Chapter 3.1. X-ray sources

Ultraviolet (UV) radiation is electromagnetic radiation with a wavelength from 10 to 380 nm, shorter than those of visible light (390 > λ > 700 nm) and infrared radiation (λ > 700 nm) but longer than that of X-rays (λ < 10 nm). Vacuum UV (VUV) is a wavelength region of less than 200 nm which requires a vacuum path because of strong absorption by air. Extreme UV (EUV or XUV) radiation is electromagnetic radiation with a wavelength longer than ∼30 nm, adjacent to soft X-rays (λ < 10 nm). VUVX is used as an integrated terminology for VUV, EUV and X-rays, considering the wide range of wavelengths covered by free-electron lasers (FELs; Section 5), making conventional distinctions by wavelength less essential. Although there are many different light sources for VUV and EUV, a common mechanism is a radiative deactivation process of excited states in the form of atoms or molecules. Applications of VUV light sources extend from laboratory spectrometers probing electronic structure by absorption or photoemission spectroscopy (PES) to industrial use for photolithography, surface modification and sterilization.

The most common VUV light sources for laboratory spectrometers are gas-discharge lamps and excimer (excited molecules) lamps. An excited state induced by an electrical discharge or high-energy electron beam forms molecules temporarily bound to themselves (dimers) or to halogens such as fluorine and chlorine (complexes). The excited compound releases excess energy by undergoing spontaneous or stimulated emission, resulting in a strongly repulsive ground-state molecule which very quickly (on the order of a picosecond) dissociates back into two unbound atoms. Gas-discharge lamps are a family of artificial light sources that generate light by creating an electrical discharge through an ionized gas or a plasma. Such lamps use a noble gas (argon, neon, krypton or xenon) or a mixture of these gases. Photolithography in the deep UV (335–532 nm) utilizes KrF and ArF lasers for wavelengths of 248 and 193 nm (Jain et al., 1982). From the early 1960s until the mid-1980s, mercury–xenon lamps were used in lithography using spectral lines at 436, 405 and 365 nm, respectively. Excimer lamps utilize spontaneous emissions from excimers. Working excimer molecules with wavelengths below 200 nm include ArF (193 nm), KrI (190 nm), ArCl (175 nm), Xe₂ (172 and 175 nm), ArBr (165 nm), F₂ (158 nm), Kr₂ (146 nm), Ar₂ (126 nm) and NeF (108 nm). The intrinsic emission lines from either discharge or laser devices are commercially available and are widely used for spectroscopy and photolithography. In the latter application, as the scaling rule shifts according to Moore's law, the wavelength shifts from 436 nm (high-pressure mercury) down to 157 nm (excimer laser) and is expected to reach 13.5 nm (EUV).

Excimer lasers, discharge-produced plasma (DPP) and laser-produced plasma (LPP) are used as EUV light sources. DPP creates plasma by discharge between electrodes, while LPP uses a YAG laser or a CO₂ laser, which irradiates a metal droplet and creates plasma. Comparing LPP and DPP, the former can provide higher power focusing onto a target, producing a high-brightness (10–100 TW cm⁻²) pulse with a high repetition rate. EUV and X-ray generation by LPP has been reviewed by Kauffman (1991). Next-generation photolithography with a 5 nm node requires 13.5 nm light. Currently, EUV light sources are limited to plasma sources where a dense plasma of multicharged cations (tin) is neutralized by electrons, but the power density is limited because of the unstable population (Tao et al., 2005; Coons et al., 2010). Because of the extremely low conversion efficiency (∼0.02%), the currently available power of LPP light sources is 10–20 W on average, compared with the industry requirement of 250 W. In laser-assisted discharge plasma (LADP), plasma is created by large-current short pulse discharge of the target (thin tin film), which allows higher powers of up to 51 W by optimizing the laser pulse specifications (Corthout, 2007).

Short pulses from plasma produced by focusing pulsed laser beams onto solid targets are an alternative X-ray source. LPP X-ray sources are extensively used in a wide range of advanced scientific and technological applications such as X-ray imaging, photolithography, dynamic studies of the mechanical properties of materials and X-ray time-resolved radiography. The mechanism of X-ray pulse generation is via fast electrons within a plasma formed by a laser pulse, which allows femtosecond short pulses beyond those of synchtron-radiation (SR) X-rays (100 ps). Laser-based X-ray plasma sources are promising as short-pulse X-ray sources for ultrafast X-ray spectroscopy. Using a femtosecond pulsed infrared laser and a water target, a microfocus (75 µm) spot with ∼10⁶ photons s⁻¹ after focusing a polycapillary optics is available (Miaja-Avila et al., 2015).

