Design descriptions of the Prometheus-L and -H inertial fusion energy drivers

ELSEVIER

Fusion Engineering and Design 25 (1994) l I I - 1 2 4

Fusion Engineering and Design

Design descriptions of the Prometheus-L and -H inertial fusion energy drivers G.J. Linford ~, D.E. Driemeyer b, S.W. Fornaca ", A.W. Maschke a " TR 14/ Inc., BM. 01/1220 Redondo Beach, CA 90278, USA h McDonnell Douglas Aerospace Co. (MDAC), St. Louis, MO 63166, USA

Abstract

Two innovative drivers have been designed for a prototype 1000 MW thermonuclear power plant planned for operation early in the next century. The Prometheus-L driver is a 4 MJ KrF master oscillator power amplifier laser system designed to operate at a 5.6 Hz repetition rate. Output pulses from the KrF master oscillator are synchronized with the pulsed-power excitation of the KrF power amplifiers and the launching of the inertial fusion energy (IFE) deuterium/tritium targets. The Prometheus-L laser architecture features 960 5 kJ electric discharge KrF power amplifiers pumping 60 crossed stimulated rotational Raman scattering H 2 amplifiers serving as beam accumulators. Pulse compression of the 60 accumulator beams is accomplished in 60 chirped, self-seeded S F 6 stimulated Brillouin scattering pulse compressors. Grazing incidence metal focusing mirrors minimize back-streaming radiation damage from the target chamber. This architecture permits the laser driver to deliver spectrally broad-band, temporally complex optical pulses in 60 beam lines to implode the direct-drive IFE targets within a 5 m radius target chamber. The Prometheus-H driver is a 7.8 MJ 4 GeV Pb ÷+ heavy ion (HI) inertial fusion energy system designed to operate at a 3.5 He repetition rate. The HI driver design is based on a short, ramped gradient, 5 MeV accelerator, followed by a longer, 2 km constant gradient, single beam linear accelerator operated in a 50 kHz burst mode to generate sequentially 18 4 GeV beamlets. A two-sided irradiation geometry was developed for indirect-drive HI targets. Six beamlets are used for the 45 ns precursor HI pulses stored in two superconducting storage rings, 12 superconducting storage rings accumulate the 12 main beamlets, with a final buncher generating the 8 ns HI pulses which arrive at the target chamber simultaneously. Final focusing is accomplished with large aperture triplet focusing magnets through Pb-vapor neutralization cells to reduce the effect of space charge, ballistically focusing onto two Pb stripping jets to create self-pinched Pb +82 beams which propagate along two preformed channels through two small openings in the HI target chamber blanket.

1. Introduction

T w o inertial fusion energy ( I F E ) driver designes have been developed for the M D A C - I e d Prometheus I F E reactor design study [1]: (1) a 4 MJ/pulse K r F laser driver, and (2) a 7.8 MJ/ Elsevier Science S.A. SSDI 0 9 2 0 - 3 7 9 6 ( 9 4 ) 0 0 0 6 0 - K

pulse heavy ion driver. The Prometheus driver designs represent the implementation o f recent advances in high energy laser and heavy ion accelerator technologies in order to enhance performance, reduce risk, and p r o m o t e cost-effectiveness. F o r the P r o m e t h e u s - L K r F laser driver, an eco-

112

G.J. L#!ford et al. / Fusion Engineering and Design 25 (I 994) I I I 124

nomical, reliable KrF laser amplifier design was selected which exploits novel non-linear optical (NLO) techniques developed during the past decade. These novel NLO techniques include: (1) use of stimulated rotational Raman scattering (SSRS) for large aperture, high energy beam synthesis, correction of optical beam distortions, and intensity smoothing to minimize optical damage by ultraviolet (UV), 2 ~ 250 nm laser light; (2) replacement of the previous elaborate angular multiplexed [2] pulse compression systems with a simple stimulated Brillouin scattering (SBS) pulse compressor self-seeded by an electronically controlled seed generator, producing significant cost savings. Implementation of these important NLO beam conditioning technologies had a significant effect on the design of the Prometheus-L driver, since these new technologies permit relatively small (approximately 5 kJ), cost effective KrF excimer laser amplifiers to be optimized for performance and economy without having to meet requirements for high output energies (around 200 kJ) or ambitious beam quality IFE goals. This permits the Prometheus-L driver to have acceptable performance at lower cost than conventional KrF laser designs. Significant improvements in driver availability is expected since individual 5 kJ Prometheus-L excimer laser amplifier modules can be replaced during routine driver maintenance operations without having to shut down the IFE reactor. In the case of the Prometheus-H driver a previous heavy ion driver design concept [3] has been given new life by incorporating some innovative techniques involving superconducting magnets, beam autoneutralization for control of space charge problems, self-pinched propagation of heavy ion (HI) beams, etc. The key elements of the Prometheus-H driver include the following innovations: (1) use of a single linear accelerator (LINAC) operated in the burst mode to generate a series of n HI pulses; (2) serial accumulation of these sequential n HI pulses in appropriate storage rings; and (3) use of triplet focusing magnets to collapse n/2 HI beams into two HI beam bundles inside pre-ionized, self-pinched channels for transport to {he target. Although the success of this single accelerator/storage ring approach is

dependent upon improvements in Metglas to reduce core losses, there are a number of attractive attributes possessed by the Prometheus-H driver design which reduce other heavy ion IFE driver costs. An example is the use of self-focused HI beam channel transport through the reactor target chamber to the indirect-drive deuterium/tritium (D/T) target. This configuration requires only two small penetrations in the blanket, thereby greatly simplifying the radiation shielding problems of the IFE reactor. The Prometheus IFE drivers have been designed to maximize cost effectiveness by employing redundancy where feasible in the important driver subsystems. These subsystems are generally optimized individually in such a manner that the overall Prometheus driver requirements are met by using a multiple number of coordinated subsystems. By optimizing the subsystems in order to maximize performance and cost-effectiveness, more reliable and economical driver designs have been developed.

