamak operation, and also seriously affect re-versed field pinches (RFPs) and other toroidal

confinement devices. We consider both passiveand active effects here.

7.1. Passi6e stabilization by LM walls

In reactor studies such as ARIES-RS [14], pas-sive stabilizing conductors are placed behind theblanket. These conductors must maintain atoroidally continuous conduction path to stabilizethe vertical instability. They are placed behind theblanket because structural metals compatible withthe fusion environment are poor conductors, anda thickness to provide substantial conductivitywould negatively affect tritium breeding if placedin front of the blanket. Also, the degradation ofthe conductivity of such metals due to radiationdamage is a problem. In particular, radiationdegradation of joints and welds may jeopardizethe required toroidal continuity. The removal ofthe stabilizing plates by this significant distancefrom the plasma, substantially reduces their stabi-lizing effect, limiting tokamak reactor designs toan elongation k of approximately two or less. It iswell known that the maximum plasma current is astrong function of elongation, and thus, so is theattainable MHD b as well as the confinementpredicted by scaling laws.

In contrast, molten lithium metal can be placedclose to the plasma since it does not degradebreeding (in fact it improves it). Furthermore, theconductivity of a liquid is unaffected by the radia-tion environment. Liquid plasma facing designsconsidered by APEX have lithium much closer tothe plasma. Alternatively, a liquid lithium vesselcould be placed just behind a solid first wall(maintaining a toroidally continuous conductionpath). Below we consider the effects of this on k

and thus on b.An n=0 vertical resistive stability code has

been written. It solves the perturbed Grad–Shafranov equation D�DC= (FF%)%DC as an ini-tial value code including inductive fields andresistive elements. The elliptic operator is invertedwith vacuum boundary conditions, and includesthe effects of external resistive coils, a resistivewall, and active feedback coils with voltages de-termined from the signals of sensor coils.Presently pressure is not included in the equi-


librium, and the toroidal current profile FF ’ istaken to be a constant. The voluminous litera-ture on vertical stability [18] shows that plasmapressure is not a major effect (and is usuallystabilizing), and hollow current profiles (as ex-pected with high bootstrap fraction operation)are expected to be more stable than a flat cur-rent profile. Thus, the results below are morepessimistic (and thus conservative) than expectedfrom more realistic profiles.

As examples, we consider the vertical stabilityof a k=3 plasma with aspect ratio A=3 and 4,with 4 cm of lithium (a typical number for thinliquid plasma facing concepts), and the liquid ata distance b/a=1.2 (i.e. a distance from theplasma of 20% of the horizontal minor radius,or about 30 cm for ARIES-RS). Liquid facingconcepts usually have liquid closer than this,and would be more stabilizing. The resistivewall time is roughly 0.5 s, and the resistive ver-tical instability growth time is about 0.66 s. Thistime scale is easily within the reach of existingvertical feedback technology (which can have re-sponse times of the order of a millisecond, orslightly less). With a standard feedback geome-try with the active coil above the plasma a dis-tance which would place it behind a 1 m shield,and a sensor coil on the outboard side (thougha distance which would place it inside the shieldbut behind the first wall), vertical stability isachieved with feedback gain about an orderof magnitude larger than in the case k=2,and with feedback response times 050 ms.This appears to be within the range of presenttechnology. Little effort has been spent optimiz-ing the parameters of the feedback system,and considerable improvement might be possi-ble.

We find the consequences of this to the at-tainable b in advanced technology (AT) modesare large. The MHD equilibrium code TOQ[19] (used routinely by the general atomics (GA)� MHD group) has been used, to obtain highbootstrap fraction equilibria for A=3 and 4.Broad pressure profiles are used which havebeen used by the GA group in b optim-ization studies for A=1.4 tokamaks. The maxi-

mum b for ballooning stability for A=4 and 3is:

bN=4.5b=5–7% S=7.3k=2bN=5.7k=3 S=13.9b=20–22%

As can be seen, the stable b is increased byabout a factor of 3. Note bN does not increasemuch, so the increased b is mainly due to in-creased current. The b and bN values for k=2are quite similar to those found in the ARIES-RS study (k=1.9, bmax=5.4%, bN,max=4.8).We do not have capabilities to examine n=1stability, so we estimate stability based on theshape factor S= (1/aB)qedge. With wall stabiliza-tion, the maximum stable bN is an increasingfunction of S and profile flatness p(0)/�p�. Ifwe extrapolate published results by Turnbull etal. [20], we infer that the much higher shapefactor for k=3 should enable n=1 stability forthe modestly higher bN value.

This has large consequences for a reactor. Fora 1 GW reactor with 1 m of inboard blanket/shield and 13 T superconductors (and the sameb as ARIES-RS for k=2):

k Majorb MW/m2 r* H-factor(ITER89P)R

1.9 5.54.8% 4 1/500 1.83.15 9.5 1/18018% 1.63

As can be seen, there is a large reduction insize and therefore mass and cost. For example,the length of superconducting wire needed is re-duced by about a factor of 2.5. The wall load-ing is in the range considered as the nominalcase for APEX design evaluations of advancedwall concepts (8 MW/m2).

Note that the r* of the k=3 reactor isthe same as JET and JT-60. Thus, this reactoris not an extrapolation in r*, but rather ingeometry. Since geometry is not a fundamentalphysics variable, we expect that extrapolat-ions in k from existing machines can be madewith much less uncertainty than extrapolationsin r*.


7.2. Acti6e stabilization of resisti6e wall modes

We now consider the effects of a flowing wallon the n=1 resistive wall instability. We employa self-consistent limit of the MHD equations toobtain an analytically solvable model of the b

driven external kink mode. The model uses highA, reduced MHD, simplified with flat current andpressure profiles and circular plasma cross-sec-tion. We note that independently, a similar modelwas investigated by Betti et al. with similarresults.

We obtain a b driven kink mode, which re-quires coupling between adjacent poloidal modenumbers m for instability. The mode can be stabi-lized with an ideal (perfecting conducting) wall.Finite resistivity and rotation are added numeri-cally, using the analytic plasma response. As ex-pected, with no rotation there is a resistive wallmode with gRWM � the resistive wall time. For apoloidally rotating wall (which adds a currenthdj=v0×d B=v0 d Br in addition to the induc-tively driven current), we find stabilization whenthe poloidal transit time for the flow to go fromthe top to the bottom 1/tp=v0/p r is fast enoughthat:

1tp

]gRWM.

This result has also been found independently byBetti. Here, we note that for 4 cm Li, this corre-sponds to velocity levels considered by the APEXdesign for liquid metal walls. Note that it is notnecessary for the flow to be facing the plasma, butrather the flow could be in a cavity behind a solidfirst wall (but close to the plasma).

This stabilization can be interpreted as an in-ability of the n=1 flux to penetrate if the metalflows from the top to the bottom more rapidlythan the growth rate, since then the metal isalways being replaced by fresh metal. Alterna-tively, the result can be interpreted as thedephasing of a toroidal instability which requiresa particular phase relationship between differentpoloidal harmonics. Since each m number isDoppler shifted by a different amount, there isnot rotating frame where the relative phasesneeded for instability can be maintained.

Note that stability requires that the conductingwall be placed somewhat closer for stability thanis the case for a perfectly conducting wall. This isdue to the fact that the mode can rotate with afrequency to remove wall stabilization for a singlepoloidal harmonic. Since only two of the threeharmonics are wall stabilized, the stabilization isnot as effective. However, in more realistic shapedequilibria, there is a much broader spectrum of mnumbers required in the eigenfunction than in thiscircular model. Since only one among the largenumber of harmonics can escape wall stabiliza-tion, we anticipate that shaped equilibria will haverotational stabilization effectiveness more nearlyequivalent to that of a perfectly conducting wall.

Stabilization of resistive wall modes would leadto several benefits. Higher b steady state equi-libria could be obtained, with very hollow currentprofiles. Steady state operation with such profilesenables high bootstrap fractions and thus lowrecirculating power. Also, hollow current profilesare theoretically predicted to give E×B shearingrates larger than instability growth rates for con-ventional drift instabilities, leading to transportbarriers and high confinement. Hollow currentprofiles are well correlated experimentally withsuch good confinement. Thus, flowing liquid wallsmay enable the conditions needed for high steadystate confinement.

We note that the codes used to obtain theabove results are being developed further to han-dle arbitrary equilibria output from an equi-librium MHD code. We also note that flowingliquid metals can behave differently under pertur-bations since they can be pushed out of the way.Analysis indicates that liquids flowing at thespeeds indicated above are not greatly affected bythis (though a stationary liquid would be), butinclusion of this effect is also in progress. This lastpoint especially indicates that there is a synergismbetween both liquid metal walls and tokamakphysics performance, both in b and confinement,as well as in the analysis of dynamics of plasmadischarges and flowing wall behavior.

We recommend that this synergism be pursuedvigorously through cooperation between the fu-sion physics and the fusion engineeringcommunities.


Fig. 30. Impurity concentration limits for different impuritiesdue to radiation loss in a tokamak (solid line) and due tosimple fuel dilution (dashed line).

A multi-faceted, self-consistent model is re-quired to make a complete evaluation of theinteractions between the edge-plasma and the liq-uid walls. We have made substantial progress indeveloping components of this general model andin using these components for initial evaluation ofsome of the critical issues. The progress is summa-rized below and presented in more detail in Ref.[1] for the following areas: two-dimensional fluidtransport simulations of properties of the hydro-genic edge plasma; two-dimensional fluid trans-port simulations of impurity penetration to thecore region arising from evaporating Flibe andLi-based walls; one-and-a-half-dimensional ki-netic and two-dimensional fluid transport calcula-tions of evaporated and sputtered impurities fromliquid divertor plates; two-dimensional simula-tions of intense power deposition to a lithiumdivertor plate during a disruption; one-and-a-half-dimensional plasma core transport modeling, be-ginning simulations of the behavior of smallliquid samples in the PISCES plasma divertorsimulator and the DIII-D tokamak.

8.1. Edge fluid transport simulations

We have used the two-dimensional UEDGEcode [22] to obtain profiles of hydrogen ion den-sity, parallel ion velocity, and separate ion andelectron temperatures. The base-case is an ITER-like tokamak where the transport simulation setsboundary conditions of power and density a smalldistance inside the magnetic separatrix and calcu-lates the resulting scrape-off layer (SOL) profiles.We have characterized two-dimensional plasmasprofiles for both high-recycling regimes (Flibe orother non-recycling divertor) and low-recycling(lithium divertor which retains incident hydro-gen). The low-recycling case results in high elec-tron temperature at the divertor and low density,with the opposite being true for high recycling.An important consideration for the low-recyclingcase is the large particle flux out of the core thatmust be maintained by an edge particle-fuelingsource such as pellets.

To assess the effectiveness of the edge plasmafor shielding the core from impurities, theUEDGE calculations are extended to include im-

8. Plasma–liquid surface interactions and edgemodeling

The thin layer of edge plasma provides theinterface between the hot-plasma core and theliquid first-walls and divertor plates. The edge-plasma properties must be accurately determinedto predict the coupling between the core plasmaand the wall, and the edge-plasma itself is affectedby both the core plasma and the wall. Liquidsurfaces can impact the edge and core plasmas byreleasing impurities through sputtering, recycling,and evaporation. Such impurities degrade fusioncore performance through enhanced radiation lossand fuel dilution. The tolerable levels of coreimpurity concentration owing to radiative energyloss [21] and to fuel dilution are shown in Fig. 30for a tokamak. Changes in the edge plasma tem-perature and gradient scale-lengths can also affectthe stability of the core-edge plasma, e.g. the L-Htransitions, ELMs, and possibly disruptions.

