Project C: Construction of MAMI C - A1-Collaboration · · 2004-04-29Project C: Construction of...

Project C: Construction of MAMI C

Dr. Karl-Heinz KaiserInstitut fur Kernphysik

Universitat MainzJ.-J.-Becher-Weg 45

55099 MainzTel.: +49 6131 [email protected]

Dr. H. EuteneuerInstitut fur Kernphysik

Universitat MainzJ.-J.-Becher-Weg 45

55099 MainzTel.: +49 6131 [email protected]

261

262 CHAPTER 2. REPORT ON THE PROJECTS

PROJECT C: CONSTRUCTION OF MAMI C 263

C.1 Overview

H. EUTENEUER, K.-H. KAISER,G. ARZ, K. AULENBACHER, R. BARDAY, R. BOLENZ, O. CHUBAROV, M. DEHN,F. FICHTNER, M. GOBEL, B. GUTHEIL, F. HAGENBUCK, R. HERR, A. JANKOWIAK,P. JENNEWEIN, W. KLAG, H.-J. KREIDEL, U. LUDWIG-MERTIN, G. MEYER, J. MULLER,A. NUCK, U. REISS-FLUHR, J. ROTHGEN, J. ROSCHE, St. SCHUMANN, B. SECKLER,G. STEPHAN, H. STEPHAN, A. THOMAS, V. TIOUKINE, M. WEIS, G. WOLL,Th. ZSCHOCKEA.S. ALIMOV∗ , G.A. NOVIKOV∗, V.I. SHVEDUNOV∗

∗Inst. of Nucl. Physics, Lomonosov-University, Moscow

The availability of an extended electron beam energy range by an upgrade of MAMI from855 MeV to 1500 MeV is a prerequisite for most of the experimental program planned in thesubprojects H1 to H7. For this upgrade the construction of a fourth stage of the MAMI ac-celerator chain is in progress. It is realized as a Harmonic Double Sided Microtron (HDSM),where the beam is recirculated 43 times through two antiparallel linacs by two achromatic pairsof 90 bending magnets. The transversal focussing was simplified by the implementation of afield gradient in the bending magnets. For a good stability of the longitudinal beam dynamics,this requires the harmonic operation of the two linacs. One linac is working at 4.90 GHz andthe other one at 2.45 GHz, the standard MAMI frequency. All components for the HDSM havealready been delivered or are under fabrication. The HDSM was designed to fit into two exist-ing underground halls, which were joined by removing walls. Their electric power and watercooling system was refurbished for the new requirements. The existing beam transport systemwas also upgraded to deal with the higher end energy.

Rf-SystemThe rf-system of the HDSM consists of 5 modules for the 2.45 GHz-linac (one klystron / oneaccelerating section) and 4 modules (one klystron / two sections) for the 4.90 GHz linac, plus a4.90 GHz matching section module at the HDSM-injection. By end of march 2004 the opera-tiveness of all components needed has been ensured. The development of a new 4.90 GHz / 60kW cw-klystron by THALES Electron Devices was finally a success, after a delay of 29 monthsand building 11 prototype specimen. For 2.45 GHz two types of 50 kW cw-klystrons are al-ready working at MAMI B. A prototype 4.90 GHz accelerating section was designed and builtfully in-house, and successfully high power tested in July 2003. The order for the 10 seriessections was given with some flexibility to two suppliers, ACCEL Instruments / Bergisch Glad-bach and PMB / Paris, which both will deliver a demonstration section till October 2004 andthe full series latest till end of 2005. For the 2.45 GHz accelerating sections the contract withINP / MSU & TORIY / Moscow had to be cancelled half way, because a series of failures anddelays. Three of six sections were delivered and successfully tested, the remaining are underfabrication at ACCEL. The two 30 kV / 27 A HV-supplies for the klystrons were manufacturedby BRUKER / Wissembourg and are mounted at IKPH. All prototype waveguide componentswere high power tested, at 4.90 GHz by the half-way prototype klystron PT6 from THALES;after that the series specimen were ordered and are delivered. The mounting of the high powerwaveguide system in the HDSM-hall will begin mid of 2004. For the low power rf-distribution-and control-system all subunits are ready manufactured, for the most part in-house. Theirmounting and connection in the accelerator hall similarly will begin soon. Concerning therf-monitor resonators 16 XY-TM110-Cavities for beam position measurements on the injectionand extraction lines of the HDSM are ready built and tuned. The altogether 12 low-Q cavities


for position and phase / intensity measurement of the 10 ns diagnostic pulses on the two linacaxes are ready at 2.45 GHz. For 4.90 GHz a prototype was built and the series is in work.

HDSM MagnetsThe iron cores of the four 90 bending magnets for the HDSM were manufactured to our fullsatisfaction by USINOR and SFAR, France in accordance with the drawings provided by ourinstitute. Mechanical controls and first magnetic measurements were conducted in close col-laboration between these companies and our institute. They have been delivered with somedelay from Nov. 2001 through December 2002. The coils were also designed cooperatively bythe institute and the manufacturer Sigmaphi, France. Four high precision and reliably operatingpower supplies were purchased from Danfysik, Denmark. The sophisticated task of transporta-tion of the four magnets along the narrow path into the accelerator halls and their assemblythere were completed by the end of January 2003 by Grohmann-Attollo, Berlin. In 2003 allhigh precision magnetic measurements needed for the construction of surface correction coilswere completed. The first correction coil, produced by water jet cutting, will be finished inApril 2004. Further correction elements to precisely control the beam orbits, like small dipoleson the return paths, shims and pole coils are under construction and will be finished by the endof 2004.

Injection and ExtractionDuring the actual period under report the design of the injection and extraction beam line hasessentially been completed, and the magnetic components were ordered for delivery by August2004. Previously, in 2001, the MAMI B beam line had been modified for the inflection of the1.5 GeV beam and its transport to the experiments. In the same period the power and coolingsystems had been upgraded and a new experimental area for the X1-collaboration had beeninstalled.

Vacuum SystemThe design of the vacuum system for the HDSM and its injection and extraction beam linesis nearly completed and most of the components have been delivered or ordered. In October2003, the first vacuum chamber for the HDSM-magnets has been constructed and welded inthe institute. The three others will be fabricated in 2004 after slight modifications initiated bythe experience made with the first one.

ControlIn addition to the large and medium sized power supplies purchased externally, five 128 channelpower supplies have been developed in the institute to feed the numerous correction dipoles andnew solid wire quadrupoles. They will be ready by April 2004. Because the number of beamsignals produced by short diagnostic pulses will be doubled to nearly 2000 by the operation ofthe HDSM, a new fast multiplexing unit has been developed for a very flexible display of thesignal groups on oscilloscopes and for digitising them to enable automatic beam corrections.

Polarised ElectronsDuring the reported period polarised source operation has seen considerable improvements.The stability of operation was such, that downtimes due to polarised source failures duringthe production runs of about 5000 hours were extremely small (< 1%). Beam parameters forthe parity violation experiment A4 were improved by an order of magnitude, which allowedfor controlling helicity correlated effects down to a level where they are small compared toother error sources. Spin rotation with a Wien filter system at the 100 keV injection energy has


been successfully demonstrated and will be used as a solution for spin tuning at MAMI C. Thespectroscopic studies of fundamental properties of photo cathodes with high resolution in thetime domain have revealed strong evidence that the present restriction to 80 % of electron spinpolarisation is not fundamental; an increase to almost 90 % has been demonstrated by a timeof flight separation technique.

C.2 The Harmonic Double Sided Microtron

C.2.1 Rf-System 2.45 GHz and 4.90 GHz

The rf-system for the two linacs of the HDSM consists of 5 modules at 2.45 GHz (one klystronfor one accelerating section) and 4 modules at 4.90 GHz (one klystron feeding a pair of sec-tions). In addition a fifth 4.90 GHz-module is necessary for longitudinal matching on the in-jection path from RTM 3 to the HDSM (cf. Fig. C.1). For the accelerating structures at bothfrequencies the same biperiodic on axis coupled type was chosen as already successfully in useat MAMI B. Concerning the klystrons, at 2.45 GHz the two types in operation at MAMI B wereprincipally ready available, whereas at 4.90 GHz a totally new development was necessary.

Figure C.1: Schematic Layout of the Harmonic-Double-Sided-Microtron.

C.2.1.1 2.45 GHz Accelerating structures

The 2.45 GHz-sections are nearly identical to those in operation at MAMI B. They consist of33 nose cone shaped accelerating cells (AC) and 32 pancake like coupling cells (CC), theirelectrical length thus being 2.02 m. The coupling factor by a pair of coupling slots in eachweb is kAC/CC = - 4.1% (magnetic). The main quite tight specifications we defined for theirfabrication were: vacuum resonance frequency 2449.532 / + 0.1, - 0.25 MHz; passband gap(g = fCC − fAC) < ±0.4 MHz; quality factor Q0 > 15000 (machining of high surface qual-ity!); match of power input with a vswr = 1.13± 0.09 overcritical and an AC-field unflatness


< ± 3 %. To avoid the difficulties arising in brazing 2 m high resonator staples, the sectionsconsist, as those of MAMI B, of three parts: two identical long side parts (“half sections”) arebolted to the input coupler AC by a special vacuum / rf-seal.

The order for manufacturing six of them was given in April 2000 to the colleagues of theaccelerator group at the Institute of Nuclear Physics / Moscow State University (INP / MSU).This team had shown an excellent qualification for the treatment of this type of structure;during the early 1990’s they built several similar sections for their own accelerator project atINP / MSU. By following up publications and during personal discussions they demonstratedto us, that these capabilities were conserved over the last years. In addition, their price for thesix sections was only 63 % resp. 38 % of that demanded by the two other bidders, ACCELInstruments GmbH and AES-Northrop Grumman. For machining and brazing the cavity seg-ments the klystron manufacturer TORIY acted as a subcontractor. The time schedule was todeliver a prototype section in April 2001 and, after its high power test (HPT), fabricating the5 series sections within 13 months. The implicit long term planning was naturally that, afterthis work finished, also the 4.90 GHz-structures would be manufactured by the INP / MSU &TORIY collaboration. However, an unexpected series of failures and delays forced us to termi-nate the 2.45 GHz-contract halfway in November 2002. The causes for these failures may besummarised as follows:

1. INP / MSU & TORIY did not succeed to establish a tuning and brazing process for thecavity segments, which resulted in a reproducible and thus pre-settable change of the reso-nance frequency and especially the passband gap from the just pressed to the brazed statusof a resonator staple. Changes in gap of up to 5 MHz occurred, a factor of four highercompared with the worst cases we had in the 1980’s during the in-house manufacturing ofthe MAMI B sections. As a consequence sophisticated procedures and tools for retuninga brazed section had to be applied, e.g. by introducing an axial hammer into the sectionsfor individually fine-bending webs, the sensitivity of AC- and CC-frequencies to this op-eration being 1:8. Clearly the level of cleanness inside the cavities was (further) reduced,and most probably by this contaminations enormous and sometimes insurmountable anddestructive (strong detuning of CC’s) multipactor phenomena occurred during the HPT’sat Moscow and Mainz.

2. Several times a severe lack of carefulness occurred: E.g. two times stainless steel centringpins were lost into a section which, because of their high specific resistance, were moltenand partly evaporated during the HPT, in this way again causing severe multipacting. Onehalf section was launched for transport in an inadequate package and arrived at Mainz bentby several millimetres, having thus a by 30 % reduced effective beam hole aperture.

3. The equipment for quality control before and for measurements during the HPT’s turnedout to be partly insufficient. The lost steel pins were already mentioned above. One sectionwas destroyed by operating it during the HPT at Moscow in an auto-oscillation loop withthe klystron, thus accidentally working at a non π/2-mode with high fields in the CC’s;this procedure was done because of lack of a stable driving oscillator and a high powercirculator for the 50 kW-klystron.

4. The delivery of auxiliary equipment (e.g. a technoscope and a high power circulator) sentto INP for loan, to improve this situation, was often strongly delayed by custom difficulties,as well as the transport of several sections sent back for a guarantee repair.


Altogether, out of twelve half sections fabricated by INP / MSU & TORIY, only six had asufficient quality to be mounted with an input coupler to give three 2.45 GHz sections, carefullyand finally satisfactorily high power tested at Mainz, the last one in February 2004. For two ofthem quite strong multipacting had to be overcome; however, by flooding them several timesjust with surrounding air – thus simulating the worst operational accident of a broken ceramicrf-input vacuum window – it was verified that the rf-conditioning was permanently successful.

Because of all these difficulties in December 2002 a second manufacturer – ACCEL Instru-ments / Bergisch-Gladbach – got the order to produce the remaining 2.45 GHz-sections neededfor the HDSM. The prototype demonstration section was envisaged to be delivered in January2004, but will be delayed to April because of the heavy workload of ACCEL by other projects.However, we are confident that after its successful testing at IKPH the series sections can becompleted by end of this year.

To make the production process of these sections by ACCEL easier and speed it up as far aspossible, their design was modified a little: they will now have four instead of only two tuningplungers, two of them for a permanent tuning correction in case of aberrations after brazingand two, as up to now, for regulating precisely the section-phase. In addition, the end cellswere modified by a thin end wall for tuning after brazing. The maximum possible tuning rangeof a section is thus increased from ± 0.4 MHz by ca. a factor of three and consequently thepertaining specifications were relaxed. During the numerous tests of the INP / MSU-sections itturned out (and was confirmed by LOOP-simulations), that only a very good symmetry of theleft and right half section of a structure, concerning resonance frequency and passband gap,guarantees acceptable field flatness and low fields in the CC’s. Therefore, the specificationswere updated in this respect (difference in fres ≤ 0.35 MHz and g ≤ 0.8 MHz for left and righthalf sections).

The new control electronics of the sections tuning plungers has been extensively tested duringthe HPT’s. The two regulating tuners are moved by step motors controlled via a PC, so thatvarious operation modes can be very flexibly programmed, e.g. a different penetration depth ofthe tuners for symmetrising or conditioning a section. The tuner regulation is the final chain-link for the phase stability of the two linacs of the HDSM. It has been long term tested tostabilize the phase of a section (formula: ∆φ = arctg(Q0 ·∆ f/ fres)) to better than ±0.04 in anenvironment causing phase wandering of ca. ±0.8. The resolution for phase control of the stepmotor driven tuners is 0.017 per step, in contrast to that a cooling water temperature variationof 0.1C would cause a phase deviation of 1.4.

C.2.1.2 4.90 GHz Accelerating structures

The 4.90 GHz accelerating structures are not just a 2:1 scaled copy of the MAMI B sec-tions, but quite substantial optimising changes were done for their resonator-geometry (cf. [1]).Therefore, in spite of extensive computer simulations, their design left uncertainties, naturallyfor details of practical manufacturing, but also e.g. concerning the multipactor and thermo-mechanical behaviour. Consequentially, to be able to give qualified specification data and fab-rication advises to external manufacturers, a prototype section was machined, tuned, brazedand high power tested fully in house (cf. Fig. C.2).