2. Description of Prometheus-L driver

The Prometheus-L driver is based upon an architecture which emphasizes scaling of output energies by increasing the number of cost and performance-optimized components, rather than scaling up existing kilojoule scale KrF laser amplifiers to construct 100 k J-sized laser amplifiers. The Prometheus-L IFE laser driver/4] is composed of seven major subsystems: (1) KrF excimer laser master oscillator (ELMO); (2) 960 KrF laser 5 kJ drivers (employing electric excimer laser (EDEL) amplifiers and consisting of excimer laser preamplifiers (ELPs), excimer laser drivers (ELDs), and excimer laser power amplifiers (ELPAs)); (3) 60 SSRS accumulators (together with their Stokes seed generators); (4) 60 SBS pulse compression cells complete with "chirper" seed generators: (5) 60 electro-optical switchyard and optical delay lines; (6) excimer laser and NLO cell gas flow subsystems, and (7) 60 sets of beam coupling optics, image relaying, and final focusing optics. A block-diagram of the Prometheus-L driver is shown in Fig. 1.

G.J. Linford et al. / Fusion Engineering and Design 25 (1994) I I 1 - 1 2 4

l kxclmer Master Usclllator (1 J, 280 ns)

113

l

A

I

~ Beamsplitter Array ] 60_19 mJ, 280 ns)

17~ Beamspl~tter Array ]

1 I

(02 J, 270 ns)

1020 Gain 80 Exorner Preamos (4 J, 250 ns) I

950 Gain 38 ELPAs I. J 1020 Gain 50 Driver Amos I / 60 Raman Oscdlators (75 kj. 250 ns) m (200 J,~55 n£) I/./-~ (05 J:!.250 n s ) I ~

j

~

J60 !qaman Osc Pump Beam~mJ50 lqaman Preamplifiers (200 J, 255 ns) J, 250 as) 50 lqarnan Accumulators I .

(100 kJ, 250 n$)

I

I

60 Pulse Cornl~essors I i I I j(6s ~.J,6 ns. 28 kJ,a0 ns)lF I~ I

I

~

60 FOCbSmQMirrors I (65 kJ 6ns; EB kJ, 80 ns)I

Direct Drive Target ' ' 50 Grazing Incidence Mirrors t45 MJ, 6180 ns, ~=o2gl~jJ( 64 kdr 6 nsl 27 kdr 80 nslJ Fig. 1. Block diagram of Prometheus-L driver illustrating generation of 4.5 MJ energy at ). = 250 nm.

As illustrated, generation of the 60 PrometheusL laser driver output beams is initiated with a single, carefully timed and synchronized 1 J pulse at 2 = 248 nm from the excimer laser master oscillator, followed by division of this initial E L M O beam into 1020 beamlets by the beamsplitter array, followed by two tiers of excimer laser preamplifiers (totalling 1080 amplifiers) which amplify the beamlet energies up to an energy of 4 J. These 1024 K r F laser pulses are further amplified by 1020 KrF excimer laser drivers to beamlet energies of 200 J each. At this stage, 60 ELD KrF laser beams are used to generate 60 30 J Stokes seed beams at 2 - - 2 5 0 n m in 60 Raman seed MOPAs, and the remaining 960 K r F driver 200 J pulses are directed into 960 excimer laser power amplifiers which amplify the 960 beams to produce 7.5 kJ K r F laser output pulses. 16 of these 7.5 kJ ELPA KrF laser beams are then combined as pump beams for each of the cross-angle 60 Raman accumulators. Each of these 60 Raman accumulators is then used to amplify the 30 J 2 = 250 nm SRRS seed beam to an output energy of 100 kJ. In this way, the Prometheus-L laser driver can synthesize 6 MJ of energy equally dis-

tributed among 60 main beam lines of the laser driver with conversion efficiencies of approximately 80%. If one of the 16 ELPA KrF laser amplifiers begins to deteriorate, it can be replaced on-line as part of regular driver maintenance without having to shut down the I r E reactor. Since the Raman accumulators are operated at a crossed angle, a missing ELPA unit will not affect beam fill factors or beam quality, and the remaining 15 ELPAs can easily be driven 0.4% harder to correct for the missing 6% normally contributed by the missing ELPA. The 60 Prometheus-L laser driver beams of the driver are synthesized by the simple Raman accumulators, each delivering an energy E ~ 80 kJ. These 60 beams are then injected into each of the 60 SBS pulse compressors where the ~m > 250 ns pulses are shortened to ~m "" 6 ns. Following pulse compression and temporal tailoring in electrooptics (E/O) switchyards, these 60 short pulse beams are image-relayed and focused through the final focusing optics to generate the temporally complex, 4 - 5 MJ laser pulses focused on the direct-drive IFE targets injected within the target chamber. A similar K r F laser driver system could

114

G.J. Linford et al. / Fusion Engineerh~g and Design 25 (1994) I I I - 124 6C]-Fold Path-Matching

B/S Array Q /

17×6o

BIB Array of 1021S, Bearnhnes ~

~Exclmer /

H -~-

,

~

.