The edge plasma, in turn, influences the liquidsurfaces through particle bombardment and lineradiation from excited ions. The bombardmentleads to sputtering and recycling, and both bom-bardment and radiation heat the surface resultingin increased evaporation. The maximum tolerableevaporation rate determines the maximum allow-able surface temperature of the liquid, and thesputtering analysis determines the required edge-plasma.


purity gas evaporating from the liquid wall. Anumber of processes are included in this model-ing. The impurity gas is emitted from the wall inthe form of atoms at typically 1 eV, although arange of energies have been used to assess theenergy of the atoms after molecular dissociationwhich is not yet modeled in any detail. Theseneutrals diffuse by elastic collisions with ions untilthey are ionized by the electrons of the edgeplasma. Once an ion, the impurity diffuses acrossthe magnetic field with anomalous diffusion co-efficients estimated from present experimentaldevices. Thus, the ions can diffuse radially intothe core or back to the liquid wall where they areassumed to be absorbed. In addition, the ions canflow along the magnetic field and out of thesystem. The electron energy lost by ionizing theimpurities through all of their charge states isincluded, so that the impinging impurities de-crease the electron temperature, especially nearthe liquid surface. A typical set of charge-stateprofiles from fluorine from a Flibe wall are shownin Fig. 31.

Similar calculations have begun for Sn–Li wallswhere only Li is evolved from the surface; it isassumed that evaporation of Sn is negligible.Lithium penetrates less easily to the core due, inpart, to its lower first-ionization potential of 5.4 V

Fig. 32. Comparison of fluorine and lithium densities at thecore boundary for different gas fluxes at the first wall. Thedotted lines are extrapolations owing to non-steady solutionswhich arise with a collapse of the electron temperature just infront of the wall for larger gas fluxes.

compared to 17.3 V for fluorine from Flibe. Sec-ondly, if one considers a Sn–Li, its evaporationrate is less than that of Flibe at a giventemperature.

The comparison between the fluorine (Flibe)cases and the lithium (Li, Sn–Li) cases with re-spect to impurity concentration is shown in Fig.32. This figure quantifies what core impurity coredensity should be expected for a given gas flux,which can be determined from known data of theevaporation rate at a given liquid surfacetemperature.

From Figs. 30 and 32, one can deduce that foran ITER-like tokamak with 150 MW of plasmapower flowing into the scrape-off layer, impuritypenetration to the core may be kept to an accept-able level if the liquid surface temperature forFlibe is 540°C or less, while for Sn–Li it is 740°Cor less. However, these results are quite prelimi-nary with one of the most important uncertaintiesbeing the fact that the transport simulations havenot yet found steady state solutions at the largergas flux regions of Fig. 32 shown by the dottedlines. These dotted line portions of the curves arejust those being used to make the estimates of themaximum surface temperature quoted above.

Fig. 31. Density of fluorine charge-states at the outer midplanefor a gas wall flux of 8×1018 m−2s−1 in the standardtokamak geometry.


Thus the highest priority of our present researchis to better resolve and understand the non-steadysolutions. Such solutions correspond to where theelectron temperature near the surface abruptlydrops below a few eV owing to impurity radiationand particle energy losses to the wall. This is a‘detached’ type of plasma, but here the detach-ment is from the side wall rather than the divertorplate.

The calculations for impurity influx from sidewalls have been mostly performed in a tokamakgeometry. More simulations are needed for alter-nate confinement geometries such as the field-re-versed configuration (FRC), spheromak, sphericaltorus, and others.

8.2. Kinetic simulations of the sheath andpresheath

Kinetic simulations are performed for the re-gion near liquid divertor plates using the test-par-ticle codes BPHI and WBC codes [23] with MonteCarlo collisions. BPHI focuses on the sheath re-gion, including ionization within the sheath,whereas WBC uses a reduced sheath model andincludes the presheath region �10 cm in frontthe plate. Both codes begin with a hydrogenplasma from a two-dimensional fluid transportcode, but then trace sputtered and evaporatedimpurities from the plates made of Flibe orlithium until they escape upstream or are rede-posited on the plates.

For the WBC code lithium analysis, the follow-ing is observed: (1) very high near-surface lithiumredeposition rate (�100%), (2) high redepositedaverage energy with highly oblique Li ion im-pingement. Result (1) is favorable showing lowpotential for plasma contamination by sputteredlithium, even for the low-collisionality, low-recy-cle regime. Result (2) gives rise to concerns aboutrunaway self-sputtering although preliminary esti-mates using initial ALPS/APEX project datashow that this will probably not occur.

WBC calculations for Flibe assessed the near-surface transport of the individual sputtered Flibeconstituents of F, Li, and Be. As with the lithiumsurface calculations, a highly preliminary sputter-ing model was used. Results using a hydrogen

plasma in the high-recycle regime (Te=30 eV,ne=3×1020 m−3) show a high redeposition frac-tion for each element. There is a lower potentialfor self-sputtering runaway due to lower redeposi-tion energies and less oblique incidence.

BPHI sheath code calculations were performedfor a low-recycle plasma divertor regime with alithium surface. Preliminary results, for one par-ticular low-recycle regime, show that a majority ofslow-moving, evaporated lithium atoms will beionized in the sheath and will be returned to thesurface due to strong sheath electric field. On theother hand, the sheath heat transmission factorwill increase due to reduced sheath potential re-sulting from the extra electrons and ions producedby in-sheath ionization. The resulting increase inheat flux is of concern in terms of a runawayeffect but this may be mitigated by the transientnature of the overheating and the fact that thelithium is flowing.

8.3. Additional on-going edge plasma simulationwork

A self-consistent sputtering erosion/redeposi-tion analysis of a lithium divertor surface isplanned, using coupled UEDGE/WBC/VFTRIM(plasma SOL fluid code/Monte Carlo kinetic im-purity code/vectorized fractal-TRIM sputteringcode) codes. This will better compute plasma con-tamination potential, tritium codeposition, andself-sputtering runaway potential.

Another important question is the response of aliquid divertor plate to a tokamak disruption. Anumber of physical processes have been includedin the HEIGHTS package [24] and simulationsperformed for a liquid lithium plate. The incom-ing power to the plate is taken as 100 GW/m2

which is typical of what would be expected in areactor-sized tokamak. As this high particle en-ergy strikes the plate, material is ablated in theform of a gas vapor, which is subsequently ion-ized by the incoming electrons. The energy re-quired for ionization of the vapor can decreasethe incoming energy to the plate by an order ofmagnitude to less than 10 GW/m2 while thispartially ionized vapor cloud becomes opticallythick. An additional reduction of the power to the


plate comes from the splashing of plate materialinto droplets due to Kelvin–Helmholtz or Ray-leigh–Taylor instabilities in the vapor. The powerloss in vaporizing these droplets can result inanother factor of 5 reduction in power reachingthe plate. The mass loss of the liquid lithium platecan likewise be reduced by about two orders ofmagnitude from the combined shielding of thevapor and the splash droplets. As a result, theeffect of a disruption on the lithium plate is notthought to be limiting. Further assessment isneeded to determine how the incoming disruptionpower, which is initially absorbed by the vaporand droplets but then re-radiated, affects nearbystructures. Also, the vapor and splashing thatresult from the disruption will migrate to othersurfaces in the machine. If all surfaces are movingliquids, they will self-clean; and using the sameliquid for the plate and the walls will eliminate theproblem altogether.

The impact of different edge-plasma conditionson the performance of the fusion core plasma isbeing studied with the one-and-a-half-dimensionalcore transport code ONETWO [25] which hasbeen used extensively for analyzing DIII-D exper-imental results. As an initial case, an ITER-liketokamak is being considered with a 20 keV oper-ating point since a lot of previous analysis hasbeen done on this configuration which provides agood simulation benchmark. The effect of thelow-recycling edge conditions using lithium plateswill be contrasted with the normal high-recyclingedge (which would likely arise if Flibe were used).Given this background, a similar analysis will beperformed for the ARIES-RS design.

Finally, it is important to benchmark modelspredicting how liquid surfaces emit impurities inthe presence of plasma discharges, and how theimpurities transport in the plasma. At present,small samples of lithium and gallium have beenused in the linear plasma device PISCES, andlithium has just been used on the DiMES probefor the DIII-D tokamak. Sputtering data is alsoavailable from particle beam measures on theUniversity of Illinois experiment. The sputteringdata from these various experiments are beingtabulated and will be used as input for the fluidand Monte Carlo codes which follow the subse-

quent ionization and transport of the impurityions. A challenge to impurity transport modelingfor the DiMES probe is that the probe is localizedto one toroidal location, so three-dimensional ef-fects do enter which can only be estimated by thepresent codes. Nevertheless, these calculations be-gin the vital process of comparing modeling re-sults with experimental data. Larger-scale liquidsamples in experiments will improve this bench-marking. There is on-going work to use liquiddivertor surfaces in other devices such as CDX-U.This type of activity is important to provide theexperimental data base to validate models predict-ing the influence of such walls in fusion-relateddevices.

9. High-temperature solid wall with lithiumevaporation (EVOLVE)

This section discusses a novel method to extendthe capabilities of a solid wall by using a high-temperature refractory alloy with heat extractionachieved by lithium evaporation.

The desire to achieve both high power densityand high power conversion efficiency leads toseveral required features of a first wall and blan-ket concept. Achieving high power density meansthat the coolant heat removal capability must behigh and the first wall material should have at-tractive thermophysical properties (high thermalconductivity, low thermal expansion, etc.).Achieving high power conversion efficiency meansthat the first wall and blanket should operate atvery high temperatures. Materials operating atvery high temperatures generally have limitedstrength and, therefore, such a concept shouldoperate at low primary stresses. This means thatthe coolant pressure should be as low as possible,and the temperatures throughout the blanketshould be as uniform as possible to reduce ther-mal stresses.

One system that has this potential is theEVOLVE (evaporation of lithium and vapor ex-traction) concept. The key feature of theEVOLVE concept is the use of the heat of vapor-ization of lithium (about ten times higher thanwater) as the primary means for capturing and


removing the fusion power. A reasonable range ofboiling temperatures of this alkali metal is 1200–1400°C, corresponding with a saturation pressureof 0.035–0.2 MPa. Calculations indicate that anevaporative system with Li at �1200°C can re-move a first wall surface heat flux of \2 MW/m2

with an accompanying neutron wall load of \10MW/m2. The system has the followingcharacteristics:1. The high operating temperature translates nat-

urally to a high power conversion efficiency.2. The choices for structural materials are limited

to high temperature refractory alloys. A tung-sten alloy, e.g. W–5%Re, is the primary candi-date as a structural material, with tantalumalloys as the back-up.

3. The vapor operating pressure is very low (sub-atmospheric), resulting in a very low primarystress in the structure.

4. The temperature variation throughout the firstwall and blanket is low, resulting in low struc-tural distortion and thermal stresses.

5. The lithium flow rate is approximately a factorof ten slower than that required for self-cooledfirst wall and blanket. The low velocity meansthat an insulator coating is not required toavoid an excessive MHD pressure drop.

The areas addressed are first wall and blanketdesign, tritium breeding, activation and waste,power conversion, first wall thermo-mechanicalbehavior, tritium extraction, and critical issues.The key features of the design are summarized inTable 11.

