Résumé

MAX IV ouvre la voie pour un microscope à rayons X dernier cri

Les sources synchrotron de rayons X de troisième génération ont révolutionné l’étude de la structure des matériaux. De nombreuses installations de ce type sont exploitées dans le monde entier, et sont utilisées par une communauté d’environ 50 000 personnes ; elles permettent aux chercheurs d’étudier les sujets les plus variés, de la structure des protéines utilisées pour le développement de médicaments à l’étude des fossiles ou du patrimoine culturel. L’installation suédoise MAX IV, qui a commencé à fonctionner cette année, est le premier anneau de stockage utilisant un achromat à courbure multiple ; grâce à cette conception novatrice, les faisceaux de rayons X produits sont plus petits et plus stables que ceux des autres installations, ce qui fait de MAX IV la source de rayons X la plus brillante du monde. D’autres infrastructures suivent la même voie, le but ultime étant la construction d’un synchrotron fonctionnant à la limite de la diffraction des rayons X.

Since the discovery of X-rays by Wilhelm Röntgen more than a century ago, researchers have striven to produce smaller and more intense X-ray beams. With a wavelength similar to interatomic spacings, X-rays have proved to be an invaluable tool for probing the microstructure of materials. But a higher spectral power density (or brilliance) enables a deeper study of the structural, physical and chemical properties of materials, in addition to studies of their dynamics and atomic composition.

For the first few decades following Röntgen’s discovery, the brilliance of X-rays remained fairly constant due to technical limitations of X-ray tubes. Significant improvements came with rotating-anode sources, in which the heat generated by electrons striking an anode could be distributed over a larger area. But it was the advent of particle accelerators in the mid-1900s that gave birth to modern X-ray science. A relativistic electron beam traversing a circular storage ring emits X-rays in a tangential direction. First observed in 1947 by researchers at General Electric in the US, such synchrotron radiation has taken X-ray science into new territory by providing smaller and more intense beams.

Generation game

First-generation synchrotron X-ray sources were accelerators built for high-energy physics experiments, which were used “parasitically” by the nascent synchrotron X-ray community. As this community started to grow, stimulated by the increased flux and brilliance at storage rings, the need for dedicated X-ray sources with different electron-beam characteristics resulted in several second-generation X-ray sources. As with previous machines, however, the source of the X-rays was the bending magnets of the storage ring.

The advent of special “insertion devices” led to present-day third-generation storage rings – the first being the European Synchrotron Radiation Facility (ESRF) in Grenoble, France, and the Advanced Light Source (ALS) at Lawrence Berkeley National Laboratory in Berkeley, California, which began operation in the early 1990s. Instead of using only the bending magnets as X-ray emitters, third-generation storage rings have straight sections that allow periodic magnet structures called undulators and wigglers to be introduced. These devices consist of rows of short magnets with alternating field directions so that the net beam deflection cancels out. Undulators can house 100 or so permanent short magnets, each emitting X-rays in the same direction, which boosts the intensity of the emitted X-rays by two orders of magnitude. Furthermore, interference effects between the emitting magnets can concentrate X-rays of a given energy by another two orders of magnitude.

Third-generation light sources have been a major success story, thanks in part to the development of excellent modelling tools that allow accelerator physicists to produce precise lattice designs. Today, there are around 50 third-generation light sources worldwide, with a total number of users in the region of 50,000. Each offers a number of X-ray beamlines (up to 40 at the largest facilities) that fan out from the storage ring: X-rays pass through a series of focusing and other elements before being focused on a sample positioned at the end station, with the longest beamlines (measuring 150 m or more) at the largest light sources able to generate X-ray spot sizes a few tens of nanometres in diameter. Facilities typically operate around the clock, during which teams of users spend anywhere between a few hours to a few days undertaking experimental shifts, before returning to their home institutes with the data.

Although the corresponding storage-ring technology for third-generation light sources has been regarded as mature, a revolutionary new lattice design has led to another step up in brightness. The MAX IV facility at Maxlab in Lund, Sweden, which was inaugurated in June, is the first such facility to demonstrate the new lattice. Six years in construction, the facility has demanded numerous cutting-edge technologies – including vacuum systems developed in conjunction with CERN – to become the most brilliant source of X-rays in the world.