Master ~

Osollator

~~ ' t ~ e r +

; , - / / . + ~ . 2 ~-"--"-~'~ - ~ + - - - - ~

+e ,v ~

'

%-

....

j

,Turo,ng

M,rr :~+

+ \ " "~ N \ \ "% Exc,mer I I~Preamghf'ers

_

__I

I_

o° +, mr,o,

I,

D,,v+r

,hi,:

i II" l ~

':

H ~ XCi[T'ter IS. Preamphf lers

I M ~,

~

11ator

H

AccumuJators Outout Pulses

d

Fig. 2. Schematic of the Prometheus-L Raman beam synthesis design.

be used for irradiation of indirect-drive IFE targets following reconfiguration of the output beam optics into double, nested cones. The relationships of these various laser driver MOPA subsystems relative to each other are illustrated in Fig. 2 using only one of the 16 KrF laser beamlines to illustrate two of the 60 laser driver main beamlines. This simplified schematic of the Prometheus-L laser driver begins with the initiation of a ). = 248 nm laser pulse from the ELMO passing through the beamsplitter arrays which divide the original pulse into 1020 beamlets, followed by the ELPs, the ELDs and the ELPAs to the Raman accumulators. Following division of the original single E L M O beam into 1020 beamlets, the E L M O output pulse determines the phase, timing, intensity, l(t), etc., of the apodized output beams from the 1010 ELPs, 1020 ELDs and 960 ELPAs. These 60 additional excimer laser amplifier chains (consisting of 60 ELPs and 60 ELDs) provide the excimer laser pumping energy for the SRRS drivers. These SRRS drivers are path-matched with the original output beams from the ELDs to assure high SRRS gain at high bandwidth [5]. Each of the Stokes seed beams is spatially filtered 1.o improve beam quality, amplified in a Raman preamplifier having a gain of

300 to a level of 30J, collimated to a 120cm aperture, and injected into each of the 120 × 120 cm 2 aperture Raman accumulators. 16 ELPA pump beams enter each of the Raman accumulator cells (RACs) at an angle 0 ~ 5 ° to the RAC axis to permit intensity averaging of the excimer laser pump beams. The beam qualities of the output beams from the ELPAs are not expected to be high, owing to the turbulent flow of gases through the excimer amplifiers, dynamic gas disturbances caused by excimer laser excitation, etc. The crossed Raman (or CRAM) accumulator, however, is able to improve the output beam quality significantly relative to the excimer laser pumps, so that the output beams from the Raman accumulators can be of excellent quality. Following the main syntheses in the Raman accumulators, each of the 60 approximately 100 kJ beams is then injected into each of the 60 SBS pulse compressors, as shown in the diagram of Fig. 3. As indicated, the large diameter (120 cm), high energy (approximately 100kJ) square output beams from the Raman accumulators are directed through a time-varying electro-optical modulator "chirper" which places an acoustical frequency shift AVsBs on the leading 10 ns edge of the Raman

GJ, Linford et al./ Fusion Engineering:and Design 25 (1994)1I!-12,1

~-,,

Turning Mirrors ........: : : : ~ ~ Beamsp tter '' 'I' J

.

// ,-" ~ . ," ~;-/" " :I A " , ,''I~:IJiI, " "

....... < S J

• ",'N."%:"" baser Pulse A~

:I'I'::::I;~''''~'[" r;"

It,k,

"

Seed

Pulse

v"" .'Raman

Raman

~-----,-~.,

S ed [17i13 Accumulator " / , ~1~. Generator ...Chirper ,~/b~" Ex'c,mer Laser ' " V v '".,.,~ " ,'" Amplifier ~ L_., lqecoIlfrtl~'[]-ng ""ilL. o,o , or ......................

\

M4 Plate..................~ T ~ l |

---! P~

DrlverLase~EXclmer "'~.:'

Short Pulse SF6 Gas .............. l~t_.f II~ rm "c>-i-''i3utput Beam Retromirror .......

~

~

.......................

Pulse Compressor

Fig. 3. Schematic of the Prometheus-L c~ossed Raman: accumulator and SBS pulse compressor arrangement,

accumulator pulse. This "chirper" thereby generates the Stokes seed for the SBS pulse compression celt. The "chirper" is operated in a ramped mode to permit precise control over the resulting compressed pulse shape. It is estimated that approximately 65% of the laser energy in the 250 ns Raman accumulator pulses is converted into the short, 6 ns output pulses in the SBS pulse compressor. The remaining 35% of the laser beam energy remains in a long pulse duration (less than 250 ns), depleted SBS pump beam and is available for use in supplying the long, precursor pulse specified for the direct-drive D/T target in the E/O switchyard. The requirements [6] of the IFE precursor pulses specify that 35% of the total .laser driver energy be delivered in a longer pulse- format as part of the direct-drive target illumination requirements set forth by the Target Working. Group [6]. In order to maximize the optical efficiency while achieving the required __ 1% direct-drive target illumination uniformity, trapezoidal intensity apadization was used together with image relaying to project images of the intensity profiles onto 60 direct-drive target locations.. The relative fill factors of the-laser beam apodization can be varied to permit finite beam

misalignmentS.on the target and still meet the target illumination homogeneity requirements. The target is injected into the: c o n v e r ~ g beam region with a velocity 0 and movement of the target during the arrival. 0fthe sh0rt laser pulses can be compensated. Implosion Of the target Was neglected since it WaS assumed that the s~face of critical density remains relatively static d ~ g the arrival of the 60 short, 7 ns duration, main laser pulses. Following pulse compression in the SBS cells, each of the 60 square laser beams enters an electroOptical switchyard for integration of the long pre~ cursor pulses with the short m ~ pulses, and then directed in 60. equal intensity beam lines onto the D/T target. A s~plified, end-to-end d i ~ of the overall Prometheus-L la,~er driver system is shown in Fig. 4. Fig, 4 illustrates the operational optical ~ a n g e , ments of a :single arm of the laser driver subsystems, but fo.r reasons of clarity does not :illustrate the actual proposed Prometheus-L. c o n f i ~ a don o f the laser driver subsystems within: the IFE reactor building. A :schematic of a possible Prometheus-L laser driver eonfi~ati.on in the IFE reactor building is s h o ~ in Fig, 5:.