Some details of this in house manufacturing are in short (cf. [2]):


a) Machining of half resonator segments with a roundabout “skin” of 0.2 mm. / Stress relievingannealing of these OFHC-Cu parts at 400 - 500C in the vacuum furnace. / Final machiningof them with a surface quality as high as possible (skin depth in copper at 4.90 GHz: 0.93 µm)/ Results: A quality factor Q0 = 10600 ± 200, ca. 99 % of the ideal value was achieved, andthe full range of resonance frequency deviations for 35 segment pairs was δ fAC = 1.2 MHz,δ fCC = 4.1 MHz.

b) Defining the frequency-preset for brazing deformations: Adjusting the pressure for fre-quency measurements of resonator segment staples to 1.4 tons; at this level the increase ofthe Q-value stops, i.e. good rf-contact. / Tuning several short stacks of resonators and brazingthem; most brazings at 780C with Ag Cu-eutectic alloy. At the same time finding the optimumquantity of alloy. / From the difference in resonance frequency and gap before and after braz-ing getting the presets ∆ fAC = + 1.3 MHz, ∆g = - 3.6 MHz. / Final result for brazing the left andright half and then the full section was: deviations from these preset values in frequency andgap by +0.5 MHz resp. +0.3 MHz.

Figure C.2: The 4.90 GHz prototype section at the high-power-teststand.

c) Additionally including the necessary frequency and gap preset for the dimensional changesof the AC and CC by high power input (Pin = 15 kW operational, 22 kW max. test value): Forthe frequency change the following formula was used, resulting from a combination of heatconduction and thermoelastic calculations and taking into account experimental results (cf [3])from the 2.45 GHz MAMI B sections:

∆ fres [MHz] = −0.0823 ·Pin[kW] · (7.17

D+

aDx +0.447) (C.1)

The first term gives the average warm up of the section for a cooling water flow of D[litre / min],and the third accounts for the effective warm up and the elastic deformation by thermal gra-dients of the copper cavity profile. Inelastic deformations of the copper being extremely soft


from brazing, by high power input, which were noted for the MAMI B sections, can onlybe determined experimentally. Finally, the second term results from the temperature jump atthe boundary layer copper / cooling water. Depending on which handbook is taken to find theReynold-, Nusselt-, Prandtl-number etc., one gets a = 7.3 to 15.2 and x = 0.8 to 1.2. However,the uncertainty by this is only ± 0.12 MHz for a total frequency change of ∆ f res = 1.09 MHz at15 kW input power and the standard coolant flow of D = 41 l / min.

d) With these data fine tuning the resonator segments, matching the input coupler, and then, af-ter brazing of the section, mounting and calibrating the eight diagnostic antenna probes. Finallymeasuring the spectrum of the section and its field flatness (mainly influenced by deviationsof the dimensions of the coupling slot pairs) by bead pull measurements for different tunerimmersion depths.

The main data obtained for our prototype were:Electrical length L = 1.071 m, consisting of 35 AC and 34 CC (cf. Fig. C.3). Shunt impedancer = 81.5 MΩ/m, quality factor Q0 = 10600. First order coupling kAC/CC = - 8.7 %, second ordercouplings kCC/CC = - 0.7 %, kAC/AC = 0 %. The AC-field unflatness was measured to ≤ ± 2 %for the standard and ≤ ± 3 % for the extreme in / out tuner positions (section detuned by± 2.2 MHz). This is a very satisfactory confirmation of our coupled loop simulations by LOOP,and justifies our effort to reach an as high as possible coupling coefficient kAC/CC by MAFIA-optimising the resonator geometry. A by 3 - 4 % lower field in the periodicity breaking endcells of the section, irrelevant for its practical use, cannot be explained till now. These cellswere tuned to the same frequency as the standard cells; this according to results of careful AC-field flatness and minimum CC-field measurements on short resonator stacks with an end cell,prior to brazing.

Figure C.3: Drawing of a 4.90 GHz MAMI C linac section.

The movable tuning plungers for 4.90 GHz were built very similar to the lately revised design(cf. [4]) of the well proven and tested 2.45 GHz-tuners (cf. [3]), a trivial difference being, thatthe diameters of the sections tuner hole and the tuner head immersed into it are 14 / 10 mminstead of 24 / 18 mm. The guiding contact fingers of the tuners are made rf-current free by achoke technique. – Together with the 4.90 GHz section also its rf vacuum window, deliveredby THALES, was tested. However, because this window is nearly identical with the rf outputwindow of the 60 kW-klystron TH2166 (see below), no problems were to be expected for a testgoing only up to 22 kW.

The high power test of this prototype section was performed in July 2003. After a quite shortconditioning time of 5 days, interrupted by many control measurements of the development


of the sections spectrum and field flatness, it accepted without problems a cw-power of upto 22 kW. There was no indication that any multipacting thresholds had to be overcome. Theresults for the dynamic reversible as well as the permanent irreversible changes of resonancefrequency and passband gap with input power are given in the diagrams a) and b) of Fig. C.4. Itshould be noted, that naturally the irreversible changes do not progress, if one does not exceedthe maximum power applied during the test.

After this trouble-free and very successful test it was possible to launch the correspondingorders to external manufacturers. For these orders, slight modifications of the sections designwere done, especially concerning the orientation of the coupling slot pairs in the input cell andthe cells with diagnostic probes, which was not optimally chosen for our prototype. Because ofthe difficulties having occurred with the 2.45 GHz-sections at INP / MSU and also for a deliverytime as short as possible, it was decided not to rely on only one company. Therefore ACCELgot the order to fabricate a demonstration prototype and, after its test and release (expected tobe ready by October 2004), at minimum five series sections. PMB / Paris, a company with lessimmediate experience in this field, will fabricate their prototype also till October 2004. De-pending on its quality, they or ACCEL will get the order for the remaining five series sections.According to the outcome of this decision, the 4.90GHz-sections will be available betweenAugust and end of 2005.

To protect the accelerator sections with their small beam hole ( /0 = 14 resp. 10 mm at 2.45 and4.90 GHz) against direct electron beam impact, a special collimator with a resp. 0.5 mm smallerhole will be inserted in front of each linac. The beam absorbing part of it consists of an 80 mmlong water cooled copper cylinder and a 20 mm thick disc made of Densimet 170K (essentiallytungsten), inserted at its front and rear end. The supervision to switch off the beam within afew seconds in case of a beam impact will be done by measurement of the radiation, and foradditional safety also by the temperature of the cooling water.

C.2.1.3 2.45 GHz Klystrons

At MAMI B two types of 2.45 GHz-klystrons are in operation: the VKS7960-M made byCPI / Varian and the TH2075 from Thomson-CSF, since 04/2001 THALES Electron Devices.Both work altogether satisfactorily with good lifetime results (now up to 80000 h). The mainparameters of these tubes (operating voltage and current, max. output power, gain and effi-ciency) are nearly identical.

The TH2075 was designed 30 years ago and therefore has a quite labour-intensive mechanicalstructure. Moreover, it showed a very critical focussing behaviour, which required often a quitecumbersome magnetic tuning and shimming operations, to get the losses of its 3.4 A-beambelow the 100 mA body-current limit. However, THALES offered for the six tubes, demandedby us for MAMI C, a widely reengineered design as TH2174. This tube has a mechanicalstructure adapted to cost-effective modern CNC-machining. The beam dynamics is the sameas for the TH2075, but a new gun with superior long life Osmium coated M-cathode and animproved collector (max. dissipated power 120 kW instead of 95 kW) are installed. Above all,however, the focussing behaviour (in our standard TH20049B solenoid) was promised to bedistinctly improved: by a more sophisticated guiding of the output waveguide through the upperpole plate of the tube, reducing thus asymmetric field distortions of the magnetic end-fields,and by the beam hole of the output cavity being enlarged from 8 mm to 10 mm. This measure


Figure C.4: The two graphs above show the high power behaviour of our 4.90 GHz prototype sectionconcerning a) reversible and b) irreversible changes of resonance frequency fres and passband gap g.


sacrifices only 3 % of the dc to rf efficiency (58 % → 55 %). The order for six TH2174 wasgiven to THALES in 12/2001 and the time schedule was, that by 12/2002 the prototype andby 7/2003 the complete series should be delivered. This planning was delayed by altogetherone year; mainly by an obvious overloading of the scientific klystron development group atTHALES, but also by two quite trivial technical mistakes: the first version of the new outputwaveguide did not fit into the focussing magnet by a mistake in the CAD design, and theperveance of the new gun was 30 % too high by an erroneous mounting distance cathode toanode. However, the Factory Acceptance Test (FAT) of the prototype in 1/2004 then showed areally good natured behaviour of the TH2174: the body-current (i.e. beam loss from cathodeto collector) is only 50 % of the typical TH2075-value for a wide range of focussing currentsin three different solenoids (old and new series), and the efficiency is 56 %. Two of these newTH2174 klystrons, the second confirming the FAT-data of the prototype, were delivered in2/2004; the remaining 4 specimen are promised by August 2004.

C.2.1.4 4.90 GHz Klystrons

The new development of a 4.90 GHz / 60 kW cw-klystron was one of the crucial technical de-mands for the rf-system of the HDSM. Our main specifications given in the call for tenderwere: a cw-output power of 60 kW at a max. operating voltage of 27 kV (for the same type ofpower supply being usable at all HDSM-klystrons), an efficiency ≥ 55 % and a gain ≥ 47 dB.

A very important demand was in addition, that the klystron has a smooth and monotonouslyrising rf-power transfer curve Pout = f (Pin)! At MAMI the klystrons are operated below sat-uration, because without countermeasures the rf-output has an amplitude ripple of ca. 1 %ppand a phase ripple of several degree, both a result of the high voltage ripple of 0.4 %pp. Thisis regulated down to less then 1 · 10−3 and 0.1 degree by feedback-loops acting on the easilyaccessible rf-input power and phase. Naturally, this way of controlling the amplitude sacrificesa considerable part of the inherent efficiency of the klystrons; the effective dc to rf efficiencyis here only 28 % resp. 38 % for “zero” and 100 µA electron beam current. However, technicalinvestigations done during the construction of MAMI B had shown, that it would be nearlyimpossible to get an rf-amplitude stability of some 1 ·10−4 with a bandwidth of several 10 kHzby e.g. steering a modulating anode of the klystron.

Neither of the two bidders, THALES and CPI / Varian expected any principal difficulty to buildsuch a klystron. This was consistent with the published literature about the fundamental powerlimitations of klystrons at very high frequencies: e.g. in [5] a max. cw-power of 500 kW is es-timated for 4.90 GHz, and there exist several really built tubes at nearby frequencies for nearlythis power. The order for developing a prototype and then building 5 series tubes TH2166 wasgiven to THALES as the sole manufacturer in 10/2000. Splitting the contract in equal partsbetween the two bidders would have raised the costs of this sub-project from 680 ke to ca.1010 ke. THALES’ time schedule was to present the prototype on a Critical Design Review(CDR) in 2/2001, manufacture it till 12/2001 and finish the production of the series tubes by8/2002.

On the CDR the paper-presentation of the TH2166 (operational parameters of tube and fo-cussing coil, outer dimensions, beam dynamical properties of the six-cavity interaction struc-ture) looked quite promising. Naturally, many technical details of the planned inner config-uration of the tube were treated as a corporate secret at this stage, as usual. – Thereafter the


process changed dramatically to worse: the manufacturing of a tube fulfilling our specificationswas delayed step by step for altogether 29 months and eleven specimen (PT1 to PT11) had tobe built. The causes for this can be summarised in note form as follows:

a) At the THALES klystron development crew an extensive alteration of generations, not an-nounced to us, took place; e.g. the designer of the TH2166 retired just during the attempts toset PT1 in operation. As it turned out later on, the first basic design errors of the interior tubestructure must be attributed to this engineer. Furthermore, the new, at the beginning rather in-experienced development engineers obviously underestimated the amount of carefulness andcritical control for building such a high power tube.

b) The first year of the delay occurred by failures resulting from a mixture of levity and slop-piness: The tests were performed with a sometimes partly defect interlock system of the highpower teststand. For a not sufficiently straight tube, because of making it in a cheap and quickbrazing setup, melting of the body by the 4.5 A beam occurred (diameter of the klystron res-onators beam hole only 6 mm). The same happened to another one, where it was forgotten toremove some ferromagnetic parts from the tube body before inserting it into the focussing coil.The next prototypes failed with vacuum window cracks, because by a swapped data base entryin the THALES purchasing system an insufficient ceramic material with too high rf-loss tan-gent was used. Another tube got unusable because of mechanical deformations caused by aninadequate fixing of the output waveguide.

c) In March 2003 PT6 fulfilled our specifications, except the very important one demanding asmooth and monotonously rising power transfer curve. This tube showed jumps and hysteresisloops for increasing resp. decreasing input power (cf. Fig. C.5), obviously caused by multipact-ing in its resonators. In spite of it being doubtful if this klystron could be used with our highprecision phase and amplitude regulation at the HDSM, it was bought aside of the main con-tract for a distinctly reduced price, because there was an imperative need to have just a powersource for our urgent 4.90 GHz accelerating section prototype and high power waveguide com-ponent tests. For this purpose the tube fully did its duty, however, the multipacting phenomenadid not diminish or vanish, but a little bit unexpected got strongly enhanced with the operationtime growing to some 100 hours.

d) THALES fully approved our claim of PT6 not being acceptable as the final prototype. Theywere very optimistic, that a coating of the klystrons penultimate and output cavity lips witha titanium layer (secondary electron emission yield of Ti < 1) would solve this “last” prob-lem. The multipacting phenomena really vanished. However, now it finally turned out, thatthe geometry and cooling channel system of these two highly rf-loaded cavities (dissipatedpower of up to 2 kW in a resonator of /0 = 26 mm and L = 18 mm!) was not adequate to handlethe heat load increased by 30 % by the Ti-coating. Therefore the tubes PT7 to PT9 got unus-able by thermal deformations and molten tuning mechanisms in cavities 5 and 6, because firstTHALES only tried to overcome the problem by a reduction of the Ti-coated area. However,PT10 works reliable up to 50 kW output power, and for PT11 THALES presented an improvedcoolant guiding, which realistically promised an output power limit very near to the 60 kWdemanded. On the FAT on 30.03.2004 this PT11 fulfilled all our specifications, except for aslightly lower saturated efficiency (53 % instead of > 55 %). It was accepted as the first seriestube, and the remaining five klystrons will be delivered till August. - It must be noted, that forthe 1.5 GeV-HDSM the real power the TH2166 must deliver to its two accelerating sections isonly 38 kW (2 x 15 kW for the rf-field and 2 x 4 kW for the max. beamloading by 43 x 100 µA).


Figure C.5: Transfer curves Pout = f (Pin) measured at the 4.90 GHz / 60 kW klystron TH2166-PT6 fordifferent values of the klystron voltage. Clearly visible are the jumps and hysteresis loops, at least twoon each trace, which move towards lower drive power and higher output power with increasing highvoltage. This clearly indicates multipacting phenomena inside the interaction structure.

The demanded power of 60 kW results of a factor of 1.35 as “Regelreserve”, to have enoughgain on the transfer curve for the rf-amplitude feedback-loop, and of 1.15 for waveguide lossesplus a margin for the Q-value variation of the industrially manufactured sections. These factorsare realistic but not very canny. However, for any thoughts concerning a higher output energyof the HDSM one must take in mind, that the rf-power needs to grow with the square of theenergy gain per turn.

In one respect the 2.45 GHz- and 4.90 GHz klystrons of THALES significantly differ in theirbehaviour, namely concerning the amount of harmonics in their rf-output. For the TH2075 /TH2174 tubes one had never to take note of such harmonic signals (and they lately were esti-mated by measurements to be below - 45 dBc). The TH2166, however, shows a strong harmoniccontent of ca. - 25 dBc. As a consequence at all diagnostic ports of the 4.90 GHz waveguidesystem appropriate filters have to be inserted. Naturally, the accelerating sections show no har-monic resonance.