The cross-section design of the EVOLVE con-cept is illustrated in Fig. 33. In the EVOLVEconcept, the first wall and primary breeding zoneare combined into one unit. Behind this unit,there is as a separate component, a high tempera-ture shield at the inboard region and a secondarybreeding blanket at the outboard region. Behindthe secondary breeding zone there is, as a separatecomponent, an additional high temperatureshield, required in order to meet the shieldingrequirements of vacuum vessel and magnets.

The first wall consists of a tube bank arrangedin the toroidal direction as shown in Fig. 34.Within each tube is another tube that supplies theliquid lithium to the first wall. There are twodifferent methods under consideration for the dis-tribution of the liquid metal at the surface. One ofthem employs a large number of jets generated bynozzles in the supply tube by which the LM isdistributed to the backside of the first wall. Withthe other one, capillary forces in a wick structure,arranged at the backside of the first wall, areemployed to transport the liquid lithium from thesupply tube to the entire surface of the first walltube. This wick is connected to the supply tubevia longitudinal slots in this supply tube. For asurface heat flux of 2 MW/m2, a toroidal segmentwidth of 3 m, and the tube dimensions givenabove, a boiling temperature of 1200°C (satura-tion pressure 0.035 MPa) results in a liquid metalvelocity in the feed tube of about 1 m/s and avapor velocity of about 500 m/s. This is aboutone-third of the sonic velocity and results in atolerable pressure drop.

The blanket consists of a number of trays,stacked poloidally, containing liquid lithium. Aspace is left between trays to allow the Li vapor tobe removed from the blanket. Each tray containsa lithium pool with a height of 10–20 cm, whichis maintained constant by a system of overflowtubes. The large volume heating of the lithiumleads to boiling. The vapor bubbles have to rise inthe pool and separate from the liquid metal at thesurface. From here the vapor flows a short dis-tance in parallel to the surface before it enters thevertical vapor manifold. Entrained liquid metalwill be separated there. Behind the trays is amanifold, approx. 20 cm thick, for collecting the

Table 11Key features of the EVOLVE concept

ValueFeature

Heat capture and removal Li vaporLi vapor pressure 0.035 MPaLi vapor velocity �500 m/s

TungstenStructural materialOperating temperature �1200°CFirst wall heat flux 2 MW/m2

Neutron wall load 10 MW/m2

Tritium breeding ratio (local 1.37two-dimensional)

�57%Power conversion efficiency


Fig. 33. Cross-sectional view of the EVOLVE first wall/blanket concept.Fig. 34. Schematic of EVOLVE first wall tubes and blanket trays containing Li.


Li vapor. The total radial thickness of the firstwall and blanket is approx. 70 cm.

Two-dimensional neutronics modeling of thefront evaporation cooled blanket of EVOLVE isneeded to properly account for the poloidal het-erogeniety and gaps between trays. The R–Zgeometrical two-dimensional model used in thecalculation includes the FW, trays with Li vapormanifold, secondary breeding blanket, shield, VV,and magnet in both the IB and OB regions. Boththe IB and OB regions are modeled simulta-neously to account for the toroidal effects. TheTWODANT module of the DANTSYS 3.0 dis-crete ordinates particle transport code system wasutilized. The overall TBR calculated for the refer-ence design using the two-dimensional model is1.37. It is based on the conservative assumptionof no breeding in the divertor region. Tritiumbreeding (69.8%) occurs in the trays (57.3% OBand 12.5% IB). The OB secondary blanket con-tributes 27.7% of the total overall TBR (20.2%behind trays and 7.5% between trays). The contri-bution of the shield is only 2.5% (1% OB and1.5% IB). Tritium breeding has a comfortablemargin that allows for design flexibility.

There are two coolant streams exiting from theblanket. The front part of the blanket, includingthe first wall and the primary breeding zone, iscooled by boiling lithium, which carries approxi-mately two-thirds of the total thermal power. Theback part of the blanket, composed of secondarybreeding zone and the high temperature (HT)shield at the outboard zone and the HT shield atthe inboard zone, is a conventional self-cooledliquid lithium blanket with an exit temperature ofalso 1200°C, which carries the other one-third ofthe thermal power. The two blanket coolantstreams will be fed to two heat exchangers totransfer the thermal energy to a helium loop. Thereason that He is used for the secondary coolantis that a closed cycle gas turbine can be used forvery efficient power conversion. The two lithiumstreams exit from the blanket operates in series,with the liquid lithium stream to heat up thesecondary He from 700 to 800°C, while the hightemperature lithium vapor super heat the same Hestream from 800 to 1000°C. The He at 1000°Cwill enter a He turbine for power conversion.

With a very high He temperature, and very highrecuperator, compressor and turbine efficiencies, avery high cycle efficiency of 57.7% is calculated.This thermal efficiency includes the pumpingpower of the secondary He stream, but does notinclude the pumping power of either of thelithium streams, which is very small in any case.

Finite element thermal and stress analyses havebeen performed for the first wall subjected tosurface heat fluxes of 1.5 and 2 MW/m2, a coolanttemperature of 1200°C, and a coolant pressure of0.05 MPa. A single tungsten tube of radius 2 cmand wall thickness of 3 mm deforming undergeneralized plane strain condition is considered.The primary membrane stress in the EVOLVEfirst wall is so low (B1 MPa) that neither low-temperature nor high-temperature ratchetingshould be a limiting criterion for the surface heatflux. The peak surface heat flux will be controlledeither by creep-fatigue (which is not consideredhere) or possibly by brittle fracture (due to he-lium-embrittlement). The temperature distributionfor a peak surface heat flux of 2 MW/m2 and aheat transfer coefficient of 40 000 W/m2/°C showsa peak temperature of 1317°C. The peak stressintensity is 158 MPa, which easily satisfies theratcheting limits. Very little ductility is needed tomaintain the allowable stress limit at a high value.For example, if the uniform elongation remainshigher than 2% or the reduction in area at failureis \1%, then the allowable stress is \300 MPa.A stress of 150 MPa would be allowable even forcompletely embrittled tungsten at 1200°C.

The EVOLVE concept is at an early stage ofevaluation. At this stage, it is important to assessthe potential of the concept, identify crucial is-sues, and to define needed R&D work to resolvethose issues. The critical issues to be addressed inthe near future are:1. Will the backside of the first wall remain

wetted under all conditions?2. Will the vapor generated in the stagnant boil-

ing pools of the primary breeding region sepa-rate fast enough from the liquid metal?

3. Will the liquid metal overflow system workand lead to equal liquid metal pressure in eachtray?


4. Is it possible to fabricate entire blanket seg-ments of tungsten or tungstenalloys in spiteof their low ductility and their limitedweldability?

5. How will the structural material behave underintense neutron irradiation?

6. Will the high after heat in tungsten cause asafety problem in case of a LOCA?

10. High-temperature solid wall with heliumcooling

This section explores extending the capabilitiesof a solid wall using high temperature refractoryalloy cooled with high-pressure helium. A primarymotivation is to explore the possibility of using ahigh-temperature helium for high-efficiency en-ergy conversion in a gas turbine cycle.

10.1. Material selection and compatibility

The material selection and compatibility issuesare discussed in Section 12. Pure tungsten ortungsten alloyed with �5% Re (to improve fabri-cability) appear to be suitable candidates. Theunirradiated mechanical properties of tungstenare strongly dependent on thermomechanical pro-cessing conditions. The best tensile and fracturetoughness properties are obtained in stress-re-lieved material. In order to be conservative, sincedata are not available on the possibility of radia-tion-enhanced recrystallization of W, and also toaccount for the presence of welds in the structure,the preliminary design is based on recrystallizedmechanical properties. There are no known me-chanical properties data on tungsten or tungstenalloys at irradiation and test temperatures above�800°C. There are no known fracture toughnessor Charpy impact data on tungsten irradiated atany temperature. Pronounced radiation hardeningis observed in W and W–Re alloys irradiated attemperatures of 300–500°C to doses of �1–2dpa, which produces significant embrittlement intensile tested specimens (�0% total elongation).Simple scaling from existing data on irradiatedMo alloys suggests that the operating temperaturefor W should be maintained above �800–900°C

in order to avoid a significant increase in theductile-to-brittle transition temperature (DBTT).The upper operating temperature limit for tung-sten will be determined by thermal creep, heliumembrittlement, or oxide formation issues. Thethermal creep of W becomes significant at temper-atures above �1400°C. Helium embrittlementdata are not available for tungsten; however,based on results obtained on other alloys, heliumembrittlement would be expected to become sig-nificant at temperatures above �1600°C (�0.5melting temperature, TM). The formation ofvolatile oxides is another potential problem intungsten at temperatures above �800°C espe-cially during an air ingress event. However, if theoxygen partial pressure in the helium coolant canbe maintained at or below 1 appm, then the rateof corrosion is calculated to be less than 2 mm/year for temperatures up to �1400°C. In sum-mary, the selected upper temperature limit fortungsten in the structure of the preliminary designHe-cooled system is 1400°C depending on theapplied stress.

10.2. He coolant impurity control

Refractory metals like W, Mo, and V are sensi-tive to grain boundary oxidation and embrittle-ment. However, if the oxygen (including H2O,CO2, CO, … etc.) partial pressure in the heliumcoolant can be maintained at or below 1 appm,then the rate of corrosion may be acceptable.With the use of Brayton cycle as the power con-version system (PCS), without the need of usinghigh temperature water as the secondary coolant,the ingress of oxygen impurities should be muchlower than the system that uses a high-tempera-ture intermediate heat exchanger. For impurityextraction, several powder metal solid getters havebeen developed. Most are based on zirconiummetal (ZrAl, ZrVFe, … etc.). With these materi-als, hydrogen can be pumped reversibly by tem-perature control. These solid getters will pumpactive gases (oxygen, oxides, N, and CxHy) irre-versibly and have been used on the tokamakexperiment TFTR. In the semi-conductor indus-try, getters have recently achieved the control ofimpurities to a level lower than 1 appb. These are


Fig. 35. Helium-cooled first wall and divertor design module.

commercial modular units with no moving partsand are self-monitoring in design.

10.3. Mechanical design and reliability

Several first wall and blanket system configura-tions were evaluated. The mechanical design isshown in Fig. 35. The helium-cooled refractoryalloy design includes a high temperature helium-cooled first wall and a lithium bath that is alsocooled with high temperature helium.

The first wall is made up of separate unitswhich, in this case, are connected to separatecooling manifolds at the back of each module.The first wall units consist of multiple parallel

passages connected through an integral manifoldto round inlet and outlet connections. The largemodules contain the lithium in a single volume,with pure lithium in the breeding zone and acombination of lithium and steel balls in theshielding zone. The temperature is relatively uni-form, although there will be some gradients, albeittransient, between the front and back structuralwalls. There are two inboard and three outboardmodules to each of the 16 sectors arranged in thetoroidal direction. The piping is routed in twocircuits. The first circuit includes the first wall andpart of the interior heat exchange tubing. Heliumat 800°C enters the first wall through the supplymanifold and exits into the first wall outlet mani-


fold at 950°C. The helium is then routed insidethe lithium can to the first supply manifold for theheat exchange tubes. The first tube circuitexits into a return manifold at 1100°C. The sec-ond tube circuit is fed at 800°C and exits at1100°C.

One of the primary goals of the APEX study isto increase the availability of fusion reactors byincreasing the mean time between failures and bydecreasing the mean time to repair. To this end,we recommended the approach of sector mainte-nance, modular maintenance for everything andpretested modules for all components.