1i6

G.J. Linford et aLf Fusion Engineering and Design 25 (1994) 111-124

Preamplifier..

Dr Ive.r

Pow.er Amplifier

,- ......................

""~L._._.;~.,,..~m=~=_N --.

""

SBS Pulse Compressor

'-,, ~

V~LLo

.

.

.

- . P o . a , , , e r ----.

.

.

.

-F~xcimer

~__~ ~ump

=~.,:"~. " / ~ ' ~ . .

~¢~i

.

Be@ms :

L[

Short "-. ',, ~.~.f~.~ ~ee~ i) I}i ...... Output '~. '..~.~~v. Raman. '"'"I Pulse ~ J ~ : ~ ~ "Accumu~tor ' i~ ~ o,se-~~ -..~

"...~

Retro Mirror

~ Rom-a:, n . .

"Chrper" -.

~

/

~

ncj" ......

_

.7~;,.'.#:..-Beamsplitter' ,%ld Plate~ ".. ' -'~, Array •

N~,~dron . " " I-!eld . ".

"

.

"" Ekclmer Master Oscdlato r

!

ZI .,/ ! Ill

", '\

Osc

\ %, " i:leam~ itter-'" .-"1~ I . : . Power .,. , IAmpllrler Recolhmat rag,.-' Telescopes..)~\,.

[] IOriver

',A.

:nee

'\! Preamplifier ................ ~

Fig. 4. End-to-end optical diagram of Prometheus-L IFE driver. Exc~mer Laser Modules

~ S~/_~

Target Chamber /'

,.~,r.a._~'~.,"~.

~ . c c u m u l a t.O~S

sas pulse \\ \ ~ ~ / / # 1 1 n ~ \ \ ' ~ % . ' % 7 / / - . , , . _ ~ 7 CelB - ~ - -

I I1

~ B u l l d i n g Fig. 5. Top view of Prometheus-L driver building layout. 3. Description of Prometheus-H driver

The Prometheus-H heavy ion [1] driver is designed to irradiate an indirect-drive IFE target from either one or two sides with both precursor and main pulse heavy ion beams having a total energy of 7.8 MJ. As shown in Fig. 6, the HI driver consists of several subsystems: (1) a singlebeam induced LINAC running in a high repetition rate burst mode producing a train of identical current pulses. (2) A set of storage rings accumu-

Iating the current pulses and acting as a series of delay lines, permitting beams that are generated over hundreds of microseconds to reach the target within hundreds of picoseconds of each other. Some of the beamlets are destined to become part of the two precursor beams, which contain roughly 30% of the beam energy and have a duration 4-5 times that of the main current pulses. Segregation of the pulses into main and precursor beamlets takes place in the storage rings. The combination of the single beam LINAC and the storage rings in the PrometheusH takes the place of the multiple beam accelerator used as the baseline in earlier studies [7-14]. (3) After the LINAC burst is complete, the storage rings release the beamlets into combination bunching/drift sections to compress the beams longitudinally. Although they share a common tunnel, the precursor and main beams have differ. ent requirements and are handled separately. Following the buncher/drift section, the beamlets are divided into two identical groups and directed toward opposite sides of the target chamber. (4) A final focusing section on each side compresses the beamlets radially, coalesces the main beamlets into a main beam, and merges this beam with the end of the precursor beam just inside the target chamber at self-pinching beam currents. The

G.J+ Linford el aL / Fusion Engineering and Design 25 (1994) 111-124

Single

,Beam

LfNAC

,,

,

i ,

• +. . . . .

._f

"

O(~'-~.'2-~

~

:

~

~.>~r .

~..~,~£L~

/

_....

~ BUik~ ~"u

Target

PrepclseBuncher Accelerator 1£6 r e : d / , ' _ ...................... ;~. . . . . . . . . . . . . . . . . . . . . .

, 14

-

,

"-~

\'~..~'/r ..~-

~ "+++ ~"

-~ ~

,

Source

;

1t7

~Stor

RlrlgS a g e

!

~,o

p.,~

~?~',~

rli -

~"~,\

Bunche(-

',/

il

Accelerat+

Final Transport ~ 180 rn ~___.~_ ~

I tIH.

Transport " ~ Tr lp let Focusing Magnets

Fig, 6. Schematic of Prometheus-H driver,

merged beams then propagate across the target chamber to opposite sides of the indirect-drive target. The entire Prometheus-H heavy ion driver system is cycled at 3.5 Hz. The Prometheus-H driver is based on spacecharge limited acceleration and transport of a heavy ion beam in a simple alternating magnetic quadrupole focusing channel. The channel containing the intense beam is formed by a slowly varying FODO lattice (focus-drift-defocus-drift). As was the case with the Prometheus-L KrF laser driver described above, the Prometheus-H heavy ion driver is composed of subsystems and individual modules which are subject to parallel optimization during the years I995-2030. In •order to meet the overall IFE reactor output power requirements, the overall Prometheus-H HI driver is cycled at 3.5 Hz. As noted previously, the channel containing the intense HI beam is formed by a slowly varying FODO lattice, whose parameters are illustrated in Fig. 7. Defined in Fig+ 7 are the quadrupole spacing L, the quadrupole length 1 = ~/L, the space between the quadrupoles L ( 1 - */), the beam envelope of radius a, the heavy ion beam trajectory, the alternating focus

F(~Array

j

Beam ]~.velol~ Ion Trajectory of Wavelength" k~.

J/A.,...................................................... ~ i...................................................................................... L:..-.