Von Laue and Knipping obtained the first diffraction pattern of a crystal using X-rays in 1912. X-ray diffraction (XRD) became a popular structural tool, which has greatly contributed to modern society from materials science to the medical industry. Structural biology originated with the first structure determination of DNA by Watson and Crick in 1953. X-ray tubes were used as a light source for powder and single-crystal X-ray diffraction and scattering experiments until SR facilities became available. Conventionally, X-rays are obtained by colliding accelerated hot electrons with a target (anode) material such as chromium, iron, cobalt, copper or molybdenum, producing characteristic X-rays with intrinsic wavelengths (2.291, 1.937, 1.790, 1.542 and 0.711 Å, respectively) and energies ranging from 5.4 to 17.4 keV. The X-ray spectrum from X-ray tubes consists of a sharp spike due to radiative decay of core excitation to K shells called characteristic X-rays and a broad background (Bremsstrahlung) known as white X-rays or continuous X-rays. For crystallography, a copper target (∼8 keV) is widely used, while for shorter wavelength X-rays heavy elements such as tungsten (∼58.9 keV) or molybdenum are used. However, multiwavelength anomalous dispersion (MAD) experiments require tunability over a wide range of up to 100 keV, which is hardly covered by changing the target material. Thus, the greatest disadvantage of electron-impact sources is the limitation in tunability and brightness, in contrast to their advantage in compactness that is convenient for laboratory applications. The development of the X-ray tube is a history of related patents. More than 19 000 patents were registered around the world up to 2013, according to searches of the European Patent Office databases. The maximum power of an X-ray tube has long been limited by the heat load of the anodes, i.e. the heat transfer of the target material on the anode. Extensive efforts to improve the efficiency of anode cooling have been carried out, which led to X-ray tubes becoming a stable and compact light source for laboratory X-ray experiments. Further, high-power X-ray diffraction experiments with a rotating-anode X-ray source allowed measurements with small crystals or thin films. On the other hand, high brightness, i.e. a micro X-ray beam from a microfocus X-ray tube and/or using focusing optics, became available. Some X-ray analysis methods (for example non­destructive testing and three-dimensional microtomography) need very high-resolution images and therefore require X-ray tubes that can generate very small focal spot sizes, typically below 50 µm in diameter. Microfocus X-ray sources operate with focus spots in the range 5–20 µm in size and with power densities in the range 0.4–0.8 W µm⁻², depending on the anode material. Although the anode power is limited to <10 W to avoid target melting (Grider et al., 1986), a high power density in the range 67–100 kW mm⁻² is available by means of a rotating anode and a metal-jet anode microfocus X-ray source (Otendal et al., 2008).

Synchrotron radiation (SR) is electromagnetic radiation that is emitted from charged particles such as electrons and positrons with a velocity close to that of light. Reviews of SR from the viewpoint of a light source (Winick & Doniach, 1980; Koch, 1983; Kim, 1989; Clarke, 2004) describe the principles and applications of SR well. A comprehensive review of X-radiation from laser plasmas (Kauffman, 1991) would be helpful to understand advanced X-ray light sources. Historically, emission from a centripetally accelerated charge has been a serious problem in accelerator physics because of radiative energy loss. Conventional nonrelativisitic dipole-like radiation with an angular dependence of sin²θ, where θ is the angle between the emission direction and the acceleration vector, changes its emitting pattern to a directional distribution as the velocity approaches that of light. The basic theoretical description of SR was made by Schwinger (1949). Equations and graphs were later presented in a sophisticated fashion by Jackson (1975). The first observation of artificial SR from accelerators was made by researchers at General Electric (Blewett, 1998). Experimental verification of SR was made by many researchers at electron synchrotron (ES) rings at Cornell University (300 MeV ES), National Bureau of Standards (NBS; 180 MeV ES) and DESY (6.6 GeV ES). As soon as the superior characteristics of SR, for example its broad and intense spectrum from microwave to X-ray, high brilliance [reported here in photons s⁻¹ mm⁻² mrad⁻² (0.1% bandwidth)], polarization and pulsed time structure, were recognized SR captured enthusiastic popularity. Initial studies took place at the so-called first-generation facilities at Moscow University (0.56 GeV), DESY (6.6 GeV), INLF, University of Rome (1.2 GeV), University of Bonn (0.5 GeV), NBS (0.18 GeV), INS, University of Tokyo (1.3 GeV) and Cornell University (0.33 GeV). In the 1980s, dedicated rings (second-generation rings) were either newly constructed or converted from storage rings for high-energy physics. In addition to the converted rings SURF-1 (NBS), ACO (LURE), DCI (LURE), ADONE (INLF), VEPP-II and VEPP-III (Novosibirsk), NINA (Daresbury), TANTALUS-I (Wisconsin) and BONN-ES (Bonn University), new rings were constructed: SOR-RING (ISSP), SRS (Daresbury), TERAS (ETL), PF (KEK), NSLS-UV (BNL), UV-SOR (IMS), NSLS-X (BNL), Aladdin (PSL) and Super ACO (LURE). Converted second-generation rings were based on the FODO lattice, while DBA (double-bend achromat) lattices, called Chasman–Green lattices, or triple-bend achromat (TBA) lattices later became dominant.

