Improvement of beam parameters

Mid-1996 coincides with the completion of the first upgrade of the ESRF source, meaning an achieved increase in brilliance by a factor 100 with respect to the Foundation Phase Report, from 1018 to 1020.

The evolution of the working point has taken place gradually, in successive steps, by lowering the coupling, lowering the horizontal emittance and matching the ß function*. This has led to a significant improvement in ESRF performances with a higher average brilliance (see Figure 83).

The current idea is to generalise low ßz ~ 2.5 in all straight sections (leading to a factor 2 increase in brilliance in the high-ß straight sections).

Further gains (each bringing about the same factor 2 improvement in brilliance) are aimed at by reducing both the horizontal emittance (3.1 nm) and coupling (~ 2 x 10-3). Such upgrades, together with the anticipated successful testing of the 10 mm gap undulator vessels, will constitute the second upgrade of the machine. At this second upgrade, performances will approach the 1021 level, seven orders of magnitude above the level of the second generation machines. Having reached 1021, the storage ring will stand two orders of magnitude below what could be considered as the ultimate performance (1023) corresponding to a ring in which the electron beam emittances are null.



Exotic modes

We are currently developing the 32-bunch mode of operation and the new "twin" hybrid mode with two single bunches equally spaced in the space left by the 1/3 filling (see Figure 84).

We are also investigating operation of the storage ring at different energies. We will explore the reasonable upper limit of the machine, which is presumably around 6.2/6.3 GeV and combine our finding with the 10 mm insertion device gap to fix our optimised working point, including electricity cost aspects.

Lowering insertion device gaps and increasing electron beam energy pushes the undulator photon spectrum to higher energies. On the other hand there is a need, in particular for experiments using circularly polarised light, to explore the lower photon energy limit. We will make the 5 GeV operation of the storage ring reliable so that it could be served at times in User Service Mode, in the same way as the single bunch mode.

Exploration of the possibility of focusing the X-ray beam downstream the sample, equivalent to an increase in the brilliance by a factor of 10, has also begun (one element of the third upgrade).

The single bunch in the storage ring (duration~ 33 picoseconds) cannot be made extremely short. For certain experiments 1 picosecond or even 100 femtosecond-long pulses would be necessary. Another possibility for performing dynamic studies is not to use the conventional pump and probe arrangement, but rather to use the standard pulse of the storage ring as the probe of the laser triggered fast reaction. The modulation amplitude would be able to be time-resolved with an ultra-fast jitter-free streak camera used as a detector.

A collaboration project for the production of such a detector is currently being carried out with the Center for Ultrafast Optical Science (CUOS), Michigan, USA. The project would take place over three years (up to the end of 1998) and end with the delivery of a jitter-free streak camera system.


Time structure beam modes


For half of the year, two transmitters are connected to the storage ring, each powering a pair of cavities. During this time the machine is operated at the standard high intensity multibunch current, which is currently at 200 mA. To cure longitudinal High Order Mode driven multibunch instability, only one third of the ring is filled. This 1 µs macroscopic pulse is used in some experiments.

During the rest of the year, one transmitter is disconnected from the storage ring to allow for maintenance, pre-conditioning of new couplers on the cavity test-stand, commissioning of klystrons and various component tests. During these periods, the storage ring is operated in time structure modes, the remaining connected klystron allowing a maximum of 175 mA at full power to be stored.

Single bunch is delivered for only a few shifts. The maximum 15 mA stored current mainly fulfils the requirements of ID9 (the Laue Diffraction beamline). A chopper allows them to get an X-ray pulse at a reduced frequency with a bunch length of the order of 100 picoseconds fwhm. The vertical chromaticity is pushed to high values in order to cure transverse instabilities. The drawback of this method is the reduction in lifetime (6 hours at 15 mA). Transverse feedback in order to stabilise the beam at a reduced chromaticity looks promising.

In the hybrid mode, a single bunch is filled in the gap of the one third multibunch macroscopic pulse. The operational current is 140 mA with 7 mA in the single bunch. The filling of the 1/3 filling with the injector still in short-pulse mode induces a refilling time of the same order as that in the standard 1/3 filling mode. This filling process has the advantage of giving a sharp square macroscopic pulse. The filling of two single bunches in the gap of the macroscopic pulse was tested in operation and looks encouraging. However, the impaired purity (due to timing problems) after this macroscopic pulse did not offer the experiments the possibility to benefit from this second short pulse.

In the standard 16-bunch mode, a 90 mA maximum stored current is delivered to the users. The present limitation is given by longitudinal High Oder Mode instabilities driven by the four storage ring cavities and by the overheating and consequent outgassing of the radio-frequency liners of the vacuum chamber. A new cavity temperature regulation system is being developed to cure longitudinal instabilities and an R&D programme on the radio-frequency liners is underway. At the restart of 1996, a tedious reconditioning of the cavities was necessary after the change of some couplers.

