R.E. Laxdal and P. Bricault
TRIUMF, 4004 Wesbrook Mall, Vancouver, B.C., Canada V6T 2A3
The ISAC radioactive ion beam facility under construction at TRIUMF combines
an isotope-separator-on-line with a post accelerator. Required in the
accelerator chain is a drift tube linac capable of accelerating unstable
nuclei with a post stripper charge to mass ratio of 
from
E=0.15MeV/u to a final energy fully variable up to
1.5MeV/u. Due to the relatively low intensities of some of the
ion species continuous (cw) operation of the accelerator is required.
A five tank interdigital H-type structure, operating at 105MHz, has been
chosen. The beam dynamics conceptual design and the results of various particle
simulations are presented.
Computer modelling of the rf structure and model measurements
have been completed and are reported.
A radioactive ion beam facility is being built at TRIUMF. [1]
In brief, the facility includes a proton beam (I 
100 
A) from the
TRIUMF cyclotron
impinging on a thick target, an on-line source to ionize the
radioactive products, a mass-separator for mass selection, an accelerator
complex and experimental areas.
The accelerator chain comprises an RFQ [2] to
accelerate beams of 
from 2keV/u to 150keV/u and a post
stripper, variable energy drift tube linac (DTL)
to accelerate ions of 
to a final energy between 0.15MeV/u to 1.5MeV/u. Both linacs are required
to operate cw to preserve beam intensity.
Both a Wideroe structure and a super-conducting structure have been considered
[4] for the post stripper linac in the ISAC project. The former
idea was abandoned due to the high power consumption and the latter
was abandoned because of the required technological development. Instead,
the IH structure has been chosen for its high shunt impedance. The structure
has been configured as a separated function DTL.
Five independently phased IH tanks operating at
provide the main acceleration. Longitudinal focussing is
provided by independently phased, double gap, spiral resonator structures
positioned before the second, third and fourth IH tanks. Quadrupole triplets
placed after each IH tank maintain transverse focussing.
A schematic drawing of the DTL is shown in Fig. 1.
When operating at full voltage, the beam dynamics resemble that of a so
called `Combined 0 
Synchronous Particle Structure'[5].
To achieve a reduced final energy, the higher energy IH tanks are turned off
sequentially and the voltage and phase in the
last operating tank is varied. The spiral resonator cavities are adjusted to
maintain longitudinal bunching. In this way, the whole energy range can be
covered with minimal longitudinal emittance growth.
synchronous phase, three double gap
spiral resonators (B) provide longitudinal focus ( 
)
and quadrupole triplets (C) provide transverse focus. The beam envelopes
define the x and y maximum half sizes of the beam and the maximum energy
spread and phase spread in the beam as a function of linac length. The
calculations are for a beam of q/A = 1/6 with
matched elliptical emittances of 0.25 
m
(normalized) transversely and 48 
keV 
nsec longitudinally.
The physical specifications of the DTL have been determined and the
beam dynamics studied using the code LANA[6]. MAFIA
has been used to model the rf characteristics of the IH tanks.
Due to the relatively small longitudinal and transverse emittances of the beam
injected into the DTL ( 
keV ns and
m (normalized))
an rf frequency of 105MHz was chosen, three times
the RFQ frequency. Each IH tank has a diameter of 94cm with
the resonant frequency tuned by optimization of the ridge geometry.
The gross specifications of the five IH tanks and the three spiral resonators
for the design particle of q/A = 1/6 are
given in Table 1. The Q and shunt impedance values of the IH structure
are calculated with MAFIA.
The chief design considerations for the DTL are the cw operation, and the
variable energy requirement. To achieve efficient acceleration and to
distribute power losses uniformally,
a constant gradient IH structure is adopted.
The gap length to cell length ratio 
is tuned to flatten the
field distribution[5]. Maximum accelerating gradients are
determined by restricting the total power per unit length to less than
20kW/m. (For power calculations the quoted MAFIA shunt impedance
values are scaled by 75%.)
The IH structure is highly efficient with a total effective gradient of
1.5MV/m and an rf power consumption estimated at only 61kW.
Table 1: Summary of parameter specifications for each IH tank and spiral
resonator cavity (B1-B3) for the design particle of q/A = 1/6.
All cavities operate at 105MHz. Here L is the
length, a is the aperture, 
is the average cell length,
is the effective field gradient and 
is the
power per unit length. Other parameters have their normal designations.
The quoted IH shunt impedance values are from MAFIA. The power/unit length
and power calculations for the IH structure assume a shunt impedance
75% of the value quoted.
In the case of the buncher cavities, the shunt impedance is quoted from
reference [7].
The variable energy requirement sets restrictions on the tank
and quadrupole lengths, and on the specifications of the bunching cavities.
Any tank and triplet combination is required to be short enough to limit the
phase spread entering
the next section to 
so that the beam can be bunched
without longitudinal emittance growth. This requirement forces a short
first tank of 9 cells and 27cm with corresponding reduction in shunt impedance.
The quadrupole triplets are designed to be very compact.