....

a

=

2~/~

Fig. 7. Schematic of the Prometheus-H quadrupole focusing channel.

and defocus magnets, the betatron wavelength 2b, and the phase advance a = 2~L/2b. As indicated in Fig. 7, the HI beam envelope has a periodicity equal to that of the quadrupoles, but the trajectory of an individual ion within the HI channel is characterized by a much longer betatron wavelength. The betatron wavelength 2b scales with the radial space charge forces on the beam and the restoring forces of the quadrupole channel and

G.J. Lit!lbrd et al. / Fusion Engineering and Design 25 (1994) 111- 124

118

ha qmd~apole int~or

,~

Field c ~ , l i n ~ 1~ s

x

t~

\-j

Poletip

1

"Pralk~et'~ Coordinate

net currents, minimum depressed tune, etc.) were used to constrain the parameters to restabilize HI driver configurations. Given the basic beam parameters at the target (total beam energy, ion kinetic energy, number of beamlets, pulse length, etc.), the cost of electricity from the HI-driven IFE plant minimizes at specific values of the three parameters. Different values are obtained if the LINAC and buncher costs are considered separately. The scaling laws, practical constraints, and river variables in terms of the three scaling parameters are given below. The space-charge limited curent lo,~x is defined as

Fig. 8. Magnetic fields at pole tips, walls and wires. ),b/4 is the maximum HI beam displacement from the axis after receiving an angular kick. An infinite betatron period (or zero phase advance) corresponds to having no net restoring force on the ion. For a coasting HI beam, the beam area is constant, although the aspect ratio changes. As the HI beam accelerates, the area decreases. The normalized emittance is defined by EN =[3;'E, where E is the emittance. The strength of the FODO channel focusing depends not only on the physical spacing of the magnets but the quadrupole gradient as well. The relative magnetic field strengths at various locations are illustrated in Fig. 8. The poletip field is defined as the magnetic field at the beam edge. The lattice parameters vary along the HI driver according to the local beam energy and/or current; the acceleration and bunching schedules (which determine V[-] and I[-]) are free parameters. The ramp portion of the LINAC commences the prebunching of the HI beams. Beyond the ramp gradient region of the LINAC, bunching continues. The HI pulse duration is given by r = l b / t l , where lb is the HI bunch length and v is the heavy ion bunch velocity. Simple expressions for the space charge-limited current /max, the emittance EN, and the quadrupole focusing strength Bpoletip, were used to derive three scaling parameters, characterizing the driver and lattice variables throughout the bunch acceleration and compression processes in .terms of the local beam energy and peak current. Practical limitations on the physical parameters (e.g., component access, mag-

I ..... = 1 . 5 6 × 10

-6 a°4L~-x[ 2

A ~,

(1)

where ao is the undepressed tune (phase advance in the absence of space charge), a is the beam radius, L is the quadrupole spacing, A is the atomic weight of the ions, L is the distance between the quadrupoles, V is the beam energy (MeV), and Q is the heavy ion charge state. To define the superconducting quadrupole limitation, the following expression for the poletip magnetic field Bpoleti p w a s used:

aAI3 Bp o l et i p =

3.13 IIQLZx/I _ f12

(2)

where the constant 3.13 was derived from recent magnet improvements [7,8], and where q is the quad packing fraction. A convenient expression for calculating the constant normalized emittance E N is

EN

-

aaSfl7 2L

(3)

Additional practical constraints [13,14] include: for adequate access to the HI beamlines, r / < 0.8 everywhere. Control of the final focus magnetic lens aberration requires that the ratio of the beam radius a to the quad length lq satisfies a/lq < 0.25 throughout the HI driver F O D O system. For the superconducting magnets, it is necesary that the current density be kept below its local limit (SSC-type conductors are assumed). The entire Prometheus-H driver (LINAC, storage ring, and buncher sections) is designed around an aggres-

G.J. Lhlford et al. / Fusion Engineering and Design 25 (1994) I 11-124

119

Table 1 Scaling of heavy ion driver parameters [13,14] #

Parameter

Symbol

Variable voltage

Constant voltage

I 2 3 4 5 6

Depressed tune Current Beam radius Beam duration Beam length Q u a d r u p o l e spacing Packing f r a c t i o n Q u a d r u p o l e length

a I

VI~ V t - lJ

I -~' I

a

V M 2 - 1 t - J14

r 1 L

V "/~

17 ~12 I-

V' - ~/2 V' ~

I- t I;- i

q

V - 3/2a ÷/t + 1/4

1312 - ~,

Iq

V - 1 / 4 - ~/2

1112

7 8

sive transport schedule [8]. The current is maintained at the local space charge limit of a magnetic quadrupole focusing channel throughout the driver, and the depressed tune is allowed to drop as the beam approaches the target. The heavy ion driver scaling parameters are summarized in Table 1. 3. I. Single beam LINA C The high repetition rate burst mode single beam Prometheus-H LINAC comprises two portions, the injector and the main LINAC. The injector includes the ion source and required preaccelertion and beam manipulation to produce a suitable beam for the main LINAC. The two sections of the main LINAC, the ramped gradient and the fixed gradient sections, differ in the allowed voltage gradient as a function of position, but are otherwise identical in function. Both sections accelerate a single space-charge-limited beam at the maximum allowed rate. The main components of the single beam L I N A C are the induction modules that accelerate the beam and the superconducting quadrupole magnets that form the transport channel. The beam is accelerated at the maximum rate in the LINAC. In the ramped gradient section, the local gradient is limited to a value proportional to the local ion energy. This is necessary so that the velocity gradient gained at different portions of the beam as it traverses a given gap is sufficient to bunch the beam without grossly affecting the energy gained by the pulse as a whole. The end of the ramped gradient section