The multi-bend achromat

Initial ideas for the MAX IV project started at the end of the 20th century. Although the flagship of the Maxlab laboratory, the low-budget MAX II storage ring, was one of the first third-generation synchrotron radiation sources, it was soon outcompeted by several larger and more powerful sources entering operation. Something had to be done to maintain Maxlab’s accelerator programme.

The dominant magnetic lattice at third-generation light sources consists of double-bend achromats (DBAs), which have been around since the 1970s. A typical storage ring contains 10–30 achromats, each consisting of two dipole magnets and a number of magnet lenses: quadrupoles for focusing and sextupoles for chromaticity correction (at MAX IV we also added octupoles to compensate for amplitude-dependent tune shifts). The achromats are flanked by straight sections housing the insertion devices, and the dimensions of the electron beam in these sections is minimised by adjusting the dispersion of the beam (which describes the dependence of an electron’s transverse position on its energy) to zero. Other storage-ring improvements, for example faster correction of the beam orbit, have also helped to boost the brightness of modern synchrotrons. The key quantity underpinning these advances is the electron-beam emittance, which is defined as the product of the electron-beam size and its divergence.

Despite such improvements, however, today’s third-generation storage rings have a typical electron-beam emittance of between 2–5 nm rad, which is several hundred times larger than the diffraction limit of the X-ray beam itself. This is the point at which the size and spread of the electron beam approaches the diffraction properties of X-rays, similar to the Abbe diffraction limit for visible light (see panel below). Models of machine lattices with even smaller electron-beam emittances predict instabilities and/or short beam lifetimes that make the goal of reaching the diffraction limit at hard X-ray energies very distant.

Although it had been known for a long time that a larger number of bends decreases the emittance (and therefore increases the brilliance) of storage rings, in the early 1990s, one of the present authors (DE) and others recognised that this could be achieved by incorporating a higher number of bends into the achromats. Such a multi-bend achromat (MBA) guides electrons around corners more smoothly, therefore decreasing the degradation in horizontal emittance. A few synchrotrons already employ triple-bend achromats, and the design has also been used in several particle-physics machines, including PETRA at DESY, PEP at SLAC and LEP at CERN, proving that a storage ring with an energy of a few GeV produces a very low emittance. To avoid prohibitively large machines, however, the MBA demands much smaller magnets than are currently employed at third-generation synchrotrons.

In 1995, our calculations showed that a seven-bend achromat could yield an emittance of 0.4 nm rad for a 400 m-circumference machine – 10 times lower than the ESRF’s value at the time. The accelerator community also considered a six-bend achromat for the Swiss Light Source and a five-bend achromat for a Canadian light source, but the small number of achromats in these lattices meant that it was difficult to make significant progress towards a diffraction-limited source. One of us (ME) took the seven-bend achromat idea and turned it into a real engineering proposal for the design of MAX IV. But the design then went through a number of evolutions. In 2002, the first layout of a potential new source was presented: a 277 m-circumference, seven-bend lattice that would reach an emittance of 1 nm rad for a 3 GeV electron beam. By 2008, we had settled on an improved design: a 520 m-circumference, seven-bend lattice with an emittance of 0.31 nm rad, which will be reduced by a factor of two once the storage ring is fully equipped with undulators. This is more or less the design of the final MAX IV storage ring.

In total, the team at Maxlab spent almost a decade finding ways to keep the lattice circumference at a value that was financially realistic, and even constructed a 36 m-circumference storage ring called MAX III to develop the necessary compact magnet technology. There were tens of problems that we had to overcome. Also, because the electron density was so high, we had to elongate the electron bunches by a factor of four by using a second radio-frequency (RF) cavity system.