C.2.1.5 Klystron High Voltage Power Supply

The two 810 kVA high voltage power supplies for the 2.45 GHz resp. 4.90 GHz set of klystronswere specified in our call for tender to be built in the well proven primary phase controlledmodulator technique, and not as switch mode power supplies. The advantage of the latter tech-nique, saving the crowbar, did in our opinion not compensate the danger of getting small highfrequency ripples beyond several 10 kHz on the output voltage, which cannot be sufficientlycompensated by our klystron phase and amplitude regulation.


Our specification was for a nominal power of 810 kVA = 30 kV x 27 A for six klystrons witheach max. 4.5 A collector current. The demanded regulation range was 20 - 30 kV, with a load-ing of 2 - 27 A and a HV-ripple of ≤ 0.4 %pp . A crowbar-circuit must limit the power depo-sition into a klystron-arc to less than 20 Joule. - The real operation of the two supplies willbe at ca. 460 kVA for the 2.45 GHz and 560 kVA for the 4.90 GHz tube set. The two powersupplies were offered by three companies: Jager / Kelkheim, FEAG-Siemens / Erlangen andBruker / Wissenbourg, the ratio of the prices being 2.0 : 1.8 : 1.0. The surprisingly low price de-manded by Bruker can be explained partly by the intelligent usage of standard substructuresalready built by this firm, and maybe partly by their desire to get access to the field of superhigh-power HV-supplies, resulting in a very tight calculation.

Some convincing details of the Bruker design were:

a) Each supply is a cascade of two 15 kV / 27 A six pulse rectifier blocks, phase shifted with aresp. autotransformer by ±15 to give an effective 12-pulse output. Thus the < 0.4 %pp HV-ripple can be achieved with a moderate smoothing capacitor of only 11 µF. Furthermore, eachof these two subunits can be separately tested at the factory into a special salt-water load up to300 kVA, 75 % of the nominal power.

b) Each 15 kV rectifier block is built up as a cascade of 24 standard 625 Volt-units, individu-ally connected to taps of the HV-transformer. By omitting some of these units the max. possibleoutput high voltage can be easily adjusted to different needs. Thus by a high operating angleof the thyristors one can always work at a mains to HV efficiency of ≥ 95 %. Moreover, byconnecting these units temporarily in parallel, a 625 V x 640 A = 400 kVA full power endurancetest for the thermal behaviour of all critical components can be done at the factory. The possi-bility of these quite extensive factory tests was important, because a load of the HV-supplieswith the full set of klystrons will take some more assembly time at IKPH.

c) The crowbar circuit is built as a series connection of thyristors, such getting rid of the usualignitrons or thyratrons, which are difficultly to handle and, because of their mercury filling,cumbersomely to be disposed of. Moreover, at the primary of the supplies a very fast IGBT-switch off is built in, which should prevent any perturbations of the mains if the crowbar isfired at full power.

During the tests of the two power supplies at the factory only a small modification of thecore of the HV-transformers was necessary, resulting in a delay of ca. 2 months. The crowbarworked satisfactorily, this demonstrated by the survival of a 0.15 mm diameter copper wireshortcutting the HV, which proves an energy deposition of less than 10 Joule from the 11 µFsmoothing capacitor. The two supplies are now under assembly at the MAMI C technical hall(cf. Fig. C.6).

C.2.1.6 High Power Waveguide System

On the 2.45 GHz-side the five rf-modules ( cf. Fig. C.7a) are nearly identical with those used atRTM2 and RTM3. Therefore here most of the components were ordered without testing themanew. An exception was the 3-port Y-circulator for 50 kW-cw power from AFT (AdvancedFerrite Technology / Backnang). For minimum attenuation and optimum isolation the prototypewas equipped with an auxiliary coil, correcting the 0.35 T field of the permanent magnet setupby ca. ± 0.005 T for minimum thermal losses at different forward and reverse power flow


Figure C.6: Floorplan of the HDSM in the experimental halls 1 and 2 together with the neighbouringpower supply rooms.

conditions. However, in discussions, if one could avoid this quite complicated active regulatorysetup, it turned out, that a nearly as good passive compensation can be done by insertion of“Thermoflux” sheets with a strongly temperature dependent permeability into the magneticcircuit. Therefore the 2.45 GHz Y-circulators were all built in this passively compensated wayand were successfully tested.

For the five modules at 4.90 GHz (cf. Fig. C.7b) the waveguide type R48 = WR187 = WG12 isused. It has more than twice the attenuation (4 dB / 100 m, i.e. 10 W/m dissipated at 1 kW powerflow) as the 2.45 GHz R26-waveguide; and also the max. electric field will be by a factor of 2higher, which the Kilpatrick-factor relieves only by 1.4. No references for a C-Band rf-systemoperating at a cw-power of several 10 kW were available. Therefore, at first only prototypesof the necessary components could be ordered resp. built in house, which had to be tested fortheir thermal and electric breakdown properties at a high power klystron. After the “half-way”prototype klystron PT6 was bought from THALES, these tests were performed immediately ina very short time. The results for the different components were:

Circulators: For a safe full power operation of the klystron TH2166 (and also the TH2174) avswr < 1.3 of their load is prescribed by THALES. Operating the tube as a driver for accelerat-ing sections, this can only be easily and safely ensured by using an appropriate circulator withan effective isolation of at least 18 dB, presenting a permanently matched load to the klystron.The 4-port differential phase shift 4.90 GHz-circulators developed by AFT were high powertested up to 56 kW into a short of arbitrary phase and showed an isolation of > 27 dB (vswr <1.10). Not any arcing was observed. Nevertheless, the waveguide arc-detectors mounted nearthem will have a quick direct connection to the rf-power interlock switch (ca. 10 µs switch offtime), bypassing the standard PC-interlock, which in the worst case can have a reaction timeof up to 6 ms.


Water Loads: The standard “diagonal quartz tube” long water loads used at MAMI B werenot any more available from THALES for a reasonable price. Therefore Spinner / Munich gotthe order to develop water loads for 2.45 GHz and 4.90 GHz in their ceramic block technique,where the waveguide is just closed by a carefully matched Al2 O3-disc, flown against by waterfrom the rear. These very compact loads were tested satisfactorily up to 50 kW resp. 60 kW.

Ceramic Phase Shifter: At 4.90 GHz the wavelength in the R48-waveguides (dimensions a x b =47.55 x 22.15 mm2) is 79.96 mm. Therefore, for the phase length from the 4.90 GHz klystron toits two associated accelerating sections to be defined by better then 1, the length of all waveg-uide components in the two arms behind the 3 dB-divider must be known to distinctly less than0.2 mm. To avoid this very difficult adjustment, in one arm a phase actuator with a range of ca.70 ≈ 16 mm will be included. It was in house developed and built and consists of a ceramicrod of /0 = 5 mm inserted with variable depth into the waveguide at an H-bend. Surprisingly, thestandard high purity Al2 O3-ceramic Degussit-AL23, used by us for many rf-resonator mea-surements (e.g. as a perturbation rod for shunt impedance measurements) without any indica-tion of Q-value deterioration, was not sufficient for this purpose. Inserted into a power flowof 20 kW it quickly heated up to 800C. A sufficiently low loss tangent ceramic available as350 mm long rods was finally delivered by Friatec (“Hochfrequenzkeramik” F99.7). The phaseactuator has a range of 70 with a sensitivity of 0.28 / mm and was successfully tested up to22 kW.

Waveguide: For all other waveguides components (E- and H-bends, 3 dB-power divider, di-rectional couplers) it was particularly tested, that their cooling structure, consisting of coppertubes brazed or soldered on the waveguide body, is sufficient to limit their warm up to lessthen ca. 10. From these tests it got evident, that for the 4.90 Ghz-system really all componentshave to be quipped with cooling tubes, whereas at 2.45 GHz e.g. the bends were just cooled viatheir neighbouring straight waveguides. A thermal problem was expected for the flexaguidesin the two side arms behind the 3 dB-coupler. They have twice the attenuation as a standardwaveguide and were specified by most manufacturers only up to 8 kW cw-power. Howeverthe specimen delivered by ProNova / Continental Microwave worked successfully up to 25 kW,cooled by carefully guided forced air.

C.2.1.7 Low Power Rf-Distribution and - Control

The rf-amplitudes for both linacs of the HDSM must have a jitter of less than 0.1 in phase andless then some 1 ·10−4 in amplitude. The klystrons TH2174 and TH2166 show an input to out-put phase variation of 9 resp. 15 per percent of change in high voltage, and their output poweris given by Pout = Ucath. · icoll. = p ·U5/2

cath., p being the perveance (ca. 0.8 - 1.1 ·10−6 A/V3/2).Therefore, with a ripple ripple of 0.4 %pp on the 26 kV high voltage, the rf-wave would fluc-tuate by 4 - 6 in phase and 1 % in amplitude.

The ten control units, suppressing these fluctuations by at least a factor of 50 up to frequenciesof several 10 kHz were built accordingly to the ones used at MAMI B; however, in a muchmore standardised way with improved diagnostic and calibration possibilities. They are readybuilt and tested. For these units, as well as for the whole low power rf-system, the meansfor in-house developement and partly also fabrication of the many necessary high frequencymeasurement and diagnostic strip-line components (directional couplers, hybrids, DC-blocks,


Figure C.7: Schematic view of the 2.45GHz and 4.90GHz rf-modules. BDC - double bidirectional cou-pler, HB-φ - Al2 O3-phase actuator, FG - flexaguide, AD - arc detector, HB - H-bend, EB - E-bend, KL- klystron, SE - section, PR - rf-probe, DL - dry load, WL - water load, FA - fan, VW - vacuum window,WG - water guard, 3 dB - 1:1 power divider, PSC - phase-shift-circulator (4 Port), YC - Y-circulator (3Port).


matched diodes, varactor phase shifters, pin attenuators and switches, amplifiers etc.) in quan-tities of 50 to 100 were established during the last two years. The basis is the widely usedsimulation software “Microwave Office” by Applied Wave Research (AWR). The advantage ofthis in-house proceeding is, that one gets the required rf-components with excellent data in anadequate narrow bandwidth of only some hundred MHz around the MAMI frequencies. On themarket one must either buy broadband components with only average compromised data or payvery high development costs. Especially useful was, that the large amount of bandpass-filtersneeded at the 4.90 GHz diagnostic ports, because of the high harmonic output of the TH2166,had not to be bought but could be produced in-house.

A schematic overview of the whole low power rf-distribution system of MAMI is shown inFig. C.8). The optimal positions and connecting ways of the necessary sub-division points areready designed. An extensive investigation was done to find a most phase stable (concerningits temperature stability) rf-coaxial cable, with low attenuation and usable up to at least 5 GHz.From ten cable types tested, the Times LMR400 turned out to be the best by a factor of eight. Itsphase / temperature stability (∆φ[] = K ·L [m] ·∆T []) is K = 0.06 /m·C and 0.10 / m·C at2.45 GHz resp. 4.90 GHz, with an attenuation of 0.24 resp. 0.35 dB/m. Just for comparison: thecoefficient K for a R26- resp. R48-waveguide is 0.07 resp. 0.13 / m·C, taking into accountonly their thermal copper expansion. For a waveguide open to surrounding air the effect canbe double, e.g. by a change of the relative air humidity of only 15 %. – The mounting of therf-distribution system will begin in autumn 2004.

Figure C.8: Schematic view of the MAMI C low power rd-distribution.

A procedure still to be worked out in detail is how to measure along the linac axes the pre-cise distance between the different accelerating sections, to be able to make a precise rf-phasesetting for them. An accuracy of one degree requires at 4.90 GHz a mechanical precision of0.17 mm. In contrast to e.g. RTM3, where the distance between adjacent sections is only100 mm and free of any installation, at the HDSM linacs focussing and diagnostic elementsneed to be installed between the sections, resulting in distances of up to 1800 mm.


C.2.1.8 Rf-Monitor-Cavities

Sixteen of the so called “high-Q” rf-monitors for cw-beam position diagnostic on the injectionand extraction beam lines of the HDSM are ready built. They are principally the same as theones already used at MAMI B: Two separated 2.45 GHz TM110-resonators in one unit; loadedQL = 4000; over-coupled with a total β = 2 to two symmetric antennas connected to a 180-hybrid for common mode rejection; having a mechanical adjustable tuning plunger for workingin different temperature environments. Only three TM010-intensity monitors are needed on thebeam lines and are available as a spare from MAMI B fabrication.

The three times two “low-Q” cavities needed on the linac axes for the detection of position,phase and intensity of the 10ns-diagnostic pulses on each of the max. 43 re-circulations, arefor the 2.45 GHz-linac just a copy of the MAMI B ones: One TM110-”resonator” with fourstrongly over-coupled antennas (QL = 30 by κ = 200) for combined X-Y-detection, and aTM010-”resonator” with a highly over-coupled loop for phase and intensity measurement, bothin one mechanical unit. For the 4.90 GHz low-Q devices some developing work had to be donebecause of mechanical constraints. The necessary minimum distance of their four end-wallantennas to the axis, because of the /0 = 13 mm beam hole and the vacuum flanges, prevented asimple scaling of the 2.45 GHz-type. A prototype was built and tested and the series specimenare in work.

C.2.2 Bending Magnets

GeneralThe four 90-bending magnets for the HDSM have been delivered from November 2001through December 2002. Transportation into the accelerator halls and assembly were com-pleted at the end of January 2003. In 2003 all magnetic measurements needed for the design ofsurface correcting coils were finished. After the construction and proving of test coils the finalcoil set for magnet no. 2 was ordered in December 2003. The main dimensions and parame-ters have been given already in [6]. For simplicity they are shown here again in Fig. C.9 andTable C.1. In the latter calculated data are replaced by measured values.

C.2.2.1 Mechanical and Electrical Properties

The iron bodies of the magnets essentially consist of two symmetrical pieces of 125 tons each,which touch at the horizontal symmetry plane. Additional plates of 200 mm thickness at theupper and lower yoke serve to equalize the flux density distribution, such that field deviations inthe gap with respect to a coordinate along the front edge are small. Stability against the gravi-tational and magnetic forces is achieved by 8 screws squeezing the two parts with 3200 tonsin total to each other. The remaining yoke bending, predicted by mechanical simulations to beabout 0.4 mm along the front edge, was taken into account in the design of the pole profile.Each magnet sits on three hydraulic supports constructed in the same way as those of RTM3.Thanks to their stability the resonance frequency for horizontal motion is about 700 Hz, so thatthe much slower seismical and building vibrations cannot excite it. The vertical vibration ofthe upper with respect to the lower pole has its main resonance frequency at about 70 Hz withamplitudes below 1 µm under operational conditions. Therefore, no deterioration of the beamis expected.


Magnetic induction in the air gap [T] 1.53 - 0.95Gap distance [mm] 85-140Mean orbit radius (max. / min.) [m] 2.23 / 4.60Maximum height / width / depth of iron body [m] 2.47 / 7.0 / 2.5Weight of iron body / copper coils [t] 250 / 6.85Electrical power consumption [kW] 72Coil current [A] 212Outer dimensions of copper conductor [mm2] 12*12Diameter of cooling channel [mm] 8Electrical current density in copper [A / mm2] 2.3Number of windings (both coils) # 512Number of cooling circuits (both coils) # 64Cooling water consumption m3 / h 15.6Pressure drop of cooling water bar 5Temperature increase of cooling water K 3.3

Table C.1: The main parameters of the HDSM-dipoles no.1 – no.4.