is defined to be at the point where the maximum gradient is reached, at the voltage Vramp. The fixed gradient section causes the energy gain to be linear with distance to the end of the LINAC. The variation in velocity gained as the non-relativistic HI beam traverses a gap is sufficiently low that the beam can be compressed without limiting the gradient. The maximum gradient is taken to 1 MV/m, a conservative value for a pulsed system. The limitation is insulator flash-over. Because less time is available to reset the cores, the loss (J/cm 3) per pulse cycle in the single beam LINAC (SBL) case is necessarily larger than the multiple beam LINAC (MBL) case. Optimization of the current waveform with respect to the losses should result in lower heat generation in the core and lower installed power costs. Protection of the induction cell components against overvoltaging is needed in case of a misfire. Recent induction LINACs, including the 10 kA Advanced Test Accelertor (ATA) at LLNL, have a damping resistor in parallel with the beam to limit the voltage, both across the cell and across critical components, in case the beam arrival time is in error. Although this provides component protection in a low repetition rate device, it will not be acceptable in a power plant application. The pulsed power requirements vary along the L I N A C as the pulse length, gap voltage, and current vary. The requirements for the acceleration modules are straightforward, but advancements must be made to achieve the high repetition rate burst mode operation. While thyratron switching may be used

120

G.J. Lhaford et al. / Fusion Enghleerhlg and Design 25 (1994) III

for low PRF applications such as a multiple beam LINAC, silicon-controlled rectifiers have been proposed for induction LINACs with PRFs up to 6 kHz for improved long-term reliability. The solid state devices have a lifetime greater than 10 j~ firings if operated within specifications. Efforts have been directed toward arranging the components so that faults in the load do not cause the pulse power components to exceed their ratings. For the factor of three higher PRF required for the SBL driver burst mode, even SCRs are probably insufficient, and FETs will probably be necessary. The high P R F pulsed power for the SBL will be more expensive on a per joule basis than the very low PRF used in the MBL: a value of $100/J was assumed in costing the pulsed power for the Prometheus-H driver.

3.2. SuperconducthTg magnets The focusing quadrupoles used to contain the beam radially during acceleration are superconducting with cos(20) windings to provide the quadrupolar magnetic field, and are covered with iron to provide flux returns. These quadrupoles fit beneath the accelerator cores at the low energy end of the LINAC where the packing fraction is high. The cost of this type of quadrupole is relatively insensitive to size. However, because the induction cores fit over the magnets, it was necessary to estimate the quadrupole radial extent in order to calculate the core volume. Because the Prometheus-H LINAC system runs at relatively low quadrupole gradients, the entire structure is compact. A specific relation between the scaling law parameters was chosen to provide a uniform beam diameter along the LINAC. Thus, in order to meet the simultaneous constraints on axial packing fraction (q <0.8) and aspect ratio (a/ lq < 0.25), the ideal length of the quadrupole must vary along the LINAC. From a spares inventory viewpoint, having unique magnets is undesirable. However, the field strength used in the quadrupoles is nowhere near the limits of the superconductors (the 2 Tesla "poletip" field in the Prometheus-H driver' LINAC is less than 60% of the maximum allowed by the conservative poletip constraint). Since each magnet acts to provide a

124

transverse momentum impulse, the quadrupole length can be fixed at a standard value (or a few values) and the field varied, keeping the product //lq constant. The high allowed aspect ratio of beam radius to quadrupole length in the Prometheus-H design (1/4 vs. 1/10 in other studies) is an area of potential concern. The issue is growth in the normalized emittance EN that is caused by magnetic lens aberrations due to the relatively short length. While there are octupole components in the lens as a result of end effects, these appear to play a significant role only if the beam is not space-charge dominated. As part of the Neutral Particle Beam Program, experiments have been performed at L A N L on the acceleration of high current, high brightness proton beams in a large aperture drift tube LINAC. The short quadrupoles are contained within the drift tubes to keep the high current beam focused. The beam there is observed to propagate without increasing its emittance significantly. While it is certain that there will have to be considerable magnet design to optimize the quadrupoles, the first-cut approximations were sufficient to proceed with the Prometheus-H design. The magnets in the heavy ion driver are required to be cooled to liquid helium temperature to be superconducting. The magnets are in cryostats that isolate them from room temperature. During the field ramping stage during startup, eddy currents generate a thermal load, but once these magnets are operating, the predominant thermal loading mechanism on the system is static heat leaks through the cryostat penetrations. Liquid helium is brought into the magnet, and cold gas exits. Higher temperature gaseous helium from the boil-off is used to anchor thermally the shields at intermediate temperatures (typically 20 and 80 K), eliminating the need for a separate liquid N~ system. Large distributed cryogenic systems (refrigerators, cold boxes, distribution lines, etc.) of this scope have been built and operated reliably at various accelerator facilities around the world (Fermilab, CERN, DESY, CEBAF), and a much larger one may soon be built for the SSC. The cryogenics system is not an issue for Prometheus-H. A properly designed cryostat will have axial thermal losses due to the

G.J. Linford et al. / Fusion Engineering and Design 25 (1994) 11i-I24 Piing

supports. F o r this reason, the cryostat should be constructed as long as possible, consistent with access requirements, minimizing the number o f penetrations between the 300 and 4 K elements. Since the cryostat cost is independent of size, the later portions of the Prometheus-H driver where the beams are carried in parallel beamlines have single cryostats containing many magnets.