A purity of the short pulses better than 10-7 is routinely delivered. A new diagnostic using an X-ray avalanche photodiode installed in the tunnel close to ID8 is used to check the purity. It has an excellent dynamic range (better than 108) and a good time resolution (10-6 at 2.8 ns of the main bunch). Nevertheless, this diode looks sensitive to radiation. Improved shielding is being made. The linac has been optimised in order to deliver an initial purity in the range of 10-3, which greatly helps the cleaning process.

A protection of the machine against vertical beam instability has been installed. It is based on the measurement of the temperature of two jaws forming a slit that intercepts the far X-ray beam tails in the vertical direction.

The lifetime in short-pulse mode is dominated by the Touschek effect. It has been affected by the reduction in coupling and the increase of the chromaticity to cure the instabilities. The installation of a third radio-frequency accelerating unit will provide additional accelerating voltage which will help to increase the lifetime.


Construction of a third radio-frequency acceleration unit for the ESRF storage ring


Four 352.2 MHz five-cell cavities and two 1.0 MW transmitters (upgraded to 1.3 MW in January 94) were initially installed on the ESRF storage ring to support an operation at the design current of 100 mA at 6 GeV. Already during commissioning, higher currents could be stored and the standard high intensity multibunch operation is now at 200 mA. Each of the two transmitters must then deliver 700 to 750 kW. However, although each cavity is fed through two input couplers, these are now operated close to their upper power limits. Note that longitudinal High Order Mode driven multibunch instabilities showing up around 70 mA are suppressed with Landau damping obtained by filling only one third of the storage ring.

For periods of about 50 % of the year, one transmitter is disconnected from the storage ring to allow maintenance, pre-conditioning of new couplers on the cavity teststand, commissioning of klystrons and various component tests. At full klystron power, 175 mA can then be stored with the remaining transmitter connected to the four storage ring cavities. During these periods, the storage ring is often operated in single bunch mode at 15 mA, in 16-bunch mode at 90 mA or in multibunch mode at a reduced intensity of about 140 mA.

In order to guarantee full availability of high intensity operation and to improve the reliability by limiting the nominal power load on the various radio-frequency subsystems, the ESRF is now constructing a third 1.3 MW transmitter feeding a third pair of cavities which will come into operation at the end of 1997. This will in particular allow us:

  • To lower the power delivered by the klystrons and to increase their lifetime,
  • to lower the power transmitted through the cavity couplers,
  • to still sustain high intensity operation in case of failure of either a transmitter or a pair of cavities, and when one transmitter is required for radio-frequency tests (no further need for an operation with only one klystron at full power),
  • to increase the accelerating voltage for a better lifetime especially in few bunch operation and to account for the additional beam loading due to further installation of insertion devices, while still operating the klystrons and couplers at moderate power,
  • to define a new operation mode with Landau damping of longitudinal multibunch instabilities for homogeneous storage ring filling, by operating one pair of cavities at frf + fo.



Development of radio-frequency liners


The radio-frequency liners (sometimes called radio-frequency fingers, depending on their design) are some specific flexible transitions, located inside the bellows in-between two adjacent vacuum vessels. They ensure a smooth electrical continuity (particularly at high frequency) for the image current which circulates along the vacuum vessel.

The present radio-frequency liners mounted in the bellow section of the storage ring are made of beryllium/copper fingers pressed against a stainless steel sleeve by means of a helical nimonic spring (see Figure 85).

This design produces sharp temperature increases for stored beams higher than 90 mA in the 16-bunch filling mode. This temperature increase generates high pressure bursts which eventually trip the beam.

To obtain a reliable operation up to 200 mA in the 16-bunch mode, a new concept has been developed based on the use of a metallic mesh: the so-called "sock" model. This model is made of a stainless steel mesh composed of 150 strands of 130 mm long wires, 0.5 mm in diameter. This mesh is welded at both ends to stainless steel sleeves. Particular care must be paid to obtain a smooth welding seam free of any small cavities which may excite High Order Modes.

The performances achieved with this sock model with a 16-bunch beam are indeed promising (see Figure 86):

  • no more pressure bursts
  • the temperature measured on the mesh varies linearly with current but reaches 180 °C for 95 mA - this temperature is too high, therefore a second prototype was made by copper-plating a stainless steel sock.

The following graph illustrates the behaviour of this model. It can be observed that, when changing from 1/3 filling to 16-bunch filling, no sharp increase in temperature is detected in the 16-bunch mode, contrary to what is observed on conventional radio-frequency liners. However, the copper plating of the complete sock reduces mechanical flexibility.