Figure C.9: The HDSM-dipole.


The pole profile, designed to obtain a field decay for compensation of the vertical edge de-focusing, consists essentially of 9 faces. These faces were machined in accordance with thespecifications defined by the institute for minimum short range field errors. The cutting inserts,their tool life, milling speed, head diameter etc. were found by tests using cast samples identicalto the iron of the magnets [7].

C.2.2.2 Coils

Apart from economical aspects the coils were designed to be highly reliable and to enable easyfailure detection and repair. They are wound in vertical layers with two conductors simulta-neously (each from its own roll), resulting in only eight windings per water circuit instead ofsixteen, allowing for a relatively small cooling bore diameter. All the 32 circuits per coil havetheir water and current connections outside the moulding so that they can be controlled sepa-rately. The connection scheme was designed by the institute starting from the scheme realizedby Bruker for the RTM3-coils. Each magnet is excited by a separate power supply deliveringmax. 260 A at 480 V.

C.2.2.3 Measurements and Controls in the Factory

The pole profile, surface quality and the linearity of reference faces have been scrutinised inthe factory for each block separately prior to removal from the milling machine. The maximumdeviations were found to be 0.1 mm for the profile and 0.1 mm for the linearity. At the upperhalf of magnet no. 3 inclusions of ceramic material were detected in an area of about 0.25 m2

of the pole. The defect has been repaired successfully by replacing the bad material on a grandscale using the shrinkage technique. After this experience, this magnet and the two followingones were in addition controlled magnetically in the factory. For this purpose the magnet wascompletely assembled and excited to about 30 % of its nominal field value. By means of a setof six induction coils mounted on a hand-driven carriage the magnet gap was scanned for fieldirregularities down to the 10−4 region. In dipole no.3 two areas were found with a field deficitof about 5 · 10−4 (see contour plot for dipole no.3 in Fig. C.11). Since errors of this order ofmagnitude can be corrected easily by the surface coils, the magnet was considered acceptable.

C.2.2.4 Transportation and Assembly in the Institute

The transportation of the 125 t blocks through the narrow apertures of the building was carriedout by a company (Grohmann-Attollo, Berlin) specialised in the movement of heavy loads(see Fig. C.10). They used sliding carriages on Teflon pillows for horizontal displacement andcompact hydraulic lifting devices for vertical movements. The combination of both allowedfor horizontal displacement of the hanging magnet block. A structure consisting of steel beamswas mounted in the entrance hall at ground level to support the lifting system. The ground levelfloor of the building had to be reinforced by pillars, guiding the forces to the stable basement.All transportations worked very well, and the magnet assembly, where the positioning had tobe done with a precision of a few millimetres, was carried out without problems.


Figure C.10: Transportation and assembly of HDSM-Dipole no.2.

C.2.2.5 Magnetic Measurements and Results

General Field PropertiesThe maximum deviation of the measured field profile from TOSCA-simulations is 1 % in theregion near the field maximum. Behind this region, where the field is continuously decreasingthe accordance to the design is 99.9 %. After optimising the coil currents the field distributionsof the four magnets now agree within 0.1 % to each other. The slightly steeper decay in allmagnets leads to a weak vertical focusing power of about 1 / 150 m and to a slightly higherphase slip of the bunches of about 0.8 with respect to the 4.9 GHz accelerating wave.

CyclingBefore starting the field mapping, each magnet was cycled for several days to obtain a stablerelation between coil current and magnet field measured at four locations in the gap. Onecycle consisted of a 6 min. current ramp to the nominal value, a 1.85 hour period of constantcurrent, an abrupt shutdown of the power supply and, finally, a 1.85 hour shutdown period. Asa measure of the reproducibility the field values taken at the end of the constant current periodwere compared to those of the preceding cycle. The deviations became smaller and smallerfrom cycle to cycle, decaying to about 10−5 after 12 cycles. This was considered to be stableenough so that the field mapping could be started.

Field MappingAfter the cycling procedure the field distribution was measured using three high precision hallprobes moved by a precise positioning machine in a grid of 12 mm·12 mm. The hall probeswere installed in a Plexiglas cube to measure the vertical field components in the magnet mid-plane as well as at a distance of 25 mm on either side of it. The cube itself slid on an air cushionon a Plexiglas plate aligned with a precision of about 0.1 mm in the gap. During the measure-ments the field stability was controlled by means of a NMR-probe. In view of a possible futureincrease of the maximum output energy the mapping has been done not only at the nominalfield value of 1.53 T but also at 1.64 T and 1.71 T.


Figure C.11: Lines of constant field difference in steps of 0.2 mT with respect to the field profile on acut along the x-axis in the centre of the magnets. Nominal field B = 1.53 T.


The contour plots depicted in Fig. C.11 show the results for the measured vertical field compo-nent in the midplane for the nominal field of 1.53 T of all four magnets.

Thanks to precise machining, relative deviations from the design field are only about 10−4 inthe inner area of the gaps. The two flat vales in the middle of dipole no.3, mentioned above,may have been caused by invisible inclusions. Much larger field decays exist, however, nearthe magnet corners. They have been predicted by TOSCA-simulations and lead to deflectionerrors of up to 2.2 mrad at low electron energy. In addition to these corner effects undulatingfield deviations can be seen in front of the pole edge. They can only be explained by deflectionsof the coil conductors of up to about 1 cm. The influence of all these fringe field errors to thebeam will be compensated by the combined action of iron shims and small correcting magnetson each beam tube (see below).

In Fig. C.12 the results of measurements at maximum fields of 1.64 T and 1.71 T correspondingto end energies of 1.61 GeV and 1.67 GeV, respectively, are shown for dipole no.1 as a repre-sentative example. As one can see in the pictures, the areas of decaying field at both sides ofthe pole area are increased and the field drop is steeper so that the correction would be moredifficult. The experiences that will be made during the construction and commissioning of the1.5 GeV-HDSM will show if such higher field errors can be handled or not. In contrast to theserelatively large corrections, only small modifications (0.06 % and 0.22 % at 1.64 T and 1.71 T,respectively) are needed to adapt the central field profiles to the ideal shape.

Figure C.12: Field deviation in steps of 0.2 mT with respect to the central field profile for two highermagnet excitations.


Antisymmetric Field ComponentsThe knowledge of the distribution of the vertical field components on either side of the mid-plane in principle allows for the calculation of the complete field distribution consisting of asymmetric and antisymmetric part [8]. A problem, however, arises from the fact that in thefringe field area, off-midplane components cannot be measured precisely enough in our ar-rangement to extract the antisymmetric fields with the desired resolution: the vertical Hallprobe position would have to be precise to better than 0.02 mm in this area to reach the re-quired precision of 0.2 mT for the transverse (antisymmetric) field components Bx and Bz. Inorder to get a solution for the inner region of the gap, the calculation was performed under theassumption that the antisymmetric field was zero in the fringe field region: From each rasterpoint of the boundary line of this area an ideal fringe field decay was added numerically. Theresult for the antisymmetric field distribution is shown in Fig. C.13 for dipole no.2 as an ex-ample. The undulations along the pole edge are produced by a 0.02 mm oscillatory instabilityof the milling head during the machining of two sub-faces at the upper pole of the magnet.In general the transverse components are below 1mT leading to vertical deflections as shownin Fig. C.14. A rough estimation of the influence of such fields to the beam optics leads to acoupling of a few percent between the horizontal and vertical phase spaces.

Figure C.13: Lines of constant fields Bx and Bz in steps of 0.2 mT and 0.05 mT resp. in the miplane ofdipole no.2.

Field CorrectionsFor the correct functioning of the HDSM the beam has to be guided very closely to its idealorbit. A path length error of 0.17 mm e. g. would result in a phase deviation of 1 at the 4.9 GHz-wave, and the displacement in transverse direction is strongly confined by the small apertures of13 mm in the return path tubes and 10 mm in the 4.90 GHz linac pipe, respectively. Therefore,


-0.4

-0.35

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

800 900 1000 1100 1200 1300 1400 1500 1600

Def

lect

ion

angl

e [m

rad]

Energy [MeV]

Figure C.14: Vertical deflection by transverse fields in dipole no.2.

in a first step the relative field errors in the bending magnets must be reduced to about 1 ·10−4.As the experience with the microtron cascade MAMI B has shown, additional corrections arenecessary to centre the beam on the linac axis in each revolution. While correction dipoles onboth ends of the return paths turned out to be sufficient for the RTM’s, additional measures(see C.2.2.7, C.2.2.8) have to be taken for the HDSM because of the decreasing length of itsreturn straights, and due to the fact that the beam passes across relatively poor fringe fields atthe magnet corners.

Surface CoilsThe fact that the complete field distribution in the inner part of the magnets was calculated fromthe measurements makes it possible to extend the correction to both symmetric and antisym-metric field errors. In order to test the method, a pair of surface coils was built covering aboutone sqare meter in the undulated field region of dipole no.2 (see Fig. C.15). Field mapping afterinstallation showed that the procedure works with the expected precision of about 1 ·10−4 (seeFig. C.16).

Figure C.15: Test coils for asymmetric field correction in dipole no.2 (left: upper coil, right lower coil),coil current 15 A.

In spite of this success it was decided to do the complete correction only for the symmetric fielderrors. As could be shown with particle tracing calculations, deflection errors by antisymmetricfield components are in the order of a few tenth of a mrad and the phase space coupling result-ing from their spatial variations are small (see above). Beyond this, the number and shape ofconductors of the combined correction coil set would be dominated by the antisymmetric part.This is due to the fact that the generation of a compensating transverse field in a certain area


-3

-2

-1

0

1

2

3

4

5

6

-3.4 -3.2 -3 -2.8 -2.6 -2.4 -2.2 -2

Ba y

[G]

z [m]

14960

14970

14980

14990

15000

15010

15020

-3.4 -3.2 -3 -2.8 -2.6 -2.4 -2.2 -2

Bs y

[G]

z [m]

Figure C.16: Antisymmetric Bay (left) and symmetric Bs

y (right) vertical field component along the lineA-B (Fig. C.15) with correction (lower curves) and without corrections (upper curves), both measured25 mm above the midplane

requires a continuous current density at the upper and lower pole faces, whereas the productionof a vertical field only needs a current loop around this region. For demonstration the 10 Acorrection current line distribution is shown in Fig. C.17 for the upper and lower asymmet-ric correction plates of dipole no.2, in comparison with the correction coil for the symmetriccorrection. In view of the rather marginal effect of the transverse components to the beam itseemed not to be worthwhile to take the risk of much more complicated coil sets.

The current distribution for the correction of symmetrical field errors was calculated fromthe difference between the measured field to a particular reference field using the formula(C.3) given in [6]. Since there is some freedom for the field configuration among the two 180

bending systems, the field profiles in the centre of the uncorrected dipoles no.2 and no.3 havebeen chosen as reference fields to simplify the shape of the correction coils. In Figure C.17the correction plates are shown for the magnets no.1 and no.2 forming the first 180 bendingsystem. Because of the high line density near the boundaries, the coils are divided into twoparts, an inner loop consisting of 20 windings for 10 A, and an outer loop of 11 windings witha correction current of 20 A. The first pair of coils, fabricated from 3 mm aluminium plates bywater jet cutting, was delivered in March 2004.

C.2.2.6 Correction Magnets

For an efficient orbit correction the construction length of the correction dipoles must be short,and the distance between the so-called first and second steerers at the return straights has to beas large as possible. As can be seen in Fig. C.18 they are designed as small H-magnets witha pole width to gap ratio of 50 mm / 15 mm for good homogeneity. As a consequence of therelatively wide pole, the horizontal steerers need all the space available on either side of itstube. Therefore, the neighbouring magnets have to be displaced along the beam tubes, forminga second row of horizontal steerers (see Fig. C.19). In the vertical direction the correctionsare expected to be smaller, so that the vertical correction dipoles can be installed towards theinner side of the return tubes. The angular corrections of the beam are expected to be notlarger than about 2 mrad in the horizontal and 1 mrad in the vertical direction. At 1.5 GeV thecorresponding electrical power is 17 Watt (11 V, 1.5 A) and 6 W (6 V, 0.9 A), respectively.


Figure C.17: Current lines for asymmetric field corrections and surface correction coils for symmetricfield errors.


Figure C.18: Small dipoles for horizontal and vertical beam angle corrections on the return path.

Figure C.19: Arrangement of horizontal and vertical correction magnets on the return paths.


C.2.2.7 Iron Shims in the Fringe Field

In case of higher field deviations that cannot be compensated at the same location, two correc-tion elements are needed to correct both, the angle and the displacement of the beam. In ourcase the correcting magnets, at least those at the linac axes and at the higher revolutions, willbe completed by shims attached at the front face of the HDSM-dipoles and reaching more orless deeply into the free space in front of the Rogowski profile (see Fig. C.20). Since the shimsare almost completely saturated (µr ≈ 1) in the area where they do not touch the front face,they simply add their magnetic flux to the fringe field. In addition, the fringe field is somewhatincreased in the outer region by the iron added to the front face of the dipole (Fig. C.21). InFigure C.22 the field distribution and the electron path are shown for the complete correction.

Figure C.20: Vertical iron shims at the magnet front face and technical realizarion.

C.2.2.8 Pole Coils

As the beam guiding in the HDSM is expected to be much more difficult than it is for the RTM,so called pole coils will be installed in the gap along the crests of the orbits (see Fig. C.23). Achange of the beam angle at this location merely produces a displacement at the magnet exitin first order. In this way, the pole coils act complementary to the correction dipoles. Becauseof the relatively small distance between the orbits, the 150 mm wide coils act to several orbitsat the same time. On the other hand, due to the relatively large magnet gap no short-rangingeffects are expected.


0

50

100

150

200

250

300

-300 -200 -100 0 100 200

Mag

netic

fiel

d B

[G]

Position normal to pole edge x [mm]

Magnetic field produced by 3mm shim

Figure C.21: Influence of iron shims for field corrections.

-200-150-100-50

0 50

100 150 200 250 300

-0.9 -0.6 -0.3 0 0.3

By

[G]

s [m]

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

-0.9 -0.6 -0.3 0 0.3

x’ [m

rad]

s [m]

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

-0.9 -0.6 -0.3 0 0.3

x [m

m]

s [m]

Figure C.22: Orbit correction in the fringe field by means of shims and correction magnet. From left toright: magnet field, beam angle x’ and beam displacement x at 855 MeV along the linac axis counted s inm from its intersection with the dipole no.2 front face (dashed lines without, solid lines with correction).

Figure C.23: Scheme of pole coil arrangement.


C.2.3 Injection, Extraction, Transport System

C.2.3.1 Injection System

The beam delivered from the RTM3 will be transported to the HDSM-hall by means of theachromatic bending system built by the X1-group [9], the former users of the HDSM halls.After passing through a 2.5 m long bore in the yoke of HDSM-dipole 3, the beam will be shifteddown by 400 mm to the HDSM midplane by two vertical deflecting magnets at a distance ofabout 4 m (see Fig. C.1 for an overview). The subsequent horizontal bending system will thenguide it to the injection point in front of HDSM-dipole 1. Two small C-magnets, one locatedclose to the linac and the other one at the beginning of the first return path, serve for theinflection to the first HDSM orbit.