The early SR research in the 1970s mostly comprised spectroscopic studies using VUV rings (<1 GeV). Crystallo­graphers soon became interested in hard X-ray rings (>1 GeV) as light sources and the dedicated X-ray facilities at SRS (Daresbury), NSLS (BNL) and Photon Factory (KEK) commissioned during 1982–1986 (second-generation rings) led to successful applications in X-ray spectroscopies (EXAFS and XANES) and crystallography. Researchers soon paid keen attention to insertion devices (undulators and wigglers) as more intense light sources. The concept of an insertion device (a periodic structure of magnet arrays) dates back to the historical paper by Ginzberg (1947). The first undulator with 50 poles was successfully installed at SSRL in 1953 (Motz et al., 1953) and two undulators were installed in the Soviet Union. A successful FEL experiment using a double-helix superconducting (SC) coil at SSRL led to the installation of an SC undulator at ACO. In 1953, Halbach and Winick proposed a prototype of a standard undulator equipped with an array of rare-earth (SmCo) permanent magnets (Halbach et al., 1983). The superior properties of undulators such as their tunablility with high brilliance, controllable polarization and high coherency had been realized by the SR community by 1985. On the other hand, the successful results with the first wiggler installed at SPEAR (SSRL) led to the proposal of a 5 GeV ring with all-wiggler radiation, which formed the concept for new rings with high ring energy (>5 GeV), low emittance (∼5 nm) and many straight sections. Because electrons circulating in a ring which lose their energy by emitting light are fed energy in a radiofrequency (RF) cavity, keeping the energy of the electrons constant, synchrotron radiation has a pulsed time (bunch) structure. Therefore, the time interval (typically a few nanoseconds) between bunches should be equal to the RF, which is elongated by reducing the number of bunches, i.e. the shortest and longest repetition rates, by a uniform and a single bunch filling, respectively. The bunch lengths are of the order of millimetres, which gives a pulse width of around tens of picoseconds. At third-generation storage rings short X-ray pulses with a time resolution of 100 ps are now available, but going beyond subpicosecond pulses to pulses such as 100 fs will only be possible with XFEL facilities.

Although wiggler radiation is not coherent, it was useful for shifting the critical energy (E_c) towards higher energy as a wavelength shifter or simply enhancing the photon flux as a multipole wiggler (MPW). As a high magnetic field or shortening a magnet gap can expand the high-energy limit, two approaches, i.e. SC magnet or in-vacuum magnet arrays (Yamamoto et al., 1992), are available. SC wigglers were installed at Photon Factory, PLS, CAMD, VEPP-2M, New SUBARU and Spring-8 utilizing high magnetic fields of up to 10.3 T. An SC magnet as a bending magnet (superbend) was also a promising wavelength-shifter device, as shown at ALS. Various types of undulators and elliptical polarization wigglers were also designed and built to control circular polarization in the 1980s (Kim, 1984; Elleaume, 1990). As an undulator with a period of a few centimetres provides X-rays with nanometre wavelength for GeV electrons, >5 GeV rings with a large circumference (∼1000 m) were proposed. As none of the existing rings met these requirements, new (6–8 GeV) rings were constructed in Europe (ESRF), the USA (APS) and Japan (SPring-8) in the late 1980s and early 1990s. These third-generation rings with a high brilliance of 10²¹ in a typical X-ray energy range naturally led to enormous outputs in crystallo­graphy and spectroscopic sciences and also opened up new experiments, for example imaging, time-resolved experiments and photon-in photon-out experiments.