For this reason, a final version of the sock is under preparation. It will be composed of 0.3 mm silver plated stainless steel wire.





Table 2 details the current status with regards to front-ends.

New carbon filtering system for high heat load beamlines: for the last six months certain beamlines have identified a need for maximisation of the flux available by increasing the number of undulator carriages in the straight section to a maximum of three, which is equivalent to a 5 metre long undulator. Whilst this has always been foreseen, the fact that the machine is also run at 200 mA has meant that the present front-end design is not suitable for this type of operation. The reduction of the insertion device gap also increases the heatload problem on the front-end and with future developments to 10 mm chambers this will only get worse. Essentially all the problems lie in passing this high power density beam through a vacuum window, traditionally made from beryllium and protected by carbon filters.

Since January 1995 the front-end group has been working on this problem both theoretically and experimentally on the front-end of ID27, which has been dedicated to this task. Initially, the present design as installed on all undulator beamlines was tested to its limit. The conclusion made was that the design is not acceptable for future needs, but this work also allowed us to compare theoretical results with experimental data.

A prototype of a new design of carbon filter/beryllium window assembly has been made and tested to extreme conditions without showing any sign of damage. A beam from a 5 metre long undulator with a 20 mm gap and at 200 mA was successfully passed through a beryllium window. The system relies on collimating the beam to a minimum size and allowing only the central cone of the undulator beam to pass through the window. This appears to be totally acceptable to the beamline users. In addition, a technique for profiling the carbon filter has been developed and a more effective material will be used. The fact that very tight collimation is being used means that the whole system has to be motorised to find the beam centre of mass.

The purchase of ten systems has been started with the first pre-production elements being assembled on ID27 followed closely by systems for ID10, ID16 and ID30 for September and November 1996.

Future developments will concentrate on the use of diamond as a leaktight vacuum window for extremely high X-ray fluxes.



Insertion devices


Table 3 gives the list of insertion device segments installed on the storage ring in June 1996. It amounts to 35 segments, corresponding to a cumulated length of 55 meters. The maximum magnetic field ranges from 0.21 T for the shortest period undulator to 4 T for the superconducting wiggler. The gradual replacement of the 19 mm thick insertion device vacuum chamber to 15 mm chambers has resulted in a small reduction of the optimised undulator period.

At the end of 1995, the first segmented phased undulators were installed on ID16 and ID30. The magnetic design for the phasing section is presented in Figure 87. Contrary to the usual phasing method, this phasing section allows an independent gap control of both undulator segments which is possible given the air space of 6 mm left between each undulator segment. The field integral generated at the junction between the undulator segments is, for any independent gap setting of each segment, smaller than 30 G x cm (10 G x cm) in the vertical (horizontal) plane. As a result, no correction coil is needed. The use of such a phasing section presents a number of other advantages. The tolerance of longitudinal positioning of one undulator with respect to the other is reduced by a factor 3 with respect to the usual phasing method. Finally, one may envisage phasing undulators made at a different time with magnet blocks from two different suppliers without difficulty. All new undulators under manufacture will be equipped with these phasing sections. One should nevertheless mention that the brilliance of two phased undulator segments is not equal to 4 times the brilliance of an individual segment. The reason comes from the non-zero electron beam emittance and energy spread. A ratio of 2.3 (instead of 4) has been measured on ID30 on harmonic 5 at a photon energy of 28 keV in agreement with the expectations. An incorrect phasing could have resulted in a ratio as low as 1.7. A larger (lower) ratio is expected at lower (higher) photon energies.

Figure 88 represents the brilliance reached by the use of two phased undulators installed at the ESRF compared to the predictions of the Foundation Phase Report. A brilliance of 1020 is reached around 4 keV and values higher than 1019 are reached in the most useful range, 2-30 keV. A large part of the improvement originates from the electron beam emittance reduction. In the medium term a further improvement of the brilliance is possible by the use of 5 m long, 10 mm gap undulators and by further reducing the electron beam emittances.



15 mm undulator vessels devoid of distributed pumping


Following problems caused by dust produced by the NEG distributed pumps in the first generation of 5 meter long, 19 mm high undulator vessels, and to adapt to the demand for a reduced undulator gap, a new 15 mm high vessel devoid of distributed pumping was designed in 1995.

The guidelines for this new design were (see Figure 89):

  • simplification of the geometry of the part under vacuum,
  • use of stainless steel with a special thermal treatment to reduce gas desorption to the minimum: vacuum firing at 950° C (in the CERN oven), bakeout cycles in the lab at 400° C, final bakeout at 400 °C after assembly on the storage ring.

A prototype was installed on the machine in March 1995 and demonstrated that the pressure distribution obtained with beam with this design, devoid of distributed pumping, is very similar to that measured on a 19 mm high vessel with NEG distributed pumps.