The X1-bending system as well as the 20.2 horizontal deflection system in the HDSM-hall aredesigned in such a way that they act as simple drift spaces of similar length in both horizontaland vertical directions. This will make it easy for the operator to match the transverse beamellipses to the HDSM acceptance. The dispersion of the vertical deflection system will not becompensated. Thanks to the small energy spread of the beam its influence on the particle po-sition will be below 0.1 mm. In Figure C.24 the beam envelopes and the dispersion trajectoriesof the injection system are depicted.

Figure C.24: Beam envelopes (left) and dispersion trajectories (right) in the HDSM injection system.The beam sizes correspond to 25 times the normal phase space areas.

As described in some detail in [6], the matching in the longitudinal phase space will be achievedby suitable rf-field gradients in a matching section, placed between the vertical deflection mag-nets, and in the 4.90 GHz linac, where the bunches are shifted to an appropriate phase on theaccelerating wave. Together with the longitudinal dispersion produced by the HDSM-dipolesno.1 & 2, this matching system consists of two variable longitudinal lenses separated by a fixedlongitudinal drift space. Optimum matching to the HDSM-acceptance is achieved, if the gra-dients of the two lenses are 215 MV / m and – 240 MV / m, respectively. While the former willbe realised by an amplitude of 2.09 MeV in the 4.90 GHz-matching section operated at zeroenergy gain, the latter will be reached at a bunch position of + 15 from the crest of the rf wavein linac 1. For this purpose the bunches will be injected somewhat earlier into the HDSM asthey would without matching. After the passage through the linac the bunch timing will becorrected by a path lengthening chicane on the first orbit between dipoles no.3 & 4.


C.2.3.2 Extraction System

As described also in [6], the extraction will be realised by a small beam deflection on theselected return path in front of HDSM-dipole 2 (see Fig. C.25), leading to a displacement ofabout 120 mm with respect to the linac axis at the dipole exit. At this position a second de-flection magnet, shown in Fig. C.26, will bend the beam away from its direction towards thelinac and feed it into the extraction beam line. There, the beam is raised by 400 mm by a pairof vertical bending dipoles and then deflected horizontally into a bore in the concrete wall tothe beam line tunnel.

In order to simplify the extraction at full energy, a dedicated extraction magnet will be installedat the highest orbit. A second dipole, constructed in such a way that it can be placed over thereturn tubes, will be used to extract any of the lower energies. Its iron core consists of fivesegments sepatated by 2 mm aluminium plates in order to improve the field homogeneity in thegap.

Figure C.25: Beam deflection for extraction in front of dipole no.2 at turn 43 (left) and at one of thelower recirculations (right).

Figure C.26: Beam deflection by the second extraction dipole.

Beam optical calculations of the extraction line turned out to be somewhat difficult, becauseits geometry was nearly completely determined by building and space constraints. Therefore,


it was not possible to install classical achromatic deflection systems in which transverse opticsand dispersion can be tuned separately to a certain extent. A solution with achromatic beamtransport to the beam line in the tunnel could only be found when a weak quadrupole was intro-duced in the dispersion region in front of the deflection magnet (not shown in Fig. C.25). Thebeam envelopes and the dispersion trajectories of the extraction system are shown in Fig. C.27.From an optical point of view the system corresponds to drift lengths of about 7 m in the hori-zontal and 12 m in the vertical direction.

Figure C.27: Beam envelopes (left) and dispersion trajectories (right) in the HDSM extraction system.The beam sizes correspond to 25 times the normal phase space areas.

C.2.3.3 Components

For the construction of the injection and extraction lines existing magnets and power supplieswill be used as far as possible. For example, the iron core of the 20.2-magnets of the injectionas well as the 3.5 vertical deflecting dipoles in the extraction line will be made by cuttingone big 82-magnet (part of the former energy loss system, used until 1987) into suitablepieces. New elements are the coils for these dipoles, the five direct cooled small C-magnetsas well as four chicane-dipoles, 24 correction magnets and 26 quadrupoles, all with uncooledsolid wire coils supplied by the in-house built multi-channel power supplies (see C.2.5.2). Allcomponents, including the required larger power supplies, have been ordered in 2003.

C.2.3.4 Beam Transport System to the Experiments

The beam line behind RTM3 had to be upgraded in order to cope with the higher electronenergy, to be delivered from the HDSM, and to connect the HDSM injection and extractionsystems. As described in more detail in [6], the first part – the so called TF-system – had to bemodified completely in order to obtain a separated beam line feeding the new experimental areaof the X1-collaboration, to inject the beam into the HDSM-halls, to further use the MAMI Bbeam bypassing the HDSM and to inflect the 1.5 GeV beam. The space used for the inflection ofthe 1.5 GeV beam made it necessary to replace the former 3-switching magnet at the end of theTF-system by a stronger one and to shift it along the beam by 5.9 m. As a consequence, the A1-system and, to a smaller extent the A3-system too, had to be realigned. The four main dipoles ofthe A1- and A3-deflection systems were upgraded to a mean bending field of 2.2 T by installing


suited pole plates and replacing the old power supplies by stronger ones. Field mapping of thefirst A3-dipole indicated that the measure was successful [10]. Fortunately, for the transportof the higher energy beam through the A2-bending system it is only necessary to replace thedipole power supplies. It still remains to replace or modify (by an aperture reduction from 40to 28 mm) some quadrupoles, which at present do not reach the field gradients necessary tofocus the 1.5 GeV beam.

C.2.4 HDSM Vacuum System

The design of the HDSM vacuum system has two major goals. First, it must provide a basepressure during operation that is low enough so that scattering effects on the residual gas haveno deteriorating effects on beam properties. Second, it should be optimised for maintenance-free operation. The technical layout of the system is therefore similar to that of RTM3 wherethe desired situation has been achieved.

The vacuum system is entirely metal sealed, with the exception of the seals of the shut offvalves. It consists of four major subsystems, the two 180 bending systems and the 2.45 /4.90 GHz linacs. These systems may be shut off from each other by valves at the ends ofeach linac. The usage of aluminium for the majority of the vacuum components prevents anefficient bake-out. However, since the necessary average pressure to avoid scattering effects israther high (< 5 ·10−7 mbar), it is possible to achieve this by a sufficiently distributed pumpingsystem. It consists of ion getter pumps of the diode type. The total number of about 50 pumpsis distributed in a way that the effects of the relatively high outgassing rate and of low gasconductance are minimized.

We therefore do expect at least the same or a slightly better average pressure than in the reliableRTM3 stage, where it is about 10−7 mbar. The capacity of the pumps is sufficient to providestable operation (estimated 20 years) over the run time of the accelerator. Evacuation of eachsubsystem is achieved with individual turbo pumps. These mechanical pumps are shut offby valves during operation in order to avoid electrical and/or mechanical noise on the beam.Detailed studies have been performed which allowed to specify the expected scattering rates,pressure distributions, maximum pressures, pump down times and the effect of synchrotronradiation [11], [12].

All elements of the pumping system have been purchased or are at least ordered.

Other work during the reporting period has been focussed on detail solutions for the subsystemsof the HDSM. A huge variety of vacuum components have to be designed. Presently, about onethird of the components are available. Most of the remaining parts have been ordered or are inthe final stages of the ordering process.

In the following the subsystems are discussed.

Linacs:The linac-vacuum system is dominated by the requirements of the rf-structures, which havelow gas conductance. Consequently a highly distributed pumping system is needed in order toavoid local pressure bumps. Seven ion pumps per linac are installed as close as it was possibleto the ends of the rf-sections. The total pumping speed amounts to 1000 l / s per linac.


The pumps are integrated into a bypass tube along the linac. This offers a more compact ar-rangement and has been made possible by the “double ended” type of ion-pump offered byindustry for several years. The conductance from the pumps to the linacs is optimized by shortpumping lines which run from the high-conductance bypass tube to the end of each rf-section.

The fraction of components with UHV-compatible conflat flanges has been increased withrespect to the RTM3. The reason for this is the lack of man power to produce special parts inhouse and the considerable price reduction of these parts in recent time.

Each rf-section is equipped with another small ion pump which is installed directly at the rf-coupling window. In case of a bad section-vacuum, e.g. by arcing or multipacting, the currentsignal of this ion pump serves as a sufficiently fast rf-power interlock to protect the sensitiveceramic of the rf-window.

Evacuation of the 2.45 GHz linac is done by a single turbo pump which is installed in thecenter of the bypass line running in parallel with the linac. For the 4.90 GHz linac, locatedat the entrance side of the HDSM-hall (cf. Fig. C.6) it was necessary to interrupt the bypassline twice, in order to get space below the linac to enable an easy access into the hall. For thisreason evacuation is achieved with three smaller turbo pumps.

Bending Systems:At present, the magnet vacuum chambers are in the final stage of manufacturing. They arefabricated from aluminium, using the experience made during the construction of the RTM3vacuum system. As can be seen at the drawing of the standard chamber for dipole no.3 inFig. C.28, long flat rods are inserted to support the 20 mm top and bottom plates against theatmospheric pressure. These spacer rods are pierced to enable the passage of the beam with aminimum free space of 7 mm on each side. (In the special chamber for dipole no.2 the posi-tion of the spacers is defined by the intersection between recirculation and extraction orbits,Fig. C.29). The two long re-entrant tubes on the right will be used to insert NMR-probes forpermanent field measurements during operation. A mirror inside the chamber in connectionwith a glass window on the right side makes it possible to map the beam spot by displaying thesynchrotron radiation emitted by the electrons when they enter the magnet fringe field. In thechambers for dipole no.2 and no.4, where they are running towards the common exit, the samewindow serves for the direct observation of the beam on its different orbits.

Figure C.28: Vacuum chamber of dipole no.3.


After welding the chamber for dipole 3 in September 2003 its geometry was remeasured bymeans of the numerically controlled theodolite system AXYZ from LEICA. Whereas the flat-ness under vacuum turned out to be very satisfying – the maximum deviation from the idealhorizontal plane was found to be not larger than 1 mm – much larger deviations became appar-ent in the transverse direction: With respect to a line defined by both ends of the multi flangebar (connection of the return tubes), the entrance flange of the common entrance hole on theright side was displaced by about 5 mm in both longitudinal and transverse directions as in-dicated by the two short arrows in Fig. C.28. Fortunately, the flange bar itself only showed acontraction of 0.6 mm and a deviation of 2 mm from its straightness. As the flange bar definesthe reference line, as mentioned above, its bending explains one half of the transverse displace-ment of the input hole. Under the assumption that the displacements of the spacer bars are inthe order of the flange bar deformation, the transmission of the beam will be possible withouthitting an obstacle. (The aperture for beam entrance is large enough to account for the 5 mmdisplacement.)

On the basis of this experience and for safety, two improvements have been introduced for themanufacturing of the remaining three chambers: First, a dense pattern of check marks on theoutside of the chamber plates will be used to a complete the measurements of the deformationsafter welding, and second, the spacer bars will be fabricated and pierced only after welding onthe basis of the new geometry. They will be inserted through flange holes added at the chamberbackside before putting the chamber under vacuum (s. Fig. C.29).

Figure C.29: Beam orbits and vacuum chamber of dipole no.2 in the modified design, enabling theinsertion of the spacer rods after welding.

Each chamber is equipped with 3 large ion pumps with a total pumping speed of 900 l / s. Thisis sufficient to pump away any desorbed gasses in operation and results in a reduced downtimeafter a venting of the chamber. The reduction of this pumping speed at the beam trajectorydeep inside the magnet is considerable, because of the low aperture of the vacuum chamber.However, estimations have shown that the base pressure will stay below 1 · 10−6 mbar evenwhen synchrotron radiation induced desorbtion contributes. The conditions should even beslightly better than in RTM3 because of the more appropriate arrangement of the spacers.


C.2.5 Control and Diagnostics

C.2.5.1 Control System

Central computer:The central computer of the MAMI control-system, which besides many other services runsthe online database of the control system, was formerly a DEC-Alpha under Open-VMS. Thisout-dated platform has now been exchanged by an Intel-CPU with a Windows 2000 Serveroperating system. For this platform a Fortran-Compiler is available (DEC-Fortran), which en-abled us to port a part of the old control-system software without much work. Another part hasbeen rewritten in C-language. The goal is still to translate all these old Fortran programs, butthis work can now be done step by step. On the modern platform there is no bottleneck, neitherin CPU-power nor in working space or network bandwidth, which may compromise the exten-sions of the control-system required for the HDSM control. The performance of the databaseis high enough even for rapidly repeating data measuring requests in high level programs.

Front ends:In the present configuration of the MAMI control-system VME is a typical bus to connectto miscellaneous hardware devices. Therefore we decided to use VME also for control ofthe HDSM. For other experimental facilities in the institute (A4 and A1) fortunately a cheapVME-Bus CPU has been developed in-house, by mounting a single mini-PC card on a VME-board and connecting its PCI-Bus via a CPLD (Complex Programmable Logic Device) to theVME-Bus. For use in the MAMI control-system some simple extensions to this CPU werecarried out to adapt to the given requirements. They are now also appropriate as a displace-ment of the out-dated SPARC-CPUs used as front-ends in the present system. The low priceof 2000e compared to commercially available PC-based VME-bus CPUs (that are more than10 times more expensive) enables us to build a sufficient number of backup systems to ensureagainst failures and to expect a long lifetime cycle of this development.

Software:

• The software driver for the HDSM-magnet power supplies is available. A routine to controlthe whole magnets together with all their components has not yet been implemented dueto missing specifications for the details of the power-on procedure.

• A new VME module to drive the HDSM rf-phase shifters has been designed and built inthe electronic workshop of the institute. A new software driver for this device was im-plemented, which is also able to control the existing MAMI B shifters based on the for-mer CAMAC-hardware. This implementation replaces the out-dated software in MAMI B(Fortran-program STEP) by the more general program written in C and will be also usedfor the phase shifters of the HDSM.

• The program CONTOUR, generating contour lines from a set of magnetic field data pointsof a dipole, has been ported to Windows using the DEC- Fortran compiler. This programwas originally developed during the build-up periods of the RTM2- and RTM3-dipoles. Itsflexibility makes it a very useful tool for the construction of the correction coils of magnets.Numerous additional programs have been developed to pre-process the magnet data [13].

• A version of the tool “Gipsy” for Microsoft Windows was bought for the modifiable syn-optic display of the status of the machine components and subsystems. Work to enable the


use of this tool in the control system is in progress. Gipsy was chosen, because it is com-patible to the old version (running under DEC-Unix), which is used in the present system[14].

• A new type of adapter to connect to devices with a RS232 or RS244 interface has been in-troduced. In the present system a multiplexer is used, combining up to 16 RS232-channelsto one RS232-channel, while the new adapter communicates via LAN instead. A specialprogram under construction builds a software interface for RS232 ports, so that there is noneed for changes in the next higher software level.

• A new type of a commercially available fast waveform digitiser for the digitisation of therf-beam-monitor signals in the microtrons has been chosen and tested. Integration into thecontrol system together with a new multiplexing hardware, built in-house, is in progress.This set-up will enable also the digitisation of the HDSM signals.

C.2.5.2 Electronic components

Three fluxgate beam current monitors with an aperture of 15 mm were constructed: one for usewith the HDSM, a second one for the beam transport system to the A1-hall and a third one forthe beam line to the A4 experimental areas. Each monitor was tested for several weeks. Theyare ready to be installed since fall 2003.

For use with MAMI C ten additional VME output modules were completed. The production ofthe same number of input modules and scanning ADCs will start after a redesign in 2004.