10.4. First wall blanket thermal-hydraulics designand analysis

10.4.1. Design inputsWith the mechanical design concept described

earlier, we determined the material volume frac-tions and power generation from different FW/blanket zones. We performed iterationcalculations between thermal hydraulics and nu-clear analysis. The normalized volumetric powerdensity for W-alloy as a function of distance xfrom the first wall is approximated by PW(x)=9e−3x W/cc per neutron wall loading in MW/m2.The normalized volumetric power density for Li-breeder is approximated by PLi(x) = 4e−3x W/cc. Other input parameters are:

Reactor power output 2 005 Mwe12 MpaHelium pressure2 528 kg/sHelium mass flow-rate

Helium Tin/Tout 800°C/1 100°CW–5ReStructural material7.49 MW/m2Max. neutron wall loading

Max. surface heat flux 2.16 MW/m2

10.4.2. First wall designThe use of helium as a FW/blanket, divertor

coolant has been proposed in various fusiondesign studies. To handle the high surfaceheat load, extended heat transfer enhance-ments by porous medium and swirl tape wereevaluated.

10.4.3. Porous mediumA porous medium enhances heat transfer from

the wall to the helium thereby reducing the filmtemperature drop and the absolute temperaturesof the first wall. The design activity reported herewas based, in part, upon development activities bytwo small US businesses. One of the companies,Thermacore, Inc., uses a porous medium to en-hance heat transfer. Thermacore designed andbuilt a series of helium-cooled modules that weretested at Sandia and elsewhere [26–29]. One ad-vance in their development of a helium-cooledheat sink was the development of designs thatconnected open axial inlet and exhaust passagesto circumferential flow passages that containedthe porous medium, as shown in Fig. 36. Theother company, Ultramet, Inc., has experience infabrication of refractory materials. Ultramet hasdesigned and built commercial products made ofrefractory metals for rocket nozzles and otherapplications in which they use a metallized foamthat is integrally bonded to fully dense material[29] as shown in Fig. 37. Their experience demon-strates that a tungsten channel with integratedporous medium structure can be fabricated.

10.4.4. Swirl tape first wall designAnother method for extended surface heat

transfer is to use a swirl tape insert. Swirl tapeincreases the heat transfer coefficient by increas-ing the effective flow velocity of the coolant andincreasing mixing. There is a large amount ofreliable data available on this method. However,the corresponding increase of coolant flow frictionfactor has to be accounted for.

For this calculation, the enhancement in heattransfer coefficient is given by, hen=2.18/Y0.09,and the increase in friction factor is given byfen=2.2/Y0.406, where Y is the twist ratio definedby pitch/2*diameter of the tube. Therefore theequivalent heq=h en* h and equivalent friction fac-tor feq= f en* f, where h and f are heat transfercoefficient and friction factor for a simple circulartube, respectively. In the following calculation, weused Y=2.

Using a maximum neutron wall loading of 7.11MW/m2, and maximum surface heat flux of 2.06MW/m2, and the swirl-tube first wall coolant ve-


locity range of 54–62 m/s, the W-alloy maximumtemperature was found to be in the rangeof 1073–1242°C. With simple tubes in theblanket, the W-alloy maximum temperature is1199°C, and the lithium maximum temperature is1228°C.

The first wall and blanket system pressure dropwas also estimated. Including frictional losses,turns, contractions, expansions, and main heliuminlet and outlet pipes, the total pressure drop was

estimated to be 0.61 MPa, which gives a DP/P of5.1%.

10.5. Thermal stress analysis of APEX first walldesign

A ‘ground rule’ of the APEX study was thatstructures should be robust, and specifically, 3mm was taken as a minimum first wall thickness(with some scientists recommending 5 mm). A

Fig. 36. Thermacore circumferential flow design.Fig. 37. Porous Ta implant, diameter is 0.75 in.Fig. 38. Effect of FW/blanket inlet temperature on PCS gross efficiency.


central challenge in the design is to relieve theprimary and secondary stresses that result fromthe high helium pressure, surface heat load andthe related steep thermal gradient in the heatedsurface. The FW is permitted to flex to relievethe thermal strain (bending stresses) form thesurface heat load.

A thermal analysis of a dual-channel FWstructure (without the porous medium included)was performed using two-dimensional planestrain models (PATRAN/ABAQUS) for a sur-face heat load of 2 MW/m2 and an internalpressure of 10 MPa; the FW was permitted toflex under the heat load. At 1000°C, the maxi-mum von Mises stress is 80 MPa, this is wellwithin the suggested stress limits stated below.Further iteration will be needed for the refer-ence case of 12 MPa pressure but the resultshould not be significantly different. The double-tube wall design will then be incorporated intothe porous medium design in the next designphase.

The thermal stress due to a prescribed tem-perature distribution along a single tube firstwall of the APEX FW/blanket was also ana-lyzed using the COSMOS finite element code.The structural model consisted of two-dimen-sional beam elements interconnected along withthe defined temperature distribution. The firstwall tube has an i.d. of 1.6 cm and an o.d. of2.2 cm. The beam elements representing thelithium case are 0.2×2.2 cm for the inner caseand 3.8×2.2 cm for the outer strong back case.The lithium case is supported by a guide struc-ture attached to the vacuum vessel. It is as-sumed that the guide structure allows freethermal expansion of the lithium case in the ver-tical and radial directions. The following W–5Re alloy properties were taken at 1000°C:Young’s modulus=392 Gpa, Poisson’s ratio=0.267, and coefficient of thermal expansion=3.96×10−6/°C.

The deformed shape and maximum stress dueto the prescribed assigned temperature distribu-tion and boundary conditions were calculated.The tangential thermal growth of the first walltube of 2.0 mm requires that the blanket mod-

ules be installed with 4.0 mm gaps in the coldcondition to prevent contact with one anotherduring operation. The radial thermal growth ofthe plasma facing tube is 4.4 mm. Since we pro-jected that the irradiated W-alloy should betreated more as a brittle than ductile structuralmaterial, we proposed that the stress criteria forevaluating calculated stress intensities for tung-sten materials be taken as one-half the ultimatestress (133 MPa) at 1000°C for welded jointsand two-thirds the ultimate stress (177 MPa)away from joints. Adopting these criteria, theallowable stress at the weld joint due to all loadcombinations is 152 MPa at 1000°C. Since theproposed support structure will allow free ther-mal expansion of the lithium case, only the tem-perature difference between the first wall tubeand lithium case will induce thermal stresses.The maximum thermal stress occurs in the firstwall tube at its junction to the lithium case andis only 6 MPa.

Although the proposed concept for supportingthe blanket induces low thermal stress, details ofhow to implement the support concept will cer-tainly result in higher thermal stresses. Also, thestresses due to dead weight, pressure, and dis-ruption loads have yet to be calculated. Thiswill be performed in the next phase of design.

10.6. Nuclear analysis

Based on the material volume fractions gener-ated, the reference design was determined by iter-ation between the thermal hydraulics task andassessed the impact of W-alloy on the nuclearheating profiles across the blanket and powermultiplication (PM), and on the tritium breedingprofiles and the tritium breeding ratio (TBR). Theimpact of Li-6 enrichment on these profiles andon TBR and PM is also assessed. In addition, weassessed the damage indices, expressed in terms ofDPA, helium, and hydrogen production rates atseveral key locations including the vacuum vessel(VV) and TF coil case. When compared to otherrefractory alloys like TZM and Nb–1Zr, the bestlocal TBR performance is with W and Li breeder.Based on a one-dimensional cylindrical base on


the outboard blanket geometry, the TBR in-creases with Li-6 enrichment and starts to satu-rate at a value of �1.43 when Li-6 enrichmentis �35%. The damage parameters, DPA rate,helium and hydrogen production rate at variouslocations were estimated in the W-alloy design.Compared to the liquid breeder Flibe, liquidlithium is the less effective material in attenuat-ing the nuclear flux at the VV and TF coil by afactor of 6–10.

The radioactive waste characteristics of thedifferent components of the machine were evalu-ated according to both the NRC 10CFR61 [16]and Fetter waste disposal concentration limits(WDR) [15]. According to Fetter limits, the firstwall, module wall, blanket, and transitional zonewould not qualify for disposal as class C waste.As a matter of fact, the W–5Re alloy producessuch a high activity that the first wall wouldhave a WDR that is more than an order ofmagnitude higher than the class C WDR limits.The high WDR is due to the 186mRe, 108mAg,and 94Nb isotopes. Only 186mRe is a product ofnuclear interactions with base elements in theW–5Re alloy.

10.7. Power con6ersion system

The major incentive for employing high-tem-perature refractory alloy FW/blanket with he-lium cooling in this design is to enable directcoupling with a CCGT (Brayton cycle) for highefficiency power conversion. This has the advan-tage of eliminating an intermediate high-temper-ature He/He heat exchanger (HX), which wouldbe a significant technical challenge. However,the potential for tritium contamination in thepower conversion system (PCS) must be ad-dressed, and appropriate design measures mustbe taken to prevent further spread of contami-nation and to facilitate maintenance of PCScomponents. Fig. 38 shows the effect of FW/blanket inlet temperature variation on PCS per-formance for the selected outlet temperature of1100°C. Based on this, the selected gross effi-ciency for the preliminary design is 57.5%.

10.8. Safety

The use of tungsten as the structural materialin this concept poses some safety challenges.Tungsten is a radiologically hazardous materialwith high decay heat, so we must ensure thatthe design is such that long-term accident tem-peratures are low enough that unacceptablylarge amounts of tungsten are not mobilizedduring an accident. Our preliminary calculationsshow that design options exist that result inlong-term temperatures below 800°C. Detailscan be found in the APEX interim report [1].


We have completed the preliminary design ofa helium-cooled refractory alloy FW/blanket de-sign. Many development issues are identified indifferent areas of the design. The following is alist of key issues, grouped by areas, which willhave to be addressed in order to become a vi-able design:

Irradiated and engineer-Materialsing design material prop-erties of W-alloy.Design criteria for W-alloy.Fabrication of W-alloycomponents.Minimum cost of W-alloycomponents includingmaterial and fabrication.Compatibility betweenhelium impurities andW-alloy.Failure rate andAvailabilitymaintenance.External coolant pipingDesignrouting.Structure support to han-dle thermal expansion.High temperature piping.Develop robust high per-formance fusion powercore W-alloy components.


Helium flow control, distri-Thermalhydraulicsbution and stability.First wall and blanket tem-perature management andstartup.

Safety Removal of afterheat dur-ing LOCA and LOFA.W-surface compatibilityPlasma and surface

interaction with high performanceplasma.

11. Gravitational flowing Li2O particulates

One of the concepts considered early in theAPEX study attempts to eliminate the structuralfirst wall by flowing Li2O particulates directlyexposed to the plasma. The concept is calledAPPLE. The Li2O particulate flow system servesas the coolant and breeder. To be able to handlesimultaneously a high neutron wall loading andhigh surface heat flux, the particulate material forthe coolant/breeder must have good thermal con-ductivity and high temperature capability. Thedesirable material properties are:

1. Low vapor pressure at high temperature.2. Low activation.3. Good tritium breeding capability.4. Low electrical conductivity.5. High thermal conductivity.6. Low tritium solubility.

After reviewing the potential candidates of theavailable coolant/breeding material, the solidbreeder Li2O was identified to have good poten-tial to fulfill most of the requirements.

Since the coolant will be facing the plasma,the low vapor pressure requirement becomesvery important. The total vapor pressure overLi2O can be very low. At 1000°C, the combinedvapor pressure of all the possible components isless than 10−5 torr. Therefore, the maximumallowable temperature of the Li2O is set at1000°C. This high allowable temperature leadsto a design with high thermal conversion effi-ciency.