Block concept

MAX IV stands out in that it contains two storage rings operated at an energy of 1.5 and 3 GeV. Due to the different energies of each, and because the rings share an injector and other infrastructure, high-quality undulator radiation can be produced over a wide spectral range with a marginal additional cost. The storage rings are fed electrons by a 3 GeV S-band linac made up of 18 accelerator units, each comprising one SLAC Energy Doubler RF station. To optimise the economy over a potential three-decade-long operation lifetime, and also to favour redundancy, a low accelerating gradient is used.

The 1.5 GeV ring at MAX IV consists of 12 DBAs, each comprising one solid-steel block that houses all the DBA magnets (bends and lenses). The idea of the magnet-block concept, which is also used in the 3 GeV ring, has several advantages. First, it enables the magnets to be machined with high precision and be aligned with a tolerance of less than 10 μm without having to invest in aligning laboratories. Second, blocks with a handful of individual magnets come wired and plumbed direct from the delivering company, and no special girders are needed because the magnet blocks are rigidly self-supporting. Last, the magnet-block concept is a low-cost solution.

We also needed to build a different vacuum system, because the small vacuum tube dimensions (2 cm in diameter) yield a very poor vacuum conductance. Rather than try to implement closely spaced pumps in such a compact geometry, our solution was to build 100% NEG-coated vacuum systems in the achromats. NEG (non-evaporable getter) technology, which was pioneered at CERN and other laboratories, uses metallic surface sorption to achieve extreme vacuum conditions. The construction of the MAX IV vacuum system raised some interesting challenges, but fortunately CERN had already developed the NEG coating technology to perfection. We therefore entered a collaboration that saw CERN coat the most intricate parts of the system, and licences were granted to companies who manufactured the bulk of the vacuum system. Later, vacuum specialists from the Budker Institute in Novosibirsk, Russia, mounted the linac and 3 GeV-ring vacuum systems.

Due to the small beam size and high beam current, intra beam scattering and “Touschek” lifetime effects must also be addressed. Both are due to a high electron density at small-emittance/high-current rings in which electrons are brought into collisions with themselves. Large energy changes among the electrons bring some of them outside of the energy acceptance of the ring, while smaller energy deviations cause the beam size to increase too much. For these reasons, a low-frequency (100 MHz) RF system with bunch-elongating harmonic cavities was introduced to decrease the electron density and stabilise the beam. This RF system also allows powerful commercial solid-state FM-transmitters to be used as RF sources.

When we first presented the plans for the radical MAX IV storage ring in around 2005, people working at other light sources thought we were crazy. The new lattice promised a factor of 10–100 increase in brightness over existing facilities at the time, offering users unprecedented spatial resolutions and taking storage rings within reach of the diffraction limit. Construction of MAX IV began in 2010 and commissioning began in August 2014, with regular user operation scheduled for early 2017.

On 25 August 2015, an amazed accelerator staff sat looking at the beam-position monitor read-outs at MAX IV’s 3 GeV ring. With just the calculated magnetic settings plugged in, and the precisely CNC-machined magnet blocks, each containing a handful of integrated magnets, the beam went around turn after turn with proper behaviour. For the 3 GeV ring, a number of problems remained to be solved. These included dynamic issues – such as betatron tunes, dispersion, chromaticity and emittance – in addition to more trivial technical problems such as sparking RF cavities and faulty power supplies.

As of MAX IV’s inauguration on 21 June, the injector linac and the 3 GeV ring are operational, with the linac also delivering X-rays to the Short Pulse Facility. A circulating current of 180 mA can be stored in the 3 GeV ring with a lifetime of around 10 h, and we have verified the design emittance with a value in the region of 300 pm rad. Beamline commissioning is also well under way, with some 14 beamlines under construction and a goal to increase that number to more than 20.

Sweden has a well-established synchrotron-radiation user community, although around half of MAX IV users will come from other countries. A variety of disciplines and techniques are represented nationally, which must be mirrored by MAX IV’s beamline portfolio. Detailed discussions between universities, industry and the MAX IV laboratory therefore take place prior to any major beamline decisions. The high brilliance of the MAX IV 3 GeV ring and the temporal characteristics of the Short Pulse Facility are a prerequisite for the most advanced beamlines, with imaging being one promising application.