The central beam diagnostic system of the HDSM and the RTMs is based on “low Q” rf-monitors for the non-invasive acquisition of beam phase, intensity and position on the accel-erator axes. During set-up and optimisation of the accelerator the electron gun is operated togenerate intense 10 ns long pulses with a repetition frequency of up to 10 kHz. This makes itpossible to resolve each turn separately by the time of flight of these diagnostic beam pulses.Whereas in the RTMs rf-monitors are only installed at both ends of their linacs, a third mon-itor is added in the middle of each HDSM-linac to get additional information about the beamposition along the quite long and narrow linac pipes. Because the number of pulse signals tobe displayed per cycle is increased by the HDSM from 954 to 1814, and because the existingsystem uses obsolete electronic components, a new fast multiplexer had to be built to displaythem in a very flexible way on three oscilloscopes. A pattern of eight signal combinations canbe stored in the multiplexing unit. The new system contains a programmable pulse generatorto produce the trigger for the gun and to multiplex the beam signals of all four stages. So theimpulse trains for phases, intensities and positions can be digitised by up to six fast wave-form digitisers, enabling the simultaneous acquisition of beam data. Since it would be nearlyimpossible to run the accelerators without this diagnostic system, a spare unit was completedtoo.

At the beginning of the year 2003 the prototype of the 128-channel power supply (8 boardswith 16 Channels) was tested, which is needed to energise the numerous correction magnetsand other low power components. Each of the bipolar channels delivers up to 7 A with a long-term stability of about 2 · 10−4 and its maximum output voltage is 20 Volts. Both parameterscan be combined freely under the condition that internal losses are below 73 W per channel.The inductance of a load should not exceed 0.2 Henry, and the sum of inductances at one 128


unit must be below 15 Henry. The line noise (50 Hz and 300 Hz) of the output current is lessthan 2 mA peak.The control board that performs the ON / OFF-procedure and the interlock functions containstwo dual slope scanning ADC’s. They are optimised for a short measurement time by a rea-sonable noise rejection with a 20 ms integration time. The measured values for current andvoltage of each channel are stored in a dual port memory and can be read at any time. Aftertest and offset corrections an absolute precision of 0.1 % for the read back of all currents hasbeen achieved.After the successful test of the prototype the construction of four more units has been started inApril 2003. Due to a change in the design the HDSM correction magnets in the meantime, thespecification could be somewhat relaxed: Each unit contains 96 channels with 1.4 A, 16 chan-nels with 3.5 A and 16 channels with 7 A maximum current. The maximum voltage is 11 V. Allpower supplies will be ready for use in April 2004.

C.2.6 Electricity, Cooling Water and Air Conditioning

During this funding period and parallel to the installations of the accelerator components forthe HDSM of MAMI C most of the new infrastructure hardware was already installed and thenecessary upgrades and modifications of the existing hardware were finished. In detail thesewere mainly:

Electricity

Operating the HDSM requires nearly 1.7 MW of additional electricity power. To scope withthis requirement, it was necessary to add a new 2 MVA transformer station to the existing ones.The space for installation was available, because at the same time the technical departmentof the university exchanged the old medium voltage switching station by a modern one withreduced space requirements. The extra power is fed to the new machine by two low voltagepower distribution systems, which were installed in the technical hall aside the accelerator hall.One of these distribution systems is exclusively used for connecting the two BRUKER highvoltage power supplies for the rf-systems of the HDSM (nominal power: two times 810 kVA,under operation 460 kVA resp. 560 kVA). The other one will feed all magnet power suppliesand control- and interlock-systems. This arrangement will help to avoid back influences on thehighly stabilised magnet power supplies due to fast power load changes and high earth currentsduring e.g. crowbar firing in case of klystron arcs. From the two 100 kVA distribution systemsof the Experimental Halls 1 and 2, one had to be relocated inside the hall to make space for thenew radiation protection electronics bunkers, and will be sufficient for all electric installationsinside these HDSM halls. It will be ready installed after the final construction of the bunkers.The other one was transferred to the technical hall to add some extra margin to the powerhandling capabilities.

Cooling Water

Obviously the 1.7 MW of extra power for the operation of the HDSM need an appropriateincrease of the cooling water plant. The old cooling plant of the pulsed linac, which was inuse for the electron beam transfer system installations, had already been totally modernisedand put into operation. However, then a delay of some months occurred, because during theprotective gas welding of the stainless steel tubing and the subsequent etching process the im-plementing company did not work carefully enough. Therefore, shortly after filling the system


with de-ionized cooling water, strong corrosion at the welded joints started and water leakagesoccurred. As a consequence the biggest part of the stainless steel tubing had to be dismountedand carefully new built. – Naturally till now a full power test of the system was not possible.However, at power levels up to 300 kW, it shows an excellent performance. The cooling watertemperature is stabilised to 28C ± 0.1C (max. deviation); measurements of the resonancefrequency of an accelerating section at the high power rf-teststand show over periods of sixhours a temperature stability of better than ±0.05C. This 1.4 MW system is supported bya completely new 600 kW installation on the technical floor of the spectrometer hall, whichserves now for the beam transfer system, the tagger hall, the experimental halls 3 and 4, as wellas for the modulator hall. After some problems with the fine tuning of the regulation parame-ters during the start-up phase of this system, it now also operates with great reliability and astability of ±0.1C.

Air Conditioning

During the design phase of the klystron high voltage power supplies by BRUKER, a majorchange in their cooling strategy was necessary. It turned out that the use of oil cooled highpower transformers, as it was foreseen similar to the ABB 610 kVA high voltage supply ofMAMI B, is not longer allowed 10m below ground, the level of the HDSM technical hall.Therefore it was necessary to adapt an air cooled design, which results in approx. 75kW ofpower losses to air. To handle this power load the old air cooling system for the pulsed linacwas refurbished and re-set in operation, realising approx. 100kW of cooling capacity. – Theeight air conditioning systems, 4kW power each, for the HDSM halls (four for each hall)are already installed and are waiting for operation. From the experiences with MAMI B it isknown, that an stabilised ambient temperature increases the accelerator stability and reducestime for machine tuning. If necessary, for each of the radiation protection electronics bunkers,a compact 2 kW system can be installed under the ceiling.

C.3 Polarised Electron Source

C.3.1 Operational Stability for Experiments at MAMI

The MAMI polarised source which is based on photoemission from thin strained layers ofGa As P achieves state of the art values of quantum efficiency (q.e. ≈ 0.5%) and spin-polarisa-tion (P = 75− 80%). The inherent emittance of such a photo source is at least equivalent, ifnot superior to a thermionic source; consequently excellent beam parameters can be achieved.

Given these initial parameters the main goal for polarised source operation is to provide stableoperating conditions for the experiments. During operation with a photo effect source thisstability is threatened by the continuous decay of q.e. which at some stage does no longer allowto continue with the experiment. Usually the time interval τ during which a decay of quantumefficiency to 1/e ≈ 0.37 of its maximum value takes place is called a “lifetime”. Obviously theeffects have to be controlled in a way, that the decrease of q.e. during the experiment can becompensated by an increase of the laser intensity.

During the period covered by this report we have been able to improve the conditions in a waythat even the experiments with highest demands on beam current could usually be served con-tinuously over their runtime. This has been achieved by a special technique described in the


next section. In addition we have been able to identify the main causes for the remaining cath-ode degradation and have started a program to improve the parameters which are responsible(section C.3.1.3).

C.3.1.1 Transmission Losses

In our photo source, the ratio of the cathode surface to the area where photoemission takesplace is rather large. The reason for this is that mechanical reduction of the active surface withe.g. a diaphragm in front of the cathode will lead to edge focussing effects which are difficultto control. In this situation scattered light causes unwanted emission from areas outside themain emission area. Therefore, a small fraction of about 0.5− 1 · 10−3 of the beam currentis lost on the beam tube behind the anode of the source. This leads to the production of gasspecies which tend to poison the cathode. Since the yield of poisonous gases is proportional tothe beam current it has to be expected that the lifetime is inversely proportional to the beamcurrent. Therefore the extractable charge would be constant under given operating conditions:C = I ∗ τ.

Under this assumptions it’s evident that a reduction of transmission losses would lead to anincrease of τ, or equivalently of C.

The source group at CEBAF first succeeded in reducing the parasitic effects by electrolyticoxidation of the cathode surface outside a small diameter [15]. The areas covered with Oxideshow quantum efficiencies which are several orders of magnitude lower than the active area.We have demonstrated that this procedure reduces transmission losses in the relevant surfacesof the vacuum system in front of the photocathode by at least a factor of 100, and that theamount of charge that could be extracted during one decay constant of quantum efficiencyincreased considerably [16].

Because it is a wet-chemical process, anodization implies a certain problem of photocathodesurface contamination. Therefore we have developed a new method which is called “maskactivation”, illustrated in Fig. C.30: During activation with Caesium evaporating from a sourcethe photocathode is stacked into a mechanical mask. This leads to the same effect as for theanodization technique. Additional advantages are that there is no Caesium coverage outsidethe active area, especially on the photocathode holder (“puck”). Because the work function ofthese materials is also reduced by Caesium, there is a small but measurable photoemission (and,potentially, field emission) from these areas which is now avoided. No migration of Caesiumoutside the activated area was observed so far.

The mask activation does allow performing measurements of the stability of a cathode withmask activation and without it (“nude cathode”), the results are shown in Fig. C.31.

The measurements were done at an average current of 200 µA. The Figure illustrates that aneffect that is correlated with current production limits the extractable charge in the first case of“nude” activation. After this, quantum efficiency was restored by cleaning the surface by heattreatment and a subsequent adsorption of fresh Caesium and Oxygen. This time the processwas carried out in the mask-activation configuration. It’s evident that a large enhancement ofextractable charge takes place [17].

The measurement of the total extractable charge during one lifetime of the photocathode re-sulted in a value of 120 Coulomb. This result was achieved by extracting a constant current


Figure C.30: Mask activation (schematic).

Figure C.31: Comparison of the decay of q.e. of a cathode, first with complete surface of the cathodeactivated (“nude”), then with “mask-activation”.


of 200 µA from a bulk Ga As with high q.e. for an effective runtime of about one week. Thewhole experiment lasted for about three weeks, because of maintenance work on the accelera-tor during daytime.

Mask activation has been used exclusively at the MAMI source since spring 2002. One addi-tional advantage of this technique is that the beam halo caused by parasitic emission is absent.This leads to a smaller emittance and as a consequence it has become much easier to achievestable machine tuning.

Experiments with a highly polarised beam at MAMI are conducted with lower beam currentsthan the ones used to identify the beam current induced effects, typically 20−30µA are used.In these experiments the extractable charge was observed to be limited to 30− 40 Coulomb,which represents only a gain of a factor of two with respect to the situation before. The reasonsfor the difference to the high current experiment seem to be caused by the parallel action ofother effects.

C.3.1.2 Identification of Parallel Acting Effects

Experiments at MAMI during 2001-2004 with their runtime of several thousand hours haveprovided us with the opportunity to observe the behaviour of photo cathodes in several verydifferent operating conditions. This resulted in the observation of different lifetimes τobs. If weassume that all effects destroying the photocathode act in parallel we have:

1τobs

= ∑i

1τi

(C.2)

The different conditions now allow to identify several of the summands τ i. Table C.2 presentsan overview of observations during several types of experiments.

The relevant time constants for our source are:

1. Vacuum lifetime τV. This parameter was observed by measuring q.e. before and after afour week beam break. Since this period is still considerably shorter than the lifetime thereis a large uncertainty in the value. We extract τV = 1200+800

−200 hours.

2. Operational lifetime without current: Here we used the opportunity that real photon ex-periments require only a very small current, which leads to the production of a negligibleamount of charge. However two additional detoriating factors exist compared with the vac-uum lifetime: First, vacuum conditions could be worsened by backstreaming gas from theaccelerator with its three to four orders of magnitude higher gas density. Second, the sourcehigh voltage is in operation leading to a potential field emission. There is an observablereduction of lifetime in this mode of operation, the effect of the additional processes justmentioned is taken into account with a new time constant τFE. From the observed reductionof τobs we extract τFE = 3000hours.

3. Operational lifetime with current. If we take into account the first two time constants τVand τFE, the time constants τobs in the three remaining experiments in Table C.2 show anadditional current dependent effect. This effect reduces the lifetime inversely proportionalto the beam current τI = kI/I;kI ≈ 200 Coulomb.


Experiment mode of operation τobs/h (additional) τi

stand by no operation 1200+800−200 residual gas in source:

(valves closed) τobs = τVReal photon Operation, I = 0.05µA 850+200

−100 back streaming, field emission:exp. (GDH) (valves open, 100 kV) τ−1

obs = τ−1V + τ−1

FENeutron Form Operation, I = 12µA 720+100

−100 Desorption, back bombardment:Factor (GN

E ) τ−1obs = τ−1

V + τ−1FE + τ−1

IVirt. Compton Operation, I = 30µA 520+80

−80Scatt. (VCS)High current I = 200µA 160+40

−20

Table C.2: Cathode lifetime under different operating conditions.

The constant kI represents the charge that can be extracted in a high current experiment when allother time constants can be neglected. It is important to note that the introduction of mask ac-tivation has improved kI by about one order of magnitude to the presently about 200 Coulomb.The current at which 1/τI = 1/τV + 1/τFE has therefore been increased by the same factor,to 65 Microamperes. Therefore present day experiments at MAMI with their typical beam cur-rents between 10 and 30 microamperes are not limited any more by beam current inducedeffects, but rather by the basic vacuum and field emission conditions. Presently this limitationis not severe, since the runtime of such experiments is smaller than τobs, so that we have beenable to serve the experiments continuously.

C.3.1.3 Improvement of Vacuum Conditions

The vacuum system of the source with its hydrogen pumping speed of approximately 800 l / scan be improved in two ways: by an increase of the pumping speed and/or by reduction of thesurface outgassing.

Both possibilities are realised if the inner surfaces of the vacuum system are coated by aTitanium / Vanadium / Zirconium alloy, the so-called Non Everaporable Getter (NEG). Afteractivation during the vacuum-chamber bakeout this coating adsorbs incident gases, thus trans-forming the chamber wall from a gas source into a pump. Measurements at CERN have re-vealed that it is possible to reduce the base pressure and the electron stimulated desorption ofa conventional UHV system by at least one order of magnitude [18].

The technology needed to coat the surfaces of the MAMI source and of its injection system ispresently being developed within the framework of a PhD thesis. The Ti Zr V NEG films weredeposited onto the interior wall of beam tubes, using a DC magnetron sputtering technique. TheNEG film composition was measured using the Electron beam scanning microscope technique(ESMA). The ESMA-results showed a NEG film composition of Ti1.3 V1.95 Zr. Sputtering for8 hours results in a sufficiently large film thickness of 1µm with a thickness uniformity of 8 %.

The pumping effect of these beam tubes has already been demonstrated in our test stand [19].Presently we are commissioning a coating stage for the main parts of the source vacuum cham-ber.


The effect of these improvements will be tested in a source which is independent from theaccelerator, in order to evaluate the potential towards an increase of cathode lifetime.