Fig. 39 shows the conceptual design of thesystem. The Li2O particulate will be fed to thereactor system through a feed tube by gravita-tional force. After the particulate enters the re-actor, it will be directed toward the inner (IB)and outer (OB) blanket by a solid baffle, madeby SiC. Upon entering the IB and OB blanketmodule, the Li2O will be divided into two sepa-rate streams. The stream facing the plasma willbe freely dropped by gravitational force, whilethe flow of the stream inside the blanket will berestricted by an opening at the bottom of theblanket module to slow down the flow. It isimportant to reduce the flow velocity of theblanket coolant to achieve a high coolant tem-perature rise for optimum power conversion.

The thermal analysis of the blanket was per-formed, and the parameters are summarized inTable 12.

The tritium breeding and activation have beencalculated. Li2O has very high lithium density,and sufficient tritium breeding can be achieved.With the expected low structural fraction in theAPEX design, the tritium breeding will not be a

Fig. 39. APPLE configuration using baffles.


Table 12Thermal hydraulics parameters for APPLE

Li2OCoolant/breeding material600Coolant inlet temperature (°C)1 000Coolant exit temperature (°C)B10−5Coolant vapor pressure (Torr)

Maximum first wall coolant velocity 5(m/s)

Maximum blanket coolant velocity 1(m/s)

First wall surface heat load (MW/m2) 2Brayton cyclePower conversion system52Power conversion efficiency (%)

Tritium inventory in the blanket (g) 5

One of the key concerns with the particulateflow concept is flow ‘control’, i.e. whether or nota particulate flow can be injected and guidedwithout a plasma-facing wall. In contrast to liquidflow, particulate flow lacks cohesion forces. Ourstudies of particulate flow dynamics have notdefinitely confirmed, nor denied, the existence ofacceptable particulate flow regimes in the complexplasma chamber geometry.

Many issues remain to be resolved for this classof particulate flow concepts. Examples of thecritical issues are:� Cooling of the solid baffle.� Impact of oxygen contamination to the plasma.� Material erosion and attrition issues.� Solid material transport.� Solid to gas heat exchanger design with the

solid in vacuum.� Particle dynamics.

12. Summary of materials considerations anddatabase

12.1. Introduction

The list of structural materials originally con-sidered for the APEX study includes conventionalmaterials (e.g. austenitic stainless steel), low-acti-vation structural materials (ferritic-martensiticsteel, V–4Cr–4Ti, and SiC/SiC composites), ox-ide dispersion strengthened ferritic steel, conven-tional high temperature refractory alloys (Nb, Ta,Mo, W alloys), Ni-based super alloys, orderedintermetallics (TiAl, Fe3Al, etc.), various com-posite materials (C/C, Cu-graphite and othermetal–matrix composites, Ti3SiC2, etc.), andporous–matrix metals and ceramics (foams). Inorder to provide maximum flexibility in the design(and to increase the possibility for significant im-provements in reactor power density), low long-term activation was not used as a defining ‘litmustest’ for the selection of candidate materials.

Due to limitations in resources and time, thematerials analysis for APEX quickly focused onrefractory alloys due to their higher thermal stresscapacity and higher operating temperature capa-bilities compared to conventional structural mate-rials. However, it should be emphasized that

serious issue. Both Li and O are low activationmaterials. The only significant activation productfrom pure lithium is the tritium, which is requiredfor the fueling of the D-T plasma. The activationfrom oxygen is very low. The only other activa-tion products are from the structural materialinside the blanket, and from the shielding materialbehind the blanket. All the structural materialsfor this design have shown to be qualified forclass C waste disposal. The summaries for theneutronics and activation analysis are summarizedin Table 13.

Table 13Summary of neutronics and activation analysis for APPLE

IB blanket thickness 40 cmOB blanket thickness 75 cm

60%Li2O density in the blanketShield composition 80% steel

20% waterTritium breeding ratio 1.215Blanket energy multiplication 1.116

166Peak end of life damage dpa in theshield at 30 FPY: IB

Peak end of life damage dpa in the 26shield at 30 FPY: OB

55 cmShield thickness, IB40 cmShield thickness, OB

VV thickness 10 cmEnd-of-life He at VV, appm 0.40Magnet protection Meet all the

design goalsMaximum class C waste disposal 0.144

rating, 10CFR61, IB shielding


Table 14Costs for simple plate products (1996 prices)

Material Cost per kg

Fe–9Cr steels B$5.50 (plate form)\$1 000 (CVI processing)�$200 (CVRSiC/SiC

composites processing of CFCs)$200 (plate form—average betweenV–4Cr–4Ti1994 and 1996 US fusion program largeheats and Wah Chang 1993 ‘largevolume’ cost estimate)

Nb–1Zr �$100$300 (sheet form)Ta�$80 (3 mm sheet); �$100 for TZMMo�$200 (2.3 mm sheet); higher cost forWthin sheet

whereas the group VI refractory metals (Mo, W)are very difficult to fabricate. A further issue withall of the refractory metals is joining, particularlyin-field repairs. Satisfactory full-penetration weldshave not been developed for W, despite intensiveefforts over a \25 year time span (1960–1985).The main issue associated with fusion zone weld-ing of the group V alloys is the pickup of embrit-tling interstitial impurities (O, C, N, H) from theatmosphere. Experimental studies are in progressin the US to develop satisfactory fusion welds forvanadium alloys.

12.1.2. O6er6iew of thermal stress capabilities of6arious alloys

The key mechanical and physical properties ofhigh-temperature refractory alloys and low-activa-tion structural materials are summarized in Sec-tion 13.3 of the APEX interim report [1]. Athermal stress figure of merit convenient for qual-itative ranking of candidate high heat flux struc-tural materials is given by M=sUkth(1–n)/(athE),where sU is the ultimate strength, E is the elasticmodulus, n is Poisson’s ratio, kth is the thermalconductivity, and ath is the mean linear coefficientof thermal expansion. In addition, temperaturelimits (usually determined by thermal creep con-siderations) can be used for additional qualitativeranking of materials. A rigorous quantitativeanalyses of candidate materials requires the use ofadvanced structural design criteria such as thoseoutlined in Section 13.2 of Ref. [1].

The mechanical properties for recrystallized re-fractory alloys have been used as the referencecase for purposes of APEX designs. Fig. 40 showsthe ultimate tensile strength for several recrystal-ized refractory and high conductivity structuralalloys as a function of temperature. The mechani-cal properties of stress-relieved (non-recrystal-lized) refractory alloys are superior to those ofrecrystallized specimens, with increases in strengthof up to a factor of 2 being typical. However, thepossibility of stress- or radiation-enhanced recrys-tallization of these alloys (along with the likelyinclusion of welded joints in the structure) doesnot allow this strength advantage to be consideredfor conservative design analyses.

conventional materials may work satisfactorily insome of the APEX concepts (e.g. austenitic stain-less steel located behind a thick wall of Flibe).Other promising advanced structural materials(e.g. ODS alloys, intermetallics) should be consid-ered in future analyses.

Numerous factors must be considered in theselection of structural materials, including:1. Unirradiated mechanical and thermophysical

properties.2. Chemical compatibility and corrosion.3. Material availability, cost, fabricability, join-

ing technology.4. Radiation effects (degradation of properties).5. Safety and waste disposal aspects (decay heat,

etc.).Work by the APEX team focused on the first

four items in this list during the initial 18 monthsof the study, and the key findings are summarizedbelow. More details are presented in Chap. 13 ofRef. [1].

12.1.1. Material costs and fabrication issuesThe APEX materials team gathered informa-

tion on the costs of many of the candidate struc-tural materials. This raw material costinformation is summarized in Table 14. The fabri-cation costs for producing finished products ofrefractory alloys (particularly W) is known to bemuch higher than for steels. The group V refrac-tory metals (V, Nb, Ta) are relatively easy tofabricate into various shapes such as tubing,


Fig. 40. Temperature-dependent ultimate tensile strengths of recrystallized refractory alloys and high-conductivity structural alloys.Data from Tietz and Wilson [31], Conway [32], Buckman [33], and Zinkle et al. [34].Fig. 41. The allowable operating temperature range for structural materials based on unirradiated/irradiated mechanical properties,void swelling and thermal conductivity degradation is denoted by the black boxes (see text). Chemical compatibility issues may causea further restriction in the operating temperature window.


The thermal stress figures of merit vary from�57 kW/m for a high strength, high conductivityCuNiBe alloy at 200°C [30] to �2.0 for SiC/SiCat 800°C. Copper alloys are not attractive choicesfor high thermal efficiency power plants due totheir high thermal creep at temperatures above400°C. The low thermal stress resistance of SiC/SiC is mainly due to the low thermal conductivityin currently available composites (primarily dueto a combination of poor quality fibers and im-precise control of the CVI deposition chemistry).The two major classes of low-activation structuralalloys, V–Cr–Ti and Fe–8–9Cr martensitic steelhave figures of merit of �6.4 (450–700°C) and5.4 (400°C), respectively. The refractory alloysoffer some advantage over vanadium alloys andferritic-martensitic steel, even in the recrystallizedcondition. For example, pure recrystallized tung-sten has a figure of merit of M=11.3 at 1000°C,and TZM (Mo–0.5Ti–0.1Zr) has a value of M=9.6 at 1000°C. The alloy T-111 (Ta–8W–2Hf) hasthe best thermal stress figure of merit among the(non-copper) alloys considered, with a value ofM=12.3 at 1000°C.

12.2. Structural design criteria

Most advanced blanket design concepts requirethe first wall to operate in temperature regimeswhere thermal creep effects may be important.Therefore, in addition to the usual low-tempera-ture design rules, high-temperature design rulesmay also have to be applied. We have adopted theITER structural design criteria (ISDC) as a basisfor the design rules to be used in APEX.

Since the design studies under APEX are pre-liminary in nature, only elastic analysis designrules are included. The design rules are dividedinto a high temperature section and a low temper-ature section, depending on whether thermalcreep effects are or are not important. The lowtemperature rules are always applicable. Hightemperature rules are also applied if thermal creepmay be significant. The low temperature designrules include limits associated with necking andplastic instability, plastic flow localization, ductil-ity exhaustion, brittle fracture, ratcheting (cyclicloading), and fatigue. The high temperature de-

sign rules include limits associated with creepdamage, creep-ratcheting, and creep-fatigue.

12.3. Summary of thermophysical properties(unirradiated and irradiated)

Analytical expressions for the temperature-de-pendent mechanical and thermophysical proper-ties for five of the structural materials consideredfor APEX have been derived from least-squaresfits of experimental data (Fe–8–9Cr ferritic/martensitic steel, V–4Cr–4Ti, SiC/SiC, Ta–8W–2Hf, and W–10Re) and documented in Chap. 13of ref. [1]. Radiation-induced void swelling is notanticipated to be a lifetime-limiting issue in therefractory metals due to their BCC structure,although there are insufficient experimental stud-ies to fully establish the void swelling behavior.Radiation hardening and associated embrittle-ment can have a major impact on all of therefractory alloys. The amount of radiation hard-ening at low temperatures (B0.3 TM) is pro-nounced in all of the refractory alloys, even fordamage levels as low as �1 displacement peratom. The amount of radiation hardening typi-cally decreases rapidly with irradiation tempera-ture above 0.3 TM, and radiation-inducedincreases in the ductile to brittle transition tem-perature (DBTT) may be anticipated to be accept-able at temperatures above �0.3 TM (althoughexperimental verification is needed). Very littleinformation is available on the fracture toughnessof irradiated or unirradiated refractory alloys.