Twelve 15 mm high, 5 meter long insertion device vessels devoid of distributed pumping were manufactured. Eight have already been installed on the storage ring.

In May 1996, a close follow-up of the Bremsstrahlung dose produced by such a newly installed vessel in a low-ß straight section was carried out. It was found that the measured Bremsstrahlung dose closely follows the decay of pressure measured at both ends of the insertion device vessel corresponding to the conditioning of the vessel. This means that the integrated pressure between input and output of the 5 meter long vessel is indeed proportional to the pressure indicated by the vacuum gauges placed at both extremities.

There are further demands to reduce insertion device gaps to gain in photon energy and flexibility. To cope with this, a 5 meter long, 10 mm high insertion device vessel devoid of distributed pumping has been designed. A prototype will be installed on the machine in September 1996.


The High Quality Power Supply (HQPS)


The HQPS system operated successfully during the 22 peak electricity cost days defined within the "ejp" (effacement jours de pointe) contract with only one failure (due to a software bug) on the first day. During last winter, the global fuel consumption was 538 000 litres to produce 1854 MWh of clean 20 kV mains. Since then, the master controller has been upgraded and is now better adapted to the ESRF needs. Thus, the problem of the first failure should not reoccur.

On 8 June 1996, a particularly severe storm affected the telephone and fire alarm detection systems as well as a few other less crucial elements on the site. Several direct-hit lightning bolts caused the beam to be lost, due to the failure of the battery-fed Uninterrupted Power Supplies (UPS). Since then, not a single beam loss has been suffered despite the large number of mains drops.

Contrary to the neighbouring institutes such as SGS-Thomson, the CNRS, etc. which are, in case of storms, instructed to close down by the storm alarm system "Meteorage", the ESRF Diesel engines only started up a total of 20 times over the last 12 months whereas more than 220 drops were recorded on the input. No beam losses were experienced.

With regard to reliability, the system is permanently watched by the controller which automatically calls the system's constructors, a Belgian company, by modem, when necessary. Minor repairs are now dealt with during the visit included in the maintenance contract (every two months). The added value of this 8 MW power plant to the availability of the beam is irrefutable.


Operation in 1995/1996

From August 1995 to July 1996, 599 shifts were dedicated to beam delivery to the users. This represents a total of 4792 hours (in addition of which 1184 were dedicated to machine studies). 40 % of these shifts (247) were delivered in the 1/3 filling mode. Nevertheless, the interest for exotic modes is increasing. For instance, there were as many shifts in 16-bunch mode in the first six months of 1996 (91 shifts) as for the whole year in 1995 (92 shifts) (Figure 90).

Compared to 1995, the fundamental modification of the operation schedule in 1996 lies in the lengthening of the runs (involving less shutdown periods). Whilst 4752 User Service Mode hours were scheduled in 1995, 5304 hours are programmed for 1996.

It is worth highlighting the fact that, despite the increasing number of Users Service Mode (USM) time, the availability of the beam has been progressively improving since 1993. It reached 89 % in 1994, 93 % in 1995 and since the beginning of 1996 this figure has been approaching the 95 % mark (Figure 91). Furthermore, the unavoidable "dead time" for refills, amounting to 2 %, is automatically deducted from the availability figure. To give an example, since the beginning of 1996, actual failures account for 3 %. These rather good figures can be explained by the implementation of several major improvements to equipment:

  • the HQPS, whose benefits are discussed above,
  • operation of the machine with two radio-frequency transmitters, each of them feeding two cavities at lower power, which proved to be very efficient in reducing the number of trips due to the klystron operating at its upper limit,
  • regular and systematic preventive maintenance on the water cooling circuits which were the third biggest cause of failures throughout 1994 and during the early stages of 1995.

Most of the runs during the considered period (August 1995 to July 1996) were excellent in terms of availability and also in terms of mean time between failures (an essential criterion for the users): 98 % was achieved in the best week (Figure 92) and the worst weekly figure stood at 87 %.

The most significant incident last year was an interruption of 14 hours (in February 1996) when an unfortunate sequence of events led to the breakdown of the power supply of the radio-frequency anode modulator. This breakdown was followed by the burning up of a transformer of the thyratron rack operating on one linac modulator. The time lost was further extended due to the fact that the spare part itself was malfunctioning.

Another major event to be mentioned, which disturbed the USM schedule but did not lead to any loss in USM hours, occurred at the beginning of 1996 when two radio-frequency couplers had to be replaced.

Finally, it should not be forgotten that a beam of 200 mA was routinely delivered throughout the first months of 1996, also corresponding to the best run in terms of availability with 96 %. Another point not to be overlooked is the achievement of being able to decrease the coupling from 2 % to 1 % (and even 0.5 % in the very best of conditions).