C.3.1.4 Lasers and Other Techniques Associated to Beam Production

As a consequence of economically based decisions of semiconductor-laser companies itbecame impossible to purchase new amplifier components for the master-oscillator / poweramplifier system (MOPA) described in the last report. Consequently we had to stop operationof the power amplifier after the last component failed in spring 2002. A replacement of theMOPA by a powered-up semiconductor oscillator had been investigated the year before [20].It had been found that sufficient output power for present day experiments at MAMI can begenerated if the rf-power on the oscillator is applied with higher power levels and improvedcoupling to the laser diode. Since then, beam production is done with the master oscillatordiode alone. This diode can produce output pulses which are synchronized to the MAMI-rfwith a FWHM pulse length of 40 ps and average power levels of 50 mW. In this operationtransmission from the source to the experiments is better than 95 %. This allows even to re-duce the phase acceptance of the chopper / buncher system, thus leading to even better beamquality and accelerator stability. Though the available power has been reduced, this is partiallycompensated by an increase in the available quantum efficiency of photo cathodes. The lasersystem is now extremely small and simple, it needs nearly no maintenance and offers verygood power-stability. RMS-stabilities of the laser – measured at a sampling rate of 1 kHz over5 minutes – are less than 10−3 [21].

It was found that the limited thermal conductivity of the interface between photocathode andelectrode structure imposes a problem for the application of higher laser powers [22]. Theincrease of temperature was measured with luminescence spectroscopy, the differential tem-perature increase was measured to be 0.3 K / Milliwatt, which limits the useful range of laserpowers to less than Pkath = 100 mW.

For experiments at very high currents which require high beam current stability, it is desirableto increase the output power of the laser system to about 300 mW. For active beam intensity sta-bilisation it is necessary to modulate the laser-beam intensity in order to stabilise the electron-beam current at the target. This is done by a transmission regulation of the laser-optical systemwith the help of an electro-optical modulator. The need for ample output power is explained bythe fact, that the point of operation of such a regulator should be at a transmission ≤ 0.5 and bythe insertion losses of the optical system. Since end of 2002 a new supplier of laser amplifiersexists, tests have shown that it is possible to achieve the necessary output power [21].

C.3.2 Analysis of Helicity Correlated Fluctuations for the A4 Experiment

In order to investigate the influence of the helicity switching process on the beam parametersan extensive study of the optical system of the source was carried out [23]. The main resultsare the following:


C.3.2.1 Current Asymmetry.

This beam current asymmetry is defined for the two helicity states (+,−) as: AI = (I+ −

I−)/(I+ + I−). The main contribution stems from the optical anisotropy of the photocathode[24] with respect to the direction of linear polarisation components. This anisotropy is due tothe inhomogeneous strain relaxation (ISR) in the photocathode. This effect may be compen-sated by rotation of a half wave plate behind the electro optical helicty switch, the so-calledPockels-cell.

An analysis with the aid of polarisation transfer matrices yielded quantitative results: The re-sulting intensity fluctuation may be parameterized by the optical imperfections in the followingway

AI(β) = AISR ((φA + ε)sin(2θk −4β)−φCsin(2θk −2β)+σC) . (C.3)

Where AISR is the anisotropy (analyzing power) of the photocathode, φA is the asymmetry ofthe optical phase shift when changing from positive to negative helicity, ε is a misalignment ofthe nominal angle between Pockels-cell axis and input polarisation, θk is the orientation of theanisotropy axis of the cathode with respect to the Pockels-cell-axis, and β is the variable angleof the compensator-half-wave plate with respect to the same axis. φC is the difference of theactual retardation of the “compensator”-half wave plate from its nominal value π. C representsthe average effect of phase shifts caused by residual birefringences of optical elements behindthe Pockels-cell and σ ≈ 1 is the circular polarisation behind the cell.

Figure C.32 shows the curve AI(β) measured at the A4-experiment. It can be fitted with afunction given in equation C.3. The parameter AISR = 0.035 is fixed by an independent mea-surement. In this situation the following values of the optical imperfections are extracted:(φA + ε) = 4,φC = 0.6,C = 1.1.

Figure C.32: Helicity correlated beam-intensity asymmetry measured in parts per million as a functionof compensator rotation in radian. Line is a fit with a function of the type given in equation C.3.

The non-zero term C implies a birefringence (phase shift) of the optical system behind thePockels-cell (telescope and vacuum window). Since it is not possible to investigate for suchan effect without breaking the vacuum-system of the source, we measured birefringences ofsimilar components in a dedicated test set-up. A birefringence of the correct order of magnitudewas indeed observed for a vacuum window.


Since it is necessary to achieve a zero-crossing of the AI(β)−curve, the term φA was delib-erately increased by choosing Pockels-cell voltages with an asymmetry (U + −U−)/(U+ +U−) = 0.034, leading to an asymmetric phase shift φA of 3.1, which is the major part ofthe observed value for φA + ε. The choice of non-optimal Pockels-cell voltages does reducethe beam polarisation by less than 0.3%. The parameter φC is within the specifications of themanufacturer of the wave-plate.

It may be noted that the value of the absolute phase shift of the Pockels-cell (90) does not enterinto equation C.3, so that the strong temperature dependence of the Pockels-effect does affectthe asymmetries only via its influence on φA. The effect of such a temperature change wouldresult in a change of 50 ppm / K if the AI-measurement is performed on the zero-crossing of thecurve in Fig. C.32.

Together with the further reduction of the amplitude and the temperature sensitivity of theAI(β)-curve by the beam current stabilisation (a suppression by a factor of 5-10), it is possibleto achieve stable conditions with average values for AI of a few ppm. The remaining AI canbe measured very accurately, thus the correction of the experimentally observed scatteringasymmetry can be done with only a small systematic uncertainty of 40 ppb [25].

C.3.2.2 Other Helicity Correlated (hc) Fluctuations

These include hc-fluctuations of e.g. the beam position. Such a fluctuation has been observedby a position-resolved observation of laser-intensity fluctuations. Very large fluctuations canbe observed in the interference patterns which form part of the halo of the laser beam. Locallythe laser intensity may vary by several percent. The halo is generated by a variety of processeslike scattering on dust, Raleigh scattering, “ghost” beam spots separated from the beam bydouble reflection on the optical surfaces. It has an asymmetric distribution around the beam-center. The helicity correlated intensity modulation may be caused by interference of scatteringbehind the Pockels-cell with light scattered before it: Since the light which passes the cellshows significant changes in its linear polarisation components, the conditions for interferencedepend on the helicity state. From the observed intensity fluctuations in a test-bench set-up wecan estimate the fluctuation of the beam position to be of the order of less than 10−3 of the beamdiameter, corresponding to an absolute movement of 300 nm at the photocathode. The positionfluctuations can be transferred by the electron optical system of MAMI into angular and energyfluctuations. These have been estimated to be of the order of < 100 nrad and < 36 eV [23], allvalues should become even smaller due to the effects of the stabilisation systems.

The contribution of helicity correlated effects to the error budget of the A4-Parity viola-tion experiment (PVA4) has therefore been successfully minimized. During the PVA4 run atQ2 = 0.23 GeV/c the contributions of the helicity correlated fluctuations of beam intensity,energy, position and angle (6 parameters) contribute only a relative error of less than 2 %[25]. Presently, this value is small compared to other error sources like limited statistics andinaccurate polarimetry.

C.3.3 Preparation of Spin Manipulation for MAMI C

Up to the year 2003 spin tuning at MAMI was exclusively achieved by energy variation ofthe third stage of MAMI (RTM 3), the so-called ∆E-tuning. It became attractive to study the


possibilities of a low energy spin rotation system because of three drawbacks of the ∆E-regime:

1. The ∆E-method does not allow to achieve transverse polarisation at low energies, a pre-requisite for a convenient characterization of photo cathodes with a Mott polarimeter.

2. ∆E-tuning is not directly applicable at low extraction energies from RTM 3, since theavailable spin-tuning range scales approximately with the square of the output energy.

3. Tuning with MAMI C would require a compensation of the energy variation of the thirdstage, since the subsequent acceleration in MAMI C with the opposite sense of beam de-flection would otherwise compensate the achieved spin angle tuning. This implies the in-stallation of a linear accelerating section in the interface between RTM 3 and MAMI C.Such an installation is possible, but it requires a large amount of resources.

Realising this inconveniences we decided to investigate the possibility of spin manipulation inthe injection system at a kinetic electron energy of 100 keV. We tried to use a Wien-filter thatis a copy of a design used for the PEGGY-II source at SLAC [26].

The Wien-filter was installed in the “Interface-0” (INT0) between the 100 keV photo sourceand the chopper system of MAMI. Figure C.33 shows the set-up.

Figure C.33: Installation of Wien-filter in the interface-0. E-field is in x-direction.

The Wien-filter consists of homogeneous and perpendicular electric (Ex) and magnetic fieldsBy, both transverse to the beam propagation (z). They are adjusted in a way that for the givenvelocity of the beam at a kinetic energy of 100 keV (β = v/c = 0.545) there is a force equi-librium between Lorentz force and electrostatic force. (E/B = v). A beam injected into thefilter remains therefore undeflected in first approximation. The spin precession frequency inthe Wien filter can be calculated to be ωs = eB/γ2mc, so that the spin precession angle is cal-culated as the product of frequency and the flight time in the filter to φs = eBL/(γ2 βmc2). The


direction of~ωs is given by the direction of the magnetic field which in our case is perpendicularto the accelerator plane.

For the filter length of L = 0.32 m a change of the magnetic field from 0 to 64 Gauss rotatesthe spin by 90. To maintain the force equilibrium an electric field of 1.08 MV/m is necessaryat the largest rotation angle. Full coverage of all possible spin angles in the accelerator planemay be achieved first by reversing the fields, thus covering the interval −π/2,π/2, second byhelicity switching of the light which transforms the spin direction by 180.

This concept of spin rotation was considered as difficult to handle because of the varyingfocal strength of the device: Off-axis particles in the (horizontal) accelerator plane (directionof electric field) experience a change of kinetic energy. In this case they do no longer fulfil thecondition for force equilibrium. The resulting force is directed towards the reference axis, andis in first approximation proportional to the electric field. Therefore the transfer matrix of aWien filter in the horizontal plane is dependent on the spin rotation angle. The focal strength isgiven by k = 1/ f = γφs sin(γφs)/L, which means in our case that the focal strength will varybetween 0 and 5.5 m−1 for the desired spin rotation range.

However, by an appropriate choice of the location of the filter the impact of variable focussingwas minimized: The beam is focussed on the position of the Wien filter even when the filteris not in use. Since the input emittance from the source is small, a tight focus of the beam isachieved at the filter position. Therefore the actual influence of the varying focussing force onthe beam emittance is rather small. Figure C.34 shows measured emittances behind the Wien-filter for different spin-rotation angles. The remaining rotation of the emittance ellipse can betolerated, since the beam emittance is much smaller than the machine acceptance. Initially aquadrupole duplet (Fig. C.33) was installed to cope with this emittance change, but it was foundthat they do not have to be used. This represents a considerable advantage for spin-tuningin routine operation. The behaviour of the system is in agreement with a three dimensionaltrajectory calculation [27].

x, [ mm ]

x’, [

mra

d ]

x, [ mm ]

-0.5 0 0.5 1

-1

0

1

2

-0.5 0 0.5 1

-1

0

1

2

90

0

30 6090

0

30

60

Figure C.34: Emittance variation behind the Wien filter for different spin rotation angles. Left:horizontalplane, right: vertical plane.

The system has been tested in several long term experiments at MAMI. Tuning from one spinangle to the other can be performed faster than with the ∆E-method. No major problems were


encountered even in operation for the demanding A4-experiment. It has therefore been decidedto use the Wien-filter as a standard spin tuning system for MAMI C.

C.3.4 Electron Beam Polarimetry

With the advent of the Wien filter system it has become possible to measure the beam-polarisa-tion in the injection system of MAMI. Such a measurement can serve to achieve three importantgoals:

1. Characterization of new photo cathodes. Photo cathodes must achieve simultaneously suf-ficient values of quantum efficiency and polarisation. Whereas quantum efficiency can bemeasured easily, polarisation measurements take a lot of effort when high energy polarime-try has to be used.

2. Monitoring polarisation drifts during the experiment. Since polarisation measurement dur-ing an experiment is often time consuming or not possible at all, a fast measurement withgood precision will facilitate the interpretation of experimental results.

3. Measurement of polarisation with high accuracy. Since the accuracy of polarimetry isproblematic, any new device with competitive accuracy will supply new input.

C.3.4.1 Precision of Polarisation Measurement

It was decided to cover the first two points by the installation of a new type of Mott polarimeterbehind the 3.5 MeV accelerator stage. Mott polarimeters detect the vertical scattering asymme-try A under the condition of transverse spin orientation in the horizontal plane. A is proportionalto the beam polarisation: A = S(θ,E)P. The analyzing power S(θ) (“Sherman”-function) hasits maximum at a scattering angle θ = 170 for the given kinetic energy E = 3.5 MeV. As anovelty with respect to other designs the upward or downward scattered electrons are detectedwith two double-focussing magnetic spectrometers. The idea of using the spectrometers is toreduce the background and to increase the energy resolution in comparison to conventionalscintillation counters.

The magnetic spectrometers were realised as 90 sector magnets, with an inhomogeneous mag-netic field in order to provide vertical focussing. The magnets are rather compact, their massbeing about 30 kilogram. A typical energy resolution on a thin target was ∆E/E = 0.02 inagreement with the design values.

For a 15 µm Gold target (“thick” target) and typical beam polarisations, asymmetries of about15 % are achieved with scattering rates in one spectrometer of about 2 kHz (primary currentI0 = 1µA). Therefore a statistical accuracy of 1 % is achieved in about one minute of measure-ment time at typical MAMI currents of several microamperes.

The reproducibility (precision) of the device is also determined by the stability of the back-ground and the beam-conditions. Tests have shown very small backgrounds (less than 2 %)on the thick target and the asymmetry seems to be stable for small variations of the beamparameters. We therefore hope for a precision of less than 1 %.


The Mott-polarimeter is intended as an offline device and is installed in a separate beam linebehind the injector. We expect that the time to install the beam onto the polarimeter and backtowards an experiment can be minimized to a few minutes.

C.3.4.2 Accuracy

The Mott polarimeter also promises a rather good absolute accuracy. A problem for absolutemeasurements with Mott-polarimeters is that it is usually done via the relation Aexp = S(θ,E)P,so that the Sherman function must be known. This knowledge results from solving the Diracequation for scattering of the electron on the nucleus and its surrounding electrons, whichcannot be done with infinite accuracy. It is has been stated, that the energy range between 3and 6 MeV is the least sensitive to this problem, since the influence of the atomic electronsis minimized in this energy range, whereas the finite size effect of the nucleus does not yetcontribute [28].

One of the experimental problems is the effect of double scattering in the target, especially for90-scattering followed by an additional angle that adds up to the detected scattering angle.This effect dilutes the measured asymmetry and is dependent on target thickness. For instance,the measured asymmetry in our polarimeter is doubled if the foil thickness is reduced to d =1µm. It is necessary to do a foil thickness extrapolation with the help of different targets toextrapolate the measured asymmetry A(d) to the value at target thickness zero, which wouldcorrespond to the “true” Sherman function. We hope to do this extrapolation with an error onA(0) of less than 1 %. The total error on the extracted beam polarisation would thus be in therange below 2 %.

In order to circumvent the almost principal limitations of Mott-polarimetry, we have tried to doan experiment following an idea of Gay et al. [29]. This technique is based on the inelastic scat-tering of transversely polarised electrons on noble gases. The observable in these experimentsis the polarisation of the fluorescent light after the excitation of the noble-gas atom, again ofthe form Pγ = SPe. No dynamical calculation is required to extract the parameter S in this case,in addition the underlying theoretical assumptions can be verified in the same experiment. Forthis reasons noble gas polarimeters can reach absolute accuracies of less than 1 %.