Towards the diffraction limit

MAX IV could not have reached its goals without a dedicated staff and help from other institutes. As CERN has helped us with the intricate NEG-coated vacuum system, and the Budker Institute with the installation of the linac and ring vacuum systems, the brand new Solaris light source in Krakow, Poland (which is an exact copy of the MAX IV 1.5 GeV ring) has helped with operations, and many other labs have offered advice. The MAX IV facility has also been marked out for its environmental credentials: its energy consumption is reduced by the use of high-efficiency RF amplifiers and small magnets that have a low power consumption. Even the water-cooling system of MAX IV transfers heat energy to the nearby city of Lund to warm houses.

The MAX IV ring is the first of the MBA kind, but several MBA rings are now in construction at other facilities, including the ESRF, Sirius in Brazil and the Advanced Photon Source (APS) at Argonne National Laboratory in the US. The ESRF is developing a hybrid MBA lattice that would enter operation in 2019 and achieve a horizontal emittance of 0.15 nm rad. The APS has decided to pursue a similar design that could enter operation by the end of the decade and, being larger than the ESRF, the APS can strive for an even lower emittance of around 0.07 nm rad. Meanwhile, the ALS in California is moving towards a conceptual design report, and Spring-8 in Japan is pursuing a hybrid MBA that will enter operation on a similar timescale.

Indeed, a total of some 10 rings are currently in construction or planned. We can therefore look forward to a new generation of synchrotron storage rings with very high transverse-coherent X-rays. We will then have witnessed an increase of 13–14 orders of magnitude in the brightness of synchrotron X-ray sources in a period of seven decades, and put the diffraction limit at high X-ray energies firmly within reach.

One proposal would see such a diffraction-limited X-ray source installed in the 6.3 km-circumference tunnel that once housed the Tevatron collider at Fermilab, Chicago. Perhaps a more plausible scenario is PETRA IV at DESY in Hamburg, Germany. Currently the PETRA III ring is one of the brightest in the world, but this upgrade (if it is funded) could bring the ring performance to the diffraction limit at hard X-ray energies. This is the Holy Grail of X-ray science, providing the highest resolution and signal-to-noise ratio possible, in addition to the lowest-radiation damage and the fastest data collection. Such an X-ray microscope will allow the study of ultrafast chemical reactions and other processes, taking us to the next chapter in synchrotron X-ray science.

Towards the X-ray diffraction limit

Electromagnetic radiation faces a fundamental limit in terms of how sharply it can be focused. For visible light, it is called the Abbe limit, as shown by Ernst Karl Abbe in 1873. The diffraction limit is defined as λ/(4π), where λ is the wavelength of the radiation. Reaching the diffraction limit for X-rays emitted from a storage ring (approximately 10 pm rad) is highly desirable from a scientific perspective: not only would it bring X-ray microscopy to its limit, but material structure could be determined with much less X-ray damage and fast chemical reactions could be studied in situ. Currently, the electron beam travelling in a storage ring dilutes the X-ray emittance by orders of magnitude. Because this quantity determines the brilliance of the X-ray beam, reaching the X-ray diffraction limit is a case of reducing the electron-beam emittance as far as possible.

The emittance is defined as Cq*E2/N3, where Cq is the ring magnet-lattice constant, E is the electron energy and N is the number of dipole magnets. It has two components: horizontal (given by the magnet lattice and electron energy) and vertical (which is mainly caused by coupling from the horizontal emittance). While the vertical emittance is, in principle, controllable and small compared with the horizontal emittance, the latter has to be minimised by choosing an optimised magnet lattice with a large number of magnet elements.

Because Cq can be brought to the theoretical minimum emittance limit and E is given by the desired spectral range of the X-rays, the only parameter remaining with which we can decrease the electron-beam emittance is N. Simply increasing the number of achromats to increase N turns out not to be a practical solution, however, because the rings are too big and expensive and/or the electrons tend to be unstable and leave the ring. However, a clever compromise called the multi-bend achromat (MBA), based on compact magnets and vacuum chambers, allows more elements to be incorporated around a storage ring without increasing its diameter, and in principle this design could allow a future storage ring to achieve the diffraction limit.