Each triplet unit has an effective length of 32cm, with a bore aperture of
28mm and a maximum gradient of 63T/m. They will occupy a 40cm
space between tanks.
The two gap spiral resonator structure is chosen for its large velocity
acceptance and large multipactor-free voltage range.
The three bunchers must operate over 
regimes given by 1.8% 
2.2%, 1.8% 3.1%,
1.8% 4.1% respectively and over voltage ranges varying
by a factor of ten or more. For ease of manufacture, three resonators with a
constant 
have been specified yielding a gap crossing
time constant of at least 75% over the whole velocity range. The
properties of the device have been studied extensively elsewhere
[7] and the quoted shunt impedance value is taken from the
literature. We are presently modelling the device with MAFIA.
Beam dynamics calculations have been done using the code LANA
with 
Na 
as the reference particle.
All transverse emittances quoted below are normalized values.
The calculated
envelopes for the full energy case
are shown in Fig.1 for matched elliptical emittances of 0.25 m
and 48 keV nsec. The longitudinal optics is typical for
a 0 structure. The beam is injected into each accelerating structure
with an energy higher than that of the synchronous particle.
The longitudinal phase space
position rotates 
/2 in each tank
and remains primarily in the second quadrant providing a stable transport.
The strong
periodic longitudinal and transverse focussing yield small
beam sizes and increased acceptance.
The true useable region of the longitudinal
acceptance corresponds to 144 keV nsec and the transverse
acceptance is 1.3 m.
An 11.7MHz time structure is imposed on the beam from the separator
( 
m) by a pre-buncher
upstream of the RFQ (35MHz)[3]. Particle simulations through the RFQ,
stripping foil and pre-DTL matching section produce realistic
particles that are subsequently run through the DTL. A summary of
before and after emittances for two MEBT conditions (Case A and Case B)
are given in Table 2. Case A includes a bunch rotator before
the stripping foil while Case B has none.
The transmission is 100% in both cases for an ensemble of 4000 particles.
A plot of the tank voltage and phase required for a given final energy are shown in Fig. 2. For a reduced voltage the particle bunch is phased negatively with respect to the synchronous phase so that as the particles lose step with the synchronous particle and drift to more positive phases they gain the maximum possible energy. This phase setting also coincides with the minimum phase spread at the exit of the following triplet section. The buncher following this tank is then used to capture the diverging beam. The following bunchers provide longitudinal transport. Simulations show that in this way the longitudinal emittance growth for typical beams can be kept below 15% for the whole energy range.
The final phase spread of the beam exiting the DTL for either none, one, two, or three bunching cavities is shown in Fig. 3. The aim is to achieve a phase spread no larger than can be bunched with a 35MHz buncher in the HEBT. The plot shows that the first and second bunchers are essential and that the third would improve the beam quality in the energy range from 0.5-0.7MeV/u.
.
Figure 3: Results of LANA studies showing the final phase width of the beam
exiting the last tank for either none, one, two or three bunchers.
Superimposed on the figure are the tank energy
regimes (vertical lines) and the
phase width corresponding to 60 of
width in a 35MHz buncher (horizontal line).
All IH tanks were simulated with MAFIA. The results of the shunt impedance
and Q calculations have been reported in Table 1.
The calculations give power density information which will then be
used to determine the cooling requirements. MAFIA was also used to predict
the appropriate dependence to flatten the field distribution in
an 11 gap full scale model (see below) and to predict the performance of
various tuners.
A full scale model of an 11 gap tank was built to test the tuning of the
field distribution and the characteristics of various mechanical tuners.
The tank is built from a rolled copper sheet 6mm thick. The end plates
are made from aluminum on which a thin foil of copper is glued. The ridge
and the drift tube are made from brass to ease the fabrication. All
pieces are bolted together to facilitate changes of the cavity geometry.
The coupling loop and the tuner are installed in the median plane, one on each
side of the tank. The field distribution from a bead pull measurement is
given in Fig.4(b). The associated values predicted
from MAFIA simulations are shown in Fig.4(a).
The drift tubes will be cooled by a water circuit coming up
through the ridge and into a drilled out portion of the stem.
A NC machined copper model of a stem and tube was made
to test the mechanical rigidity of the structure under various heat and
water loads. The tests show that a water flow of 3 
/min is sufficient
to cool the drift tube. The water flow does not produce any measurable
mechanical vibrations.
values for each gap (a).
The separated function DTL concept provides a low power solution to
achieve variable energy heavy ion acceleration in the low regime
without significant increase in the longitudinal emittance.
A mechanical concept for the DTL is being discussed. The first module
consisting of Tank 1, a quadrupole triplet and a double-gap buncher is
scheduled to be completed by the end of 1997. The DTL is scheduled to be
fully installed by the middle of 1999.
The authors would like to thank U. Ratzinger for pleasant and informative discussions concerning IH structures, and P. Ostroumov and D. Gorelov for their assistance with the LANA code. Thanks go also to summer researchers N. Sy and T. Duncan for their patient computations with MAFIA and LANA.