12.4. Coolant/structure chemical compatibility

In general, the refractory alloys have very goodcompatibility with the liquid metals and salts ofinterest for fusion applications (Li, Pb–Li, Sn–Li,Flibe). Impurity pickup (O, C, N, etc.) is the keyengineering issue in most cases for refractory al-loys in contact with these coolants as well as forHe-cooled concepts.

Formation of volatile oxides can lead to pro-nounced surface erosion of group VI metals (Mo,W) at elevated temperatures. The evaporation rateincreases rapidly up to �2000 K in both Mo andW. The high-temperature oxidation of Mo and W


Table 15Maximum allowable temperatures of structural alloys (bare walls) in contact with high-purity liquid coolants, based on a 5 mm/yrcorrosion limit (Sn–Li corrosion limits are based on experimental studies conducted with liquid Sn)

Pb–17 Li Sn–Li (Sn)Li Flibe

450°CF/M steel 400–500°C550–600°C 700°C (304/316 stainless steel)V alloy �700°C �650°C ? ?Nb alloy \600°C(\1 000°C in Pb)\1 300°C 600–800°C \800°C

\600°C(\1 000°C in Pb) 600–800°C?\1 370°C ?Ta alloy\1 370°CMo \600°C B800°C? \1 100°C?

\600°C �800°CW \900°C?\1 370°C\800°C? \760°C?�550°C? ?SiC

was analyzed using a thermodynamic model. Ifboundary layer scattering effects are ignored, theevaporation rate exceeds 100 mm/year at �1500K in both materials for 1 ppm oxygen in He at apressure of 10 MPa. Boundary layer effects mayreduce the evaporation rate by several orders ofmagnitude. The calculations suggest that limita-tions on mass transport through the boundarylayer may reduce the erosion rate to less than 10mm/year at wall temperatures up to 2600 K inboth Mo and W. Although the model does nottake into account many of the physical features ofreal wall-coolant interactions, such as roughness,bends, and temperature variations along the flow,it is reasonable to assume that the evaporationrate of W and Mo will be below a few microns peryear, when operated at temperatures as high as1200–1300°C.

Oxygen pickup in the group V metals (V, Nb,Ta) causes matrix hardening, which in turn pro-duces an increase in the ductile-to-brittle transi-tion temperature (DBTT). The matrix oxygencontent must be kept below �1000 wt ppm inorder to keep the Charpy V-notch DBTT belowroom temperature. Due to the high affinity of thegroup V metals for oxygen, it is not realistic toavoid oxygen pickup from non-lithium coolantson the basis of thermodynamics. However, thekinetics of the oxygen pickup can be kept accept-ably low either by maintaining the temperaturebelow �0.4 TM or by keeping the oxygen partialpressure sufficiently low so as to prevent signifi-cant impingement of oxygen on the metal surface.A conservative analysis indicates that an oxygenpartial pressure of �10−10 torr would be suffi-

cient to keep oxygen pickup to acceptably lowlevels in group V metals for expected structuralmaterial lifetimes (10–50 years).

The experimental database on corrosion ofstructural alloys in contact with liquid metals andFlibe was reviewed. The refractory alloys haveexcellent compatibility with liquid lithium up tovery high temperatures. The maximum operatingtemperatures of various alloys in Li, Pb–Li andFlibe is summarized in Table 15. There is a strongneed for experimental data on the chemical com-patibility of the various structural alloys withSn–Li and Flibe although several materials ap-pear to be compatible with these coolants attemperatures of interest for APEX. The refractoryalloys do not appear to have good compatibilitywith Sn–Li.

12.5. Summary and conclusions

The estimated minimum and maximum temper-atures for several of the structural materials con-sidered for APEX are summarized in Fig. 41. Thelower temperature limit is based on radiationhardening/fracture toughness embrittlement(K1CB30 MPa-m0.5) due to low temperature irra-diation. This embrittlement effect would be ex-pected to occur for damage levels above �1 dpa.There is a large uncertainty in the lower tempera-ture limit for radiation embrittlement in W due tolack of mechanical properties data at irradiationtemperatures above 700°C. The upper tempera-ture limit is based on thermal creep considerations(1% creep in 1000 h for an applied stress of 150MPa). Depending on the choice of coolant, this


upper temperature limit could be reduced due tocorrosion issues. On the other hand, even highertemperatures might be conceivable for applica-tions which have very low applied stress. Thecorresponding minimum and maximum tempera-ture limits for Fe–8–9%Cr ferritic/martensiticsteel are �250 and �550°C. The upper tempera-ture limit could be increased by using oxide dis-persion strengthened ferritic steel, which has goodcreep strength to temperatures in excess of 650°C.The recommended minimum and maximum tem-perature limits for SiC/SiC composites are �600°C (due to radiation-induced thermalconductivity degradation effects) and �900°C(due to void swelling concerns), although addi-tional irradiation data are needed to firmly estab-lish these temperature limits.

13. Safety and environment considerations andanalysis

Safety and environmental issues are being con-sidered up front in the APEX study as new ideasand designs evolve so that the goal of safety andenvironmental attractiveness is realized. A com-prehensive safety analysis requires detailed de-signs [35]. Since the objective of APEX is toexplore and evolve new ideas, rather than developdetailed designs, the role of safety analysis issomewhat different. Safety analysis is used in twoways: (1) for screening concepts by looking forsafety issues that could be ‘show stoppers’, i.e.meeting safety guidelines does not look feasible,and (2) for providing guidance to the design ideadevelopers on areas of improvements to enhancesafety and environmental attractiveness.

13.1. LOCA calculations

The initial focus has been on the ability of thedesigns to remove decay heat. The goal here is toensure that temperatures remain below levels atwhich oxidation-driven mobilization becomes un-acceptable. A number of concepts were examinedto determine the ability of the design to removeheat from the plasma-facing surface during anaccident. If surface temperatures are low enough,

mobilization of hazardous material is minimized.The CHEMCON code [36] used in these calcula-tions was developed to analyze decay heat driventhermal transients in fusion reactors.

LOCA calculations were carried out for fourdifferent APEX concepts:1. He-cooled, refractory alloy first wall/blanket

(slowly moving liquid lithium breeder withtungsten alloy structure).

2. APPLE concept (SiC structure with flowingLiO2 particulate breeder; total blanket thick-ness of �40 cm).

3. CLiFF concept (V structure with thin, �2cm, liquid breeder).

4. One of the thick liquid (Pocket) concepts (athick, �50 cm, layer of liquid breeder flowsover a ferritic steel back wall).

The decay heat distributions for the four designsanalyzed are shown in Fig. 42. Note that thedecay heat is shown per unit volume of the zone,including structure, voids, coolant channels, etc.

The optimal result, from a safety point of view,is when long-term accidents temperatures are ade-quately low without relying on active (safety-grade) cooling systems. The initial calculations foreach design assumed no active cooling. If thetemperatures were unacceptably high, variouscooling options were then examined. Peak tem-peratures and the amount of time above 800°Cfor the APPLE, CLiFF, thick liquid wall, andHe-cooled refractory alloy designs are shown inTable 16 (EVOLVE has not yet been analyzed).Because of the large amount of tungsten used inthe He-cooled refractory alloy design, active cool-ing was necessary to keep accident temperaturesto an acceptable level. Similarly, it is primarily theTenelon in the shield that is contributing to thehigh decay heat in the CLiFF design. Activecooling of the vacuum vessel reduces peak tem-peratures to 875°C, however temperatures areabove 800°C for 3.5 days.

The choice of Tenelon (which is a high man-ganese steel; manganese has high decay heat) inthe shield behind CLiFF is independent of theidea of thin liquid wall, and thus can be easilyreplaced by another shielding material. Althoughthe peak temperature during the transient for theAPPLE design is above 800°C, the duration is less


Fig. 42. Decay heat distribution per unit volume for the four concepts analyzed.Fig. 43. Volume of the shield, vacuum vessel and magnet components as a function of wall load.


Table 16Peak temperature and time above 800°C for Apple, CLiFF,He-coded, and thick liquid designs

Peak temperatureConcept Time above 800°C(h)(°C)

APPLE 1.21275CLiFF 875a 84a

B1bHe-cooled 800b

0675Thick liquid

a See text; this is due to the shielding material, Tenelon,which can be easily replaced.

b With active cooling of the blanket region.

ume of activated material. The adoption of ‘lowactivation’ materials strategy, while important toreduce the radiotoxicity of the most active compo-nents, should be done as part of a broader strategythat also minimizes the volume of waste materialthat might be categorized as radioactive, even if lowlevel.

There is a need to explore new and innovativeconcepts that can substantially reduce the activa-tion of the large ex-vessel components that con-tribute significantly to the overall volume ofactivated material and to extend the capability ofconventional conceptual fusion designs with properoptimization to achieve the same goal.

A rough order of magnitude comparison ofFW/blanket waste volumes for thick liquid walldesigns with ARIES-RS is shown in Table 17.Although additional study is needed to solidifythese numbers, the initial indication is that thickliquid wall designs could significantly reduce thiswaste volume.

In addition to a reduction in the FW/blanketwaste volume, higher power density designs resultin a more compact machine which reduces thevolume of the shield, vacuum vessel and magnetcomponents. Fig. 43 plots the volume of thesecomponents as a function of wall load. Comparisonof volume of these components at the 10 MW/m2

value with the volume at the ARIES-RS value of5 MW/m2 shows a reduction in overall volume ofthese components by about 50%. This value as-sumes that all of these components are permanentlifetime components. For permanent componentsthe volume scales inversely with the wall load.However, in ARIES-RS, part of the shield is nota lifetime component. In this case, the overallvolume reduction afforded by high wall load mightbe reduced to about 30%.

14. Summary

The objective of the APEX study has been toidentify and explore novel, possibly revolutionary,concepts for fusion chamber technology that cansubstantially improve the attractiveness of fusionenergy systems. The first phase of the study wascarried out in 1998–1999 by a multi-disciplinary

than 2 h, and the relatively low radiological hazardof SiC makes this acceptable. The temperature inthe thick liquid wall design never exceeded 675°C.

Although the neutron and surface heat loads arehigher in APEX designs than those in conventionalfusion designs, these preliminary LOCA calcula-tions indicate that safety criteria (and more specifi-cally, no-evacuation guidelines) can likely be met.For the He-cooled refractory alloy design, this willlikely require the use of a safety-grade system toremove decay heat during accidents. It may benecessary to avoid the use of Tenelon in the shieldin designs such as CLiFF; in that case, activecooling may not be necessary. For others, such asthe thick liquid concept, a safety-grade system isprobably not necessary. It is desirable to make anysuch system passive to increase the reliability of thesystem.

These preliminary scoping calculations are by nomeans sufficient for determining whether thesedesigns will meet safety guidelines. They are meantas a starting point, and are used to make recom-mendations to designers so that safety is ‘built into’designs as they mature. As more design detailbecomes available, further safety analyses will beneeded to ensure that safety requirements are met.