Together with the progress in SR sciences, researchers realized that the major trend at light sources should be brilliance and short pulses. Imaging and time-dependent experiments depend on brilliance and short pulses, respectively. Higher brilliance and shorter pulses allow researchers to study smaller sized samples with a narrower time window. Researchers therefore began to seek a novel light source with low emittance and a short pulse structure. This trend opened up the third chapter of SR history, the fourth-generation light sources, which include FELs, energy-recovery linacs (ERLs) and low-emittance rings (LSRs), together with a novel multi-bend achromat (MBA) lattice used in the so-called third+ generation. Linac-based FELs potentially have the desired parameters in peak brilliance, coherence and short pulses free from the limitations of storage rings arising from energy and momentum perturbations due to the emission of SR, so the construction of soft X-ray and hard X-ray facilities took place ahead of other sources. The successful applications of XFELs quickly produced a number of scientific outputs that were not attainable using storage rings. On the other hand, storage rings had also advanced using advanced lattice (MBA) and short-gap undulators so that intermediate-energy (∼3 GeV) rings could provide a high brilliance that is comparable with third-generation rings. These compact modern rings are called third+-generation rings. As third+-generation rings are less expensive, many new facilities were rapidly constructed around the world. Further, upgrades of the third-generation rings and newly constructed LSRs were able to reach emittances as low as 0.1–5 nm and brilliances of >10²² close to the diffraction limit due to their novel lattice design, i.e. hybrid MBA (HMBA). It is difficult to describe the future of this rapidly growing science: R&D efforts at fourth-generation light sources chasing brilliance and short pulses will not slow down, but complementary light sources may also be used as demand is still growing and different spectral specifications would easily find applications.

An FEL is a type of laser with a lasing medium that consists of very high speed electrons moving freely through a magnetic structure; hence the term free-electron laser (Motz et al., 1953; Madey, 1971). FELs are characterized by tunability over a wider frequency range than other lasers, ranging in wavelength from microwaves through terahertz radiation and infrared to the visible, ultraviolet and X-ray. The beam passes through an undulator, a periodic array of magnets with alternating poles across the beam path. When an external laser is used or the SR radiation is sufficiently strong, the transverse electric field of the radiation strongly interacts with the transverse electron current created by the sinusoidal wiggling motion, causing some electrons to gain and others to lose energy to the optical field, influenced by the ponderomotive force. The FEL radiation intensity grows, causing additional microbunching of the electrons, which continue to radiate in phase.

The first operation of an FEL (Deacon et al., 1977) was lasing at 3.6 µm, followed by lasing at a visible wavelength, 488 nm, and in the UV region. For X-ray FELs (XFELs), the amplification is due to the principle of self-amplified spontaneous emission (SASE; Bonifacio et al., 1984; Kim, 1986). At the time of writing (2016), XFEL facilities in operation with wavelengths below 0.1 nm are LCLS (SLAC; Emma et al., 2010) and SACLA (SPring-8), but more XFEL facilities are under construction, such as the European XFEL, PAL FEL and PSI FEL. Soft X-ray FEL facilities up to 6 nm are also in operation: FLASH (DESY) and FERMI (Elettra). XFEL sources produce X-ray pulses that are more than a billion times brighter than the most intense SR source. Such extremely intense X-ray pulses, along with the appearance of the serial femtosecond X-ray crystallography (SFX) technique, has enabled radiation damage-free protein crystallography for nanocrystals and microcrystals. The high peak brilliance X-ray pulses from an XFEL are destructive, but single-shot X-ray diffraction patterns recorded by a CCD allows serial femtosecond protein crystallography, which makes it possible to obtain X-ray diffraction patterns of excited states or intermediate states in a pump–probe mode or a stopped-flow mode (Chapman et al., 2011). In an energy-recovery linac (ERL), electrons from the photocathode are bent by a bending-magnet array into a superconducting linac and accelerated to full energy in a single pass. The accelerated electrons are then guided around a one-turn lattice, providing high-brilliance X-ray short pulses