J

3.3. Storage rings As each of the 18 beamlets exists the single beam LINAC, it is captured and stored in one of 14 identical rings. 12 beamlets have individual rings, and three beamlets are injected into each of the two remaining rings. The two groups of the three beamlets form the two precursor beams. The injection timing is such that the three beams are stacked one after another, then all 18 beams are sent to the bunching LINACs simultaneously. The simplest possible sublattice was assumed for sizing the storage rings: focus-drift-bend-drift-defocus-drift-bend.drift. This is replicated twice, and a straight section was included to facilitate injection and extraction of the beams from the rings. This :geometry is repeated four times. 34 1.0 m length, cos 20 superconducting quadrupoles (identical to those at the end of the LINAC) have a spacing of 4.6 m around the ring. 30 2.25 m long 6 Tesla cos 0 superconducting dipole bending magnets have a 50% packing fraction (the magnetic rigidity of the 4 GeV Pb 2+ ions is 65 Teslameters). The storage rings are stacked vertically and share common cryostats, although the beamlines are separated. The maximum storage time in one of these rings is 18 times the interbuncli spacing time. At a burst rate of 20 kHz, for example, the storage duration of the first bunch is 50 #s × 18 = 0.9 ms. During this time, the coasting bunch makes approximately 400 revolutions. There is a hazard that the H I beam may couple to resonances in the circular system, leading to beam degradation, emittance growth, and scrape-off. The consequences of scrape-off are more serious in a circular machine since the beam repeatedly passes the same location. A slight loss of beam to the wall could evolve enough gas to further perturb the beam on a subsequent pass, eventually

121

/

~chronized

berial Loading of Incoming ion Bunches

~Rlng Cycles at 3.5 Hz

Output Bunches

Fig, 9. Schematic of storage ring geometry.

causing closure o f the aperture through this vacuum instability. The A L A D D I N storage ring at the University of Wisconsin was plagued for years by this effect. In a linear device, the gas has a chance to dissipate before the next beam pulse arrives. In addition, the beam may debunch as it coasts around the ring. In the central part of the beam, the axial electric field is nearly zero, but toward the bunch boundaries, the axial forces try to accelerate the head and decelerate the tail. To remain at the space charge limited current, the beam will occasionally have to be "kicked" at its ends to counteract the cumulative effects of the axial space charge force. The beam head must be retarded and the tail accelerated. This can be done with only a few cores per ring, because the kicker voltage is low, and the affected portion of the beam is small. There is little stress on the induction modules and the resulting core volume is negligible. This adjustment does not have to be made frequently, possibly once per revolution. In the main L I N A C and the buncher, this voltage can be provided by the induction modules already present. Separate correction modules are required only in the storage rings because the beams coast at constant energy and constant length, and there is no need for the adjustments to be made in the main modules. The short length of the correction pulse makes it desirable to use separate kicker modules in the remainder of the HI driver. It has been predicted that a high-current heavy ion beam will couple to

122

G.J. Linford et al. / Fusion Eng#leer#~g and Design 25 (1994) III

the impedance of gap structures in the LINAC, forming a longitudinal analogue of the head-totail cumulative beam-breakup observed in electron LINACs. Small-scale variations in beam energy along the pulse (with wavelength long enough that the space charge forces do not reduce them quickly) would be transformed to current variations by the drift, and it is possible that an instability could grow to some finite level before it saturates. The outcome would be to increase the longitudinal emittance of the pulse, and make it more difficult to compress. The problem would presumably be worse in a storage ring, where the total path length is much longer, and the possibility of a regenerative instability also exists. Mechanisms for instability control include: (1) lowering the impedance of the gap, and (2) using feedback in the cells to damp the modes. A method must be provided to assure that the beams in all rings are synchronized so that they are at the correct position when they are ejected from the ring. Small differences in beam energy or ring circumference could otherwise accumulate, leading to a condition in which the beams are separated in time by as much as 2.3/~s due to one beam passing the extraction point

124

before the others. Rf storage rings use the principle of phase stability to maintain the particle in the right position; a slight periodic ramp over the pulse would accomplish the same thing in the long-pulse induction systems, accelerating late pulses and retarding early ones, requiring careful control of the beam parameters. Finally, there is the problem of injection and ejection [13,14]. In the high current ring, the quadrupoles are necessarily close together, leaving very little room to couple into and out of the ring. Bringing the beams out between adjacent quadrupoles does not appear to be practical. A pulsed dipole magnet producing 10 T would have to be a 2.7 m long in order to allow the beam to clear the quadrupoles. The magnetic field energy alone is of the order of 40 MJ per magnet per pulse. Even if it is assumed that the energy is not dissipated, the field would have to be established within r = 2/~s.

3.4. Final bunching and focus When the storage rings eject the 14 pulses, the beamlets are sent to one of two induction LINAC based bunchers. The first buncher compresses the twelve main beams, the second compresses the

Ouadr upole Focustng Magnets Beam Bun£hes //

4 Oegr~-_ Cony Pt

c__,r'~[,o~ng Jet qelf-Focused Str rpped *92 Heavy Ion Beam A W Diameter

Neutrahzmg tJas Focusing Bunched Icon Beams

=

& Double-Slaed Indirect Dr~ve Target

Par Dally-Bunched Precursor Beams

Target Chamber

Fig. 10. HI driver final focus through quadrupole magnets, Pb neutralizingcells, ballistic loci onto strippingjets, stripping to Pb *82 self-pinched transport to target.