13.2. Waste disposal issues

The environmental impact of waste material isdetermined not only by the level of activation, butalso the total volume of activated material. Atokamak power plant is large, and there is apotential to generate a correspondingly large vol-


Tab

le17

Rou

ghor

der

ofm

agni

tude

com

pari

son

ofF

W/b

lank

etw

aste

volu

mes

for

diff

eren

tco

ncep

ts

App

roxi

mat

est

ruct

ure

Red

ucti

onin

was

tevo

lum

eof

FW

and

FW

/bla

nket

Con

cept

type

Pea

kw

all

load

repl

acem

ent

tim

est

ruct

ure

frac

tion

blan

ket

com

pone

nts

for

liqui

dw

all

high

(MW

/m2)

pow

erde

nsit

yde

sign

sre

lati

veto

AR

IES-

RS

5.5

10%

2.5

FP

YA

RIE

S-R

Sin

finit

en/

aT

hick

liqui

dw

alls

wit

hno

stru

ctur

e0%

1070

104%

40F

PY

Thi

ckliq

uid

Flib

ew

alls

wit

h4%

stru

ctur

ebe

hind

wal

ls�

1.5

FP

Y10

Thi

ckliq

uid

wal

lsw

ith

1%st

ruct

ure

togu

ide

1%10

the

flow


integrated team of scientists and engineers from12 US organizations with participation of expertsfrom Germany, Russia, and Japan. A set of goalsfor the Chamber Technology were adopted tocalibrate new ideas and to measure progress.These goals include: (1) high power density capa-bility with neutron wall load \10 MW/m2 andsurface heat flux \2 MW/m2; (2) high powerconversion efficiency (\40%); (3) high availabil-ity (i.e. low failure rate and fast maintenance);and (4) simple technological and materialconstraints.

A number of promising ideas for new innova-tive concepts have already emerged from the firstphase of the APEX study. While these ideas needextensive research before they can be formulatedinto mature design concepts, some of them offergreat promise for fundamental improvements inthe vision for an attractive fusion energy system.These ideas fall into two categories. The firstcategory seeks to totally eliminate the solid ‘bare’first wall. The most promising ideas in this cate-gory are in the ‘concept rich’ class of flowingliquid wall variations. The second category ofideas focuses on extending the capabilities, partic-ularly the power density and temperature limits,of solid first walls. A promising example is the useof high temperature refractory alloys (e.g. tung-sten) in the first wall together with an innovativeheat transfer and heat transport scheme based onvaporization of lithium.

14.1. Liquid walls

The liquid wall idea evolved during the APEXstudy into a number of concepts that have somecommon features but also have widely differentissues and merits. These concepts can be classifiedaccording to: (a) the type of working liquid, (b)the thickness of the liquid flow, and (c) the type ofrestraining force used to control the liquid flow.

14.1.1. Basic principles and conceptsThe practical candidates for the working liquid

are the liquid metals lithium and Sn–Li (Sn–Liwas introduced into APEX because it has verylow vapor pressure), and the molten salt Flibe.Many different considerations must be taken into

account when assessing the performance of vari-ous liquid wall ideas. The hydrodynamics andheat transfer characteristics of high conductivity,low Prandtl Number liquid metal flows will de-pend heavily on the interactions with the magneticfield. In contrast, low-conductivity, high PrandtlNumber Flibe flows will be dominated by turbu-lence considerations. The Z number and ioniza-tion potential of any vapor generated from theliquid surface will affect significantly the plasmacontamination levels. The relative hydrogen solu-bility in the working liquid will play a significantrole in the structure of the edge and the stabilityof the plasma discharge.

In addition, high conductivity liquid metal(LM) flows have the potential to affect the localmagnetic fields and the plasma stability in a po-tentially positive manner. LM walls appear capa-ble of allowing stable tokamak operation withincreased elongation under reactor conditions.Modeling results indicate that the magnitude ofimprovement can be large with up to a factor ofthree improvement in stable b (from 5–7% to20–22%) at aspect ratio 4 and 3, respectively.Flowing liquid metals can potentially stabilizeresistive wall modes as well allowing higher b

steady state equilibria with very hollow currentprofiles. Steady state operation with such profilesenables high bootstrap fractions and thus lowrecirculating power. Also, hollow current profilesare theoretically predicted to give E×B shearingrates larger than instability growth rates for con-ventional drift instabilities, leading to transportbarriers and high confinement.

The selection of the thickness for the liquid walllayer flow (directly facing the plasma and in frontof a solid ‘backing wall’) leads to different con-cepts that have some common issues but manyunique advantages and challenges. Both thin andthick liquid walls can adequately remove highsurface heat flux. A primary difference betweenthin and thick liquid walls is the magnitude ofattenuation of neutrons in the liquid before theyreach the backing wall. The ‘thin’ liquid wallconcept is easier to attain, but ‘thick’ liquid wallconcepts greatly reduce radiation damage andactivation of the structure. Assuming a 200 DPAdamage limit for structure replacement, the use ofabout 40 cm of Flibe or Sn–Li can make the


structure behind it a lifetime component. Further-more, the volume of the radioactive waste fromthe FW/blanket system is greatly reduced.

Widely different liquid wall concepts are alsoobtained by applying various forces to drive theliquid flow and restrain it against a backing wall.An example is the gravity–momentum driven(GMD) concept, where the liquid is injected at thetop of the chamber at an angle tangential to thecurved backing wall. The fluid adheres to thebacking wall by means of centrifugal force and iscollected and drained at the bottom of the cham-ber. The criterion for the continuous attachmentof the liquid layer is simply that the centrifugalforce pushing the liquid layer towards the wall isgreater than any destabilizing gravitational force.

Using Flibe as the working fluid, the GMDconcept has been modeled with a three-dimen-sional, time-dependent Navier–Stokes solver thatuses the Reynolds Averaged Navier Stokes(RANS) equations for turbulence modeling andthe volume of fluid (VOF) free surface trackingalgorithm for free surface incompressible fluidflows. Example solutions at 8 m/s inlet velocitydemonstrate that a stable, thick fluid configura-tion can be established and maintained through-out a tokamak reactor configuration. Neverthe-less, gravitational acceleration and mass continu-ity lead to some amount of jet thinning as itproceeds from the top to the bottom of the reac-tor. Jet thinning can be overcome by increasingthe initial jet velocity, and a fairly uniform thickliquid film can be obtained throughout the plasmachamber if the jet is injected at 15 m/s. Thethinning can also be minimized by the MHD dragfrom the Hartmann velocity profile in a flow withconducting toroidal breaks. More analysis of thiseffect is needed for Flibe.

Numerical analyses were also performed forLM flows in GMD to determine whether or notan insulator is needed for free surface MHDflows, and to define lithium’s initial velocity thatenables a uniform thickness to be maintainedthroughout the plasma chamber in the presence ofthe toroidal magnetic field. The preliminary anal-ysis based on simplified magnetic field geometrieswith only toroidal or radial fields shows that theMHD drag effect significantly increases the layer

thickness and causes the associated reduction inthe velocity. Thus, there is a need of insulators fora free-surface LM flow if a toroidally segmentedpoloidal liquid metal flow configuration is consid-ered (other clever options may be possible that donot need an insulator). For an insulated openchannel, calculations indicate that a uniform 40cm-thick lithium layer can be maintained alongthe poloidal path at a velocity of 10 m/s.

Heat transfer calculations indicate that poloidalflow options like the GMD will have a surfacetemperature rise in the range of 150°C for lithium,and from 25 to 150°C for Flibe (depending onturbulence assumptions) when flowing at 10 m/s.A better understanding of free surface heat trans-fer (including the hydrodynamics near the freesurface/plasma interface) is needed to more con-cretely determine these values.

Variations of the GMD for the low aspect ratiospherical torus (ST) and cylindrical FRC includeadding an additional azimuthal (toroidal) velocityto produce rotation. The ‘swirl flow’ results in asubstantial increase in the centrifugal accelerationand better adherence to the backing wall, whenthe wall curvature in the poloidal direction islarge and the toroidal curvature is comparable tothe poloidal curvature.

The thin wall analog of the GMD is the con-vective liquid flow first-wall, or CLiFF, concept,where the goal is to eliminate the presence of asolid FW facing the plasma through which thesurface heat load must conduct. This goal is ac-complished by means of a fast moving (convec-tive), thin liquid layer flowing on the plasma sideof the FW. Such a thin layer is easier to controlthan a thick liquid system, but still provides arenewable liquid surface immune to radiationdamage and sputtering concerns, and largely elim-inates thermal stresses and their associated prob-lems in the first structural wall. The CLiFF classof liquid wall concepts is viewed as a more near-term application of liquid walls, and is suitablefor some currently operating plasma devices.

MHD analysis for LM-CLiFF has shown thatthe MHD drag can be significant if there is aradial magnetic field component — one normalto the free surface. Analyses indicate that a metal-lic backplate is acceptable with insulated toroidalbreaks if the radial magnetic field is no more than


0.1–0.15 T. The acceptable field magnitude woulddrop to 0.015 T for the case of toroidally continu-ous flow. Other important MHD issues such asflow across field gradients (1/R dependence of thetoroidal field for example), temporal fluctuationsduring start-up and plasma control will be ad-dressed in the next phase.

Penetrations will be required for plasma-sup-port functions such as heating and fueling. Novelschemes for accommodating penetrations in liquidwalls have been proposed. For example, modifica-tions to the back wall topology to guide the flowaround elongated penetrations are found to beeffective. Computational three-dimensional fluiddynamic simulation results for the CLiFF conceptwith Flibe show significantly reduced liquid layerdisturbance, no splash at the stagnation point,and no unwetted regions downstream the penetra-tion. These results are encouraging and providean excellent start for studying penetrations inthick liquid walls, where the volume of fluid ismuch larger.

The electromagnetically restrained (EMR), ap-plicable only to liquid metals, is another exampleof liquid wall concepts. EMR utilizes a J×Bforce field to push the liquid against the backingwall. An injected poloidal current interacts withthe main toroidal magnetic field to generate thisforce, resulting in liquid layer adherence to theback wall at potentially lower velocities than re-quired for the GMD.

Other active control schemes with injected cur-rents have been proposed as well, and will con-tinue to be investigated with new modeling toolsbeing developed for the task.

14.1.2. Moti6ation for liquid wall researchThere are many attractive features of liquid

walls that have motivated this research:� High power density capability:

Eliminate thermal stress and wall erosion aslimiting factors.Smaller and lower cost components (cham-bers, shield, vacuum vessel, magnets).

� Improvements in plasma stability andconfinement:

Enable high b, stable physics regimes if liq-uid metals are used.

� Increased potential for disruption survivability.� Reduced volume of radioactive waste.� Reduced radiation damage in structural

materials:Makes difficult structural material problemsmore tractable.

� Potential for higher availability.It is not clear yet that all these advantages can

be realized simultaneously in a single concept.However, the realization of only a subset of theseadvantages will result in remarkable progress to-ward attractive fusion energy systems.

14.1.3. Key issues for liquid wallsThe scientific and engineering issues for liquid

walls are many. Of all the potential issues, anumber of them stand out as the highest priorityfor near-term liquid wall research.1. Plasma–liquid interactions including both

plasma–liquid surface and liquid wall–bulkplasma interactions. Plasma stability andtransport may be seriously affected and poten-tially improved through various mechanismsincluding control field penetration, H/Hepumping, passive stabilization, etc. More care-ful estimates for the allowable amount of liq-uid evaporation and sputtering need to beobtained and benchmarked.

2. Hydrodynamics flow feasibility in complex ge-ometries including penetrations. The issue ofestablishing a viable hydrodynamic configura-tion threatens feasibility for all concepts, but itdiffers significantly for thick versus thin andfor molten salts versus liquid metals. The mainissue facing liquid metals is of course that ofMHD interactions. Without toroidal axi-sym-metry of the flow and field, reliable insulatorcoatings will be required on all surfaces incontact with the LM layer. Eddy currentforces perpendicular to the surface can pull theLM off the surface, even when complete axi-symmetry is assumed in the toroidal direction.Additionally, gradients in toroidal field canexert a significant drag on the free surfaceflow. For thick liquid walls, the main issuesconcern the formation and removal of theliquid flow in the plasma chamber, and theaccommodation of penetrations.