Our experiment was done in collaboration with the IPN-Orsay, who contributed the polarimeter– POLarimeter Optique, POLO – together with its optical detection system. It is the first timethat the usage of such a polarimeter at an injection-system of a high energy accelerator wasattempted. It was shown that the Figure of merit, which is defined as FOM = S2R/I0 (R =detected rate, I0 primary current), is about one order of magnitude higher than in the case ofthe Mott polarimeter discussed above. This is a clear advantage, as it would result in shortermeasurement times.

However, the experimental difficulties where considerable, especially since the experiment hasto be done in an energy range close to the threshold of the noble-gas transition, typically around20 eV. This generated several background sources, which – though identified – require a verycareful set-up of the experiment. This calls for considerable improvement before a sufficientreproducibility can be achieved. Therefore we decided to postpone this project in favour of the3.5 MeV Mott-polarimeter. A detailed report of the POLO experiment can be found in [30].


C.3.5 Photocathode Research

It’s worth noting that the efficiency of experiments at MAMI can still be increased by about20 % if polarisations in excess of 90 % could be achieved instead of the 80 % presently avail-able. Since the photoemission process from semiconductors may be divided into three parts –photoabsorption, diffusion to the surface, tunnelling into vacuum – each of these processes maybe responsible for depolarisation. Besides the beneficial aspect for MAMI-operations these in-vestigations represent fundamental research topics in semiconductor and surface-science.

The method we apply is ultra fast spin-resolved spectroscopy. The apparatus, the so-calledtest-source, is described in [31].

C.3.5.1 Pulse Response Studies

A systematic study of time resolved photoemission on different photocathode structuresyielded the following main results [32]:

a) The electron pulses produced by photo cathodes of different active layer thicknesses d be-have according to a diffusive behavior with an average sojourn time of the electrons that isgiven by < t >= d2/kD. D ≈ 40cm2/s is the diffusion constant of the electrons (minority car-riers) in the active region and k is a dimensionless constant which depends on the boundaryconditions on the interfaces (active layer-substrate/active layer-vacuum), the maximum valueis k = 12. Numerically this results in < t >= 22 ps for a one micrometer thick structure. Thequadratic dependence on layer thickness has been demonstrated for structures down to 200 nmthickness, which deliver typical pulse response times of a few picoseconds. Pulses from the150 nm thick strained layer structures are shorter than 2.5 ps (FWHM).

b) This result is limited by two experimental aspects: First the time resolution of the appara-tus, and second by the transit time spread that is induced by different kinetic energies of theelectrons at the cathode surface.

c) Depolarising effects during the diffusive transport to the surface happen on much largertimescales (τs = 50−100 ps) than the emission time. Under these conditions the formula P =P0/(1+ < t > /τs) is a good approximation for reduction of beam polarisation during diffusion[23] (P0 is the polarisation in the conduction band directly after photoemission). From theexperimentally observed < t > < 2 ps it can be deduced that the diffusion process cannot bethe major source of polarisation loss.

d) In our ultra fast-spin-resolved measurements a reduction of polarisation by about 4 % in 2 pswas observed. This result was not in agreement with the commonly accepted values for τs. Wefound an explanation in the transit time spread of the electrons, which correlates the arrivaltime of the electrons with their kinetic energy on the photocathode surface. The kinetic energyspread of the electrons is only about 100-200 meV, depending on the value of the photocathodenegative electron affinity (NEA). The time spread is then generated by the different initial ve-locities of the photoelectrons. The transit time spread is exclusively generated on the very firstpart of the acceleration stage of the source until the velocity differences become negligible.The time spread is about 1 ps for our extraction gradient of 1 MV/m. So the bunch observedat the analyzer represents the convolution of diffusive time spread and of transit time spread,which are of the same order of magnitude. The energy loss of electrons is itself correlated with


depolarisation. It takes place in the band bending region at the surface of the semiconductor,and that is believed to happen on a subpicosecond time scale. The correlation of energy andpolarisation has been measured for our semiconductor structure with a high resolution electro-static spectrometer [33]. Taking into account these effects leads to a reasonably good fit to theobserved data [32].

C.3.5.2 Space Charge Expansion

Presently an optimization of the detection system is taking place that has the goal to improvethe intensity-resolved as well as the spin-resolved time resolution [34], [35]. However, it seemsdifficult to achieve a value well below 1 ps which would be necessary to clearly disentangleenergy correlated and diffusive effects for the strained layer cathodes.

Nevertheless it seems possible to increase the bunch length by space charge effects withoutdestroying all the information that would be represented by the true pulse profile. In this casethe experimental resolution is sufficient to gain new information.

The expansion of the pulse uses the fact that the average distance of two electrons in a ultrashort bunch is increased during the acceleration process, since it is a flat disk at low velocitiesand a cylinder of a height l = βctp after the acceleration. This means that the average distanceof the electrons is increased during acceleration, thus converting the Coulomb energy intokinetic energy of the motion in longitudinal direction.

The order of magnitude of the effect can be estimated as follows for our transverse beam sizeof 300µm: At a bunch charge of 4 fC and a pulse length of 1 ps the average Coulomb energyis about E = 1 eV/electron; E scales linear with the bunch charge, since the total energy isEbunch ∝ N(N − 1) ≈ NE (N = number of electrons in the bunch). The velocity spread willtherefore scale – neglecting relativistic effects for this crude estimation – like the square-rootof the bunch charge. For a complete transfer into the longitudinal motion (which neglects apossible transfer into transverse degrees of freedom) it amounts to about 0.1 % at β = 0.55which results in a time spread of about 6 ps/m, corresponding to 18 ps for the three meter longdrift space to the analyzer in our apparatus. Figure C.35 shows the experimentally measuredexpansion. The points follow the expected square-root behavior in the region where the resultsare not dominated by the experimental resolution.

The corresponding kinetic energy spread had already observed in [31], it amounts to severalhundred eV at a bunch charge of 4 fC if observed from the laboratory frame.

If the initial order of the electrons is conserved during the acceleration and the drift, the ex-panded pulse allows to determine, e.g., the polarisation of the first part of the bunch. In this partwe expect to find the electrons which arrived first at the surface in the diffusive process andhave also avoided energy losses in the surface region. Such electrons should carry the highestpolarisation. The result of the spin- and intensity resolved measurement on an expanded pulseis shown in Fig. C.36.

The interpretation of the intensity profile is difficult, since the transformation by the spacecharge is nonlinear. However, the spin-resolved curve shows an increase at the beginning ofthe pulse and a corresponding decrease at the end. It seems that the expansion enabled to sep-arate highly polarised fractions from depolarised parts of the beam. The measured maximum


Figure C.35: Space charge expansion.

Figure C.36: Expansion by space charge: Time resolved intensity and spin-polarisation measurement.


polarisation of 89.5%±4% is one of the highest values ever achieved with a NEA photocath-ode.

It is planned to modify the electron source in a way to introduce a short drift space at lowenergy (100 eV) to use the natural velocity differences of the NEA cathode instead of the spacecharge effects. A 10 cm drift at 100 eV would separate the electrons by about 30 ps. This valuewould be large enough to cut away the low polarised part of the bunch by the MAMI-chopper.The potential advantage of this regime towards the space-charge expansion lies in the greaterflexibility if compared to the space-charge regime.

C.3.5.3 New Photocathode Material

We have participated in an international research project with the goal of developing optimizedphotocathode structures [36]. The photocathode SL-5-537 – the pulse response of this structurewas already presented in Fig. C.36 – has resulted in an improvement of photocathode parame-ters with respect to strained layer structures. The cathode was developed at the Joffe Institute inSt. Petersburg together with the Sankt Petersburg Technical University (SPTU). The photocath-ode consists of a superlattice structure (In Al Ga As / Al Ga As) of 15 periods. The compositionof this quaternary / ternary compound is chosen in a way to optimize miniband-splitting. Fig-ure C.37 shows that this has been achieved, the splitting (region with high spin-polarisation)is about 150 meV wide. The quantum efficiency approaches 1 % at polarisations of 80 %. Theoptimum wavelength range is between 780 and 800 nm. This is well adapted for commerciallyavailable laser-diodes and also for laser amplifiers.

We intend to do first tests at MAMI with this photocathode during this year.

Figure C.37: Polarisation and quantum efficiency of new superlattice structure SL-5-337.

[1] H. Euteneuer et al.: “The 4.90 GHz Accelerating Structure for MAMI C”, ProceedingsEPAC2000, Wien, p. 1954


[2] H. Euteneuer et al.: “Experiences in Fabrication and Testing the Prototype of the4.90 GHz Accelerating Sections for MAMI C”, Proceedings LINAC2004, Lubeck (to bepublished)

[3] H. Euteneuer et al.: “Experiences in Fabricating and Testing the Rf-Sections of the MainzMicrotron”, Proceeding LINAC86, Stanford, p. 508

[4] Jahresbericht 2000 - 2001, Institut fur Kernphysik, Johannes Gutenberg-UniversitatMainz, 2002, p. 25

[5] G. Faillon, “Rf-Sources for Proton Linacs”, Proceeding LINAC86, Stanford, p. 122

[6] Arbeits- und Ergebnisbericht 1999 - 2001 (SFB443), Institut fur Kernphysik, JohannesGutenberg-Universitat Mainz, p. 177 ff

[7] P. Jennewein: “Testmessungen an Probeplatten des DSM”, Interner Report MAMI 03/01

[8] M. Seidl: “Untersuchungen zur Strahldynamik am Mainzer Mikrotron”, Dissertation, In-stitut fur Kernphysik, Johannes Gutenberg-Universitat Mainz, 2003

[9] F. Hagenbuck: “Entwicklung eines neuartigen bildgebenden Verfahrens zur digitalen Sub-straktionsradiographie mit Ubergangsstrahlung am Mainzer Mikrotron MAMI”, Disser-tation, Institut fur Kernphysik, Johannes Gutenberg-Universitat Mainz, 2001

[10] U. Ludwig-Mertin et al.: “Beam Transport Magnets with 2.2 Tesla”, Proceedings EPAC2002, Paris p. 2361

[11] K. Aulenbacher: “Vakuumsystem am HDSM”, Interner Report MAMI 05/2001

[12] A. Jankowiak: “Synchrotronstrahlungsinduzierte Gasdesorption im HDSM und Vergleichzum RTM3”, Interner Report MAMI 04/2001

[13] F. Hagenbuck et al.: “Magnetic Field Correction of the Bending Magnets of the 1.5 GeVHDSM”, Proceedings EPAC 2004, Lucerne (to be published)

[14] Jahresbericht 1996 - 1997, Institut fur Kernphysik, Johannes Gutenberg-UniversitatMainz, p. 139

[15] C. Sinclair: “Performance of the CEBAF polarized electron source under high averagecurrent”, A. Gute and S. Lorenz and E. Steffens(eds.): Proceedings of the InternationalWorkshop “Polarized sources and targets”, (1999), p. 222

[16] K. Aulenbacher et al.: “New results from the Mainz polarized electron facilities”, Pro-ceedings Spin 2000, Osaka, AIP publishing Vol. 570, p. 949

[17] K. Aulenbacher et al.: “Status of the polarized source at MAMI”, K. Y.I. Makdisi,A.U.Luccio, W.W.Mackay, editors: AIP proceedings Vol. 675 ,p. 1088, Proceedings of15th international Spin physics symposium (SPIN-2002) Melville, New York (2003)

[18] C. Benvenuti et al.: “A novel route to extreme vacua: the non evaporable getter thin filmcoatings”, Vacuum 53 (1999), p. 219

[19] R. Barday: “Verbesserung der Lebensdauer der MAMI Quelle polarisierter Elektronen”,Dissertation, Institut fur Kernphysik, Johannes Gutenberg-Universitat Mainz (in prepara-tion).

319

[20] M. Wiessner: “Weiterentwicklung des Halbleitersynchrolasersystems an MAMI”,Diplomarbeit, Institut fur Kernphysik, Johannes Gutenberg-Universitat Mainz, (2001)

[21] G. Arz: “Aufbau eines hochfrequenzsynchroniserten MOPA-systems mit hoher Sta-bilitat”, Diplomarbeit, Institut fur Kernphysik, Johannes Gutenberg-Universitat Mainz,(2004) (in preparation)

[22] K. Winkler: “Photoluminszensspektroskopie an Strained Layer Photokathoden”, Diplom-arbeit, Institut fur Kernphysik, Johannes Gutenberg-Universitat, Mainz,(2002)

[23] K. Aulenbacher: “Erzeugung intensiver hochpolarisierter Elektronenstrahlen mit ho-her Symmetrie unter Helizitatswechsel”, Institut fur Kernphysik, Johannes Gutenberg-Universitat Mainz, 2004 (to be published)

[24] R. Mair et al.: “Anisotropies in Strain and Quantum Efficiency of Strained Ga As Grownon Ga As P.” Phys. Lett. A212 (1996) p.231

[25] F. Maas et al.: “Measurement of Strange Quark Contributions to the Nucleon’s FormFactors at Q2 = 0.230 (GeV/c)2”, submitted to Phys. Rev. Let.

[26] C. Prescott: private communication.

[27] V. Tioukine et al.: “Operation of the MAMI Accelerator with a Wien Filter Based SpinRotation System”, (to be published)

[28] M. Steigerwald: “MeV-Polarimetry at Jefferson-Lab.”, Proceedings Spin 2000, Osaka,AIP publishing Vol. 570, p. 935

[29] T.J. Gay et al.: “Optical Electron Polarimetry with Heavy Noble Gases”, Phys. Rev. A53,3, p. 1623 (1996)

[30] B. Collin: “Contenu etrange du nucleon: mesure de l’asymetrie de violation de paritedans l’experience PVA4 a MAMI, etude et developpement d’un polarimetre optique”,PhD-thesis Universite Paris XI-Orsay (2003)

[31] P. Hartmann et al.: “Picosecond Polarized Electron Bunches from a Strained LayerGa As P Photocathode”, Nuclear Instruments and methods A 379, p. 15 (1996)

[32] K. Aulenbacher et al: “Pulse Response from thin 3/5 Semiconductor Photocathodes”,Journal of applied physics 92,12, p. 7536 (2002)

[33] Y. Mameev et al.: “Highly Polarized Energy Resolved Near-Threshold Electron Photoe-mission from Strained Ga As”, A. Gute and S. Lorenz and E. Steffens(eds.): Proceedingsof the International Workshop “Polarized sources and targets”, (1999), p. 246

[34] M. Weis: “Optimierung der Zeitauflosung fur eine synchro-streak Apparatur mitHilfe von CCD-Kameras”, Diplomarbeit, Institut fur Kernphysik, Johannes Gutenberg-Universitat Mainz, (2004) (in preparation)

[35] R. Bolenz: “Optimierung der Effizienz eines Mottpolarimeters zur spinaufgelosten Ultra-kurzzeitspektroskopie”, Diplomarbeit, Institut fur Kernphysik, Johannes Gutenberg-Universitat Mainz, (2004) (in preparation)

[36] European Union: INTAS-Project 99-00125

Project C: Construction of MAMI C - A1-Collaboration · · 2004-04-29Project C: Construction of...

Documents

Transcript of Project C: Construction of MAMI C - A1-Collaboration · · 2004-04-29Project C: Construction of...