G.J. Linford et al. / Fusion Engineering and Design 25 (1994) I I I

two precursor beams. The two bunchers are in close proximity, sharing the same tunnel. The active portions of the two bunchers have different lengths. The main beams coast 115 m before entering the buncher and exit at the same location where the precursor beams leave their respective buncher. Both bunchers are configured with independent beamlines, with superconducting quadrupole focusing magnets sharing common cryostats, and the induction module cores surrounding the beamlines between cryostats. The quadrupole lattice parameters at the entrances to each buncher are the same as at the SBL exit. Each buncher L I N A C adds a position-correlated energy across its beam pulse. The buncher cells are ramped in time, adding zero to the front of the pulse and producing the maximum 1 MV/m gain at the end of the pulse. The main (precursor) buncher extends 8 5 m (180m), so there is a 4.2% (9%) energy difference from the front to the back of the

124

123

pulse. In practice the energy of the L I N A C would be dropped 2.1% to compensate for the net energy gain in the buncher. The precursor beams would still be 2.1% high in energy and the precursor buncher would be readjusted accordingly. As the bunches drift, the backs of the HI pulses begin to catch up with the fronts, increasing the current. The bunch compression process commences within the buncher and continues to the target. Fig. l0 provides a schematic of the ballistic loci through the Pb neutralization cell onto the Pb stripping jets, prior to injection of the self-pinched beams into the target chamber. Fig. ll illustrates the final HI focusing geometry. The final transport of the Pb +82 beams to the target consists of several steps. The majority of the beamlets are bunched, transported to the reactor area in two groups, tightly focused, and finally ballistically combined in a small spot just outside the target chamber. The remaining beamlets are temporally

INJECTION SYSTEM OUADRUPOLE IN FINAL TRANSPORT

HEAVY 10N BEAM (7 GAS NEUTRALIZATION

FIRST WALL PAN BLANKET REACTOR VACUUM VESSEL -

TRIPLET FINAL' ' ~ FOCUS MAGNETS FINAL FOCUS-'--''" VACUUM ENCLOSURE VACUUM PUMP

PORT SHIELDPLUG--" J BULK SHIELDWALL

Fig. 1 I. Isomeric view of Prometheus-H final focusing geometry.

J ' ~ X"

124

G.J. Lh~Jbrd et al./ Fusion Enghwering and Design 25 (1994) III

combined in the storage rings and form the two precursor beams with a longer duration and a lower total energy content. The beam is transported to the target in a self-pinched mode through a preformed channel.

4. Summary Two new IFE driver designs have been completed which promise to permit significant increases in reliability and cost-effectiveness for fusion power plant applications. The PrometheusL excimer laser driver utilizes a non-linear optics to achieve performance enhancement and cost reductions. The Prometheus-H heavy ion driver minimizes costs [1,8-12] by using a single beam LINAC [12-16] and superconducting storage rings [ 17].

References [1] L. Waganer et al., Inertial Fusion Energy Reactor Design Studies, Vol. I, II, and Ill, DoE/ER-54101 MDC 92E0008, March 1992, McDonnell Douglas Aerospace Team, St. Louis, MO 63166-0516. [2] G.J. Linford, Multi-MJ KrF laser driver for a 2050 IFE reactor, Proc. 14th IEEE/NPSS Symp. on Fusion Engineering, Vol. I1, 1991, pp. 891-894. [3] A.W. Maschke, personal communication, 1991. [4] See, for example, L.A. Rosocha, J.A. Hanlon, J. McLeod, M. Kang, B.D. Kortegaard, M.D. Burrow and P.S. Blowling, Aurora multikilojoule KrF laser system prototype of inertial confinement fusion, Fusion Technol. 11 (1987) 497-532.

124

[5] S.J. Pfeifer, K.J. Allardyce, R. Johnson, G.J. Linford, H. lnjeyan, J.A. Betts, H. Komine, G. Lombardi, W. Long and E. Stappaerts, Raman Beam Combining Program, AFWL TR 86-142, Final Report to Air Force Weapons Labortory, Kirtland Air Force Base, NM 87117, September 1986. [6] R.C. Davidson, ICF Design Studies Recommended Guidelines, Department of Energy Office of Fusion Energy', Germantown, MD, 1990. [7] J. Hovingh, V.P. Brady et al., Heavy ion LINACs as drivers for inertial fusion power plants, Fusion Technol. 13 (1988) 255 278. [8] C.G. Fong and L.R. Reginato, Invetigation of induction cells and modulator design for heavy ion accelerators, Proc. 14th IEEE/NPSS Symp. Fusion Engineering, 30 September- 3 October 199 I, San Diego, pp. 1171 - 1171. [9] L.D. Stewart, M.J. Monsler and S. Humphries, Jr., Parametric studies of an induction LINAC heavy ion beam driver for ICF. Proc. 14th IEEE/NPSS Symp. on Fusion Enegineering, 30 September 3 October 1991, San Diego, pp. 1166-1169. [10] R. Bieri and W. Meier, Heavy-ion driver parametric studies and choice of a base 5 MJ driver design, Fusion Technol. 21 (1992)1589 1593. Ill] R. Bieri and L. Stewart, Heavy-ion driver design and scaling, Fusion Technol. 21 (1992) 1583 1588. [12] Y. Seki, Fusion power reactor studies in Japan, Fusion Technol. 21 (1992) 1707-1714. [13] E.P. Lee, S. Yu, H.L. Buchanan, F.W. Cahbmers and M. N. Rosenbluth, Filamentation of a heavy ion beam in a reactor vessel, Phys. Fluids 23 (1980) 2095-2100. [14] E.P Lee, T.J. Fressenden and L.J. Laslett, Proceedings of the 1985 particle accelertor conference, IEEE Trans. Nucl. Sci., NS-32, pp. 2489-2492, 1985. [15] E.P. Lee, E. Close and L. Smith, Proc. 1987 IEEE Particle Accelerator Conf., Washington, DC, pp. 1126-1130. [16] E.P. Lee, C. Fong, S. Mukherjee, W. Thur, Conceptual design of blend, compression and final focus components of ILSE, Proc. 1987 IEEE Particle Accelerator Conf., Chicago, Ill, pp. 971-975 (Target working group). [17] A. Maschke, Superconducting Supercollider Magnet Performance, personal communication, 1992.