3. Heat transfer at free surface and temperaturecontrol. Liquid surface temperature and va-porization are tightly coupled plasma edge andfree surface hydrodynamic problems that re-quire knowledge of the radiation spectrum,surface deformation, velocity and turbulencecharacteristics. Being a poor thermally con-ducting medium, the Flibe surface temperaturehighly depends on the turbulent convection.However, the normal velocity at the free sur-face as well as the turbulent eddies near thesurface can be greatly suppressed. The issue ofheat transfer at free surfaces is a serious con-cern, especially considering that the prelimi-nary plasma-edge modeling predicts arelatively low limit on the surface temperaturefor Flibe.

14.2. High-temperature solid wall concepts

APEX has also explored ideas for extending thepower density and operating temperature capabil-ities of solid walls. Achieving high power densitymeans that the coolant heat removal capabilitymust be high and the first wall material must haveattractive thermophysical properties. Since mate-rials operating at very high temperatures generallyhave limited strength, such concepts should oper-ate at low primary stresses. This requires that thecoolant pressure be as low as possible, and thetemperatures throughout the first wall and blan-ket be as uniform as possible to reduce thermalstress.

Analysis of materials shows that the only struc-tural materials suitable for high-power density,high-temperature operation are refractory alloys.A tungsten alloy, e.g. W–5% Re, was selected asthe primary candidate material, with tantalumalloys as the back-up. The minimum and maxi-mum operating temperature for W and otherstructural materials were estimated. For W, thelower and upper operating temperature limits areabout 900°C and 1250°C, respectively, dependingon the choice of the coolant and the appliedstress. The lower temperature limit is based onradiation hardening/fracture toughness embrittle-ment due to low temperature irradiation. There isa large uncertainty in the lower temperature limit

for radiation embrittlement in W due to lack ofmechanical property data at irradiation tempera-tures above 700°C. The upper temperature limit isbased on thermal creep considerations and, de-pending on the coolant, could be further reduceddue to corrosion issues.

Two coolant schemes were evaluated. The firstuses helium with the motivation to explore thepossibility of using high temperature helium forhigh-efficiency energy conversion in a gas turbinecycle. The key difficulties with helium cooling arethe very high pressure (�12 MPa) and largetemperature rise, which push the requirements onthe refractory alloy structural material to therange of uncertainty in available data.

A more promising idea is an innovative coolingscheme based on the use of the heat of vaporiza-tion of lithium (about 10 times higher than water)as the primary means for heat removal. This idea,called EVOLVE (evaporation of lithium and va-por extraction) was explored in APEX in somedetail and will continue to be investigated.

Calculations indicate that an evaporative sys-tem with Li at �1200°C can remove a first wallsurface heat flux of \2 MW/m2 with an accom-panying neutron wall load of \10 MW/m2. Thesystem has the following characteristics:1. The high operating temperature leads to a

high power conversion efficiency.2. The choices for structural materials are limited

to high temperature refractory alloys.3. The vapor operating pressure is very low (sub-

atmospheric), resulting in a very low primarystress in the structure.

4. The temperature variation throughout the firstwall and blanket is low, resulting in low struc-tural distortion and thermal stresses.

5. The lithium flow rate is approximately a factorof ten slower than that required for self-cooledfirst wall and blanket. The low velocity meansthat an insulator coating is not required toavoid an excessive MHD pressure drop.

A preliminary conceptual design was developedand analyzed for EVOLVE. Key issues that needto be addressed in the future in order to assess thepotential of the concept include: (1) three-dimen-sional heat transfer and transport modeling andanalyses for the two-phase flow including MHD


effects; (2) feasibility of fabricating entire blanketsegments of W alloys; (3) effect of neutron irradi-ation on W alloys; and (4) analysis of safety issuesassociated with the high afterheat in tungsten incase of a LOCA.

14.3. Future work

The APEX team has already initiated its effortsfor the next phase, which will focus on moredetailed exploration of liquid walls and EVOLVE.The effort will include modeling, analysis, labora-tory experiments, as well as collaborating with thephysics community to conduct liquid wall relevantexperiments in existing plasma physics devices.

Acknowledgements

This work was supported by the Department ofEnergy, Office of Fusion Energy Sciences, throughvarious research grants, and involved the hardwork and creativity of many fusion researchers atthe participating organizations. Special thanks toSam Berk (US-DOE), whose deep understanding offusion technology issues was instrumental in initiat-ing the present study. Many thanks also to CharlesBaker, Kenzo Miya, Tomoaki Kunugi and GrantLogan for their continued guidance and valuableadvice in support of the APEX study. The APEXteam is grateful to many members of the fusioncommunity for their encouragement, support, andcontributions.

References

[1] M.A. Abdou, et. al., On the exploration of innovativeconcepts for fusion chamber technology: APEX interimreport. UCLA-ENG-99-206, University of California-LosAngeles, November, 1999.

[2] M.A. Abdou, The APEX Team, Exploring novel highpower density concepts for attractive fusion systems, Fus.Eng. Des. 45 (1999) 145–167.

[3] N.C. Christofilos, A design for a high power densityASTRON reactor. Lawrence Livermore National Labora-tory, Livermore, CA, UCID-15731 . Also issued in re-vised form as UCRL-72957 (1971) for submittal toNuclear Fusion but never published in that journal.Ralph Moir published this paper for Christofilos in:, J.Fus. Energy 8 (1989) (1970) 97.

[4] A.E. Robson, in: B. Bruneli, G.G. Leotta (Eds.), TheLINUS Concept, Unconventional Approach for Fusion,Plenum, New York, 1982.

[5] R.W. Moir, Mirror plasma apparatus. US Patent4.260.455, 7 April, 1981.

[6] R.W. Moir, Rotating liquid blanket for a toroidal fusionreactor, Fus. Eng. Des. 5 (1987) 269.

[7] R.W. Moir, Liquid first walls for magnetic fusion energyconfigurations, Nucl. Fus. 37 (1997) 557–566.

[8] A.P. Fraas, The BLASCON—An exploding pellet fusionreactor. Oak Ridge National Laboratory, Oak Ridge,TN, TM-3231, 1971.

[9] W.R. Meier, J. Maniscalco, Reactor concepts for laserfusion. UCRL-79694, Lawrence Livermore National Lab-oratory, Livermore, CA, 1977.

[10] M. Monsler, J. Maniscalco, J. Blink, J. Hovingh, W.Meier, P. Walker, Electric power from laser fusion: TheHYLIFE concept. Proceedings of IECEC Conference,San Diego, CA, August, 1978, p. 264. See also, J.A. Blinket. al., The high yield lithium injection fusion energy(HYLIFE-I) reactor. Lawrence Livermore National Lab-oratory, Livermore, CA, UCID-53559, 1985.

[11] V.A. Evtikhin et al., Liquid lithium tokamak reactor. 16th

IAEA Fusion Energy Conference, Montreal, Canada,October, 1996.

[12] R.W. Moir, HYLIFE-II inertial confinement fusion reac-tor design, Fus. Tech. 19 (1991) 617.

[13] L. Zakharov et al., Magnetic propulsion of conductingfluid and the theory of controlled tokomak fusion reac-tors. Meeting on liquid lithium controlled tokomak fusionreactors & MHD. International Sherwood Fusion/PlasmaConference, Atlanta, GA, 21 March, 1999.

[14] F. Najmabadi and the ARIES Team, Overview of theARIES-RS reversed-shear tokamak power plant study,Fusion Engineering and Design 38 (1997) 3–25.

[15] S. Fetter, E. Cheng, F. Mann, Long term radioactivewaste from fusion reactors, Fus. Eng. Des. 13 (1990) 239.

[16] Nuclear Regulatory Commission, 10CFR part 61, Licens-ing requirements for land disposal of radioactive waste.Federal Register FR 47, 57446, 1982.

[17] N. Morley, et al., Initial liquid metal magnetohydrody-namic thin film flow experiments in the MeGA-loopfacility at UCLA, Fus. Eng. Des. 27 (1995) 725–730.

[18] See for example, S.W. Haney, J.P. Freidberg, PhysicalFluids B1 (1989) 1637.

[19] R.L. Miller, J.W. Van Dam, Nucl. Fus. 28 (1987) 2101.[20] A. Turnbell, et al., Nucl. Fus. 38 (1998) 1467.[21] H.P. Summers, M.v. Hellermann, in: R. Janev, H. Dar-

win (Eds.), Atomic and Plasma–Material Interaction Pro-cesses in Controlled Thermonuclear Fusion, Elsevier,Amsterdam, 1993, pp. 87–118.

[22] T.D. Rognlien, J.L. Milovich, M.E. Rensink, G.D.Porter, J. Nucl. Mat. 196–198 (1992) 347.

[23] J.N. Brooks, Physical Fluids B2 (1990) 1858; T.Q. Hua,J.N. Brooks, Physical Plasmas 11 (1994) 3607.

[24] A. Hassanein, Fus. Tech. 30 (1996) 713.


[25] H. St. John, T.S. Taylor, Y.R. Lin-lui, A.D. Turnbull,Proceedings of the 15th IAEA Conference, 26 September–1October 1994, Seville, Spain, IAEA-CN-60/D-P22, p. 603.

[26] M.T. North, J.H. Rosenfeld, Test results from a heliumgas-cooled porous metal heat exchanger, in: Khonsary(Ed.), High Heat Flux Engineering III, vol. 2855, 1996.

[27] D.L. Youchison, J.M. McDonald, Thermal performanceand flow instabilities in a multi-channel, helium-cooled,porous metal divertor module. To be published in Proceed-ings of International Symposium on Fusion Technology,19–24 September, 1999, Rome, Italy.

[28] D.L. Youchison, M.G. Izenson, C.B. Baxi, J.H. Rosenfeld,High heat flux testing of helium-cooled heat exchangers forfusion applications, Fus. Tech. 29 (1996) 559.

[29] H. Kurishita, et al., Effect of neutron-irradiation on lowtemperature toughness of TiC-dispersed molybdenum al-loys, J. Nucl. Mat. 239 (1996) 253.

[30] B.L. Cox, F.W. Wiffen, J. Nucl. Mat. 85–86 (1973)901.

[31] T.E. Tietz, J.W. Wilson, Behavior and Properties of Refrac-tory Metals, Stanford University Press, Stanford, CA,1965.

[32] J.B. Conway, in: R.H. Cooper, Jr., E.E. Hoffman (Eds.),Proceedings on Refractory Alloy Technology for SpaceNuclear Power Applications, CONF-83008130, Oak RidgeNational Laboratory, Oak Ridge, TN, 1984, p. 252.

[33] R.W. Buckman, Jr., in: C.L. Briant et al. (Eds.), HighTemperature Silicides and Refractory Alloys, MRS Sym-posium Proceedings, vol. 322, Materials Research Society,Pittsburgh, PA, 1994, p. 329.

[34] S.J. Zinkle et al, in: Fusion Materials Semi-Annual Pro-gress Report (for period ending 31 December, 1997),DOE/ER-313/23, Oak Ridge National Laboratory, OakRidge, TN, 1997, p. 99.

[35] D.A. Petti, K.A. McCarthy, ITER safety: lessons learnedfor the future, Fusion Technology 34 (3, part 2) (1998).

[36] M.J. Gaeta, B.J. Merrill, CHEMCON User’s ManualVersion 3.1. INEL-95/0145, September 1995.

.

on the exploration of innovative concepts for fusion … publications/2001/abdou...the presentation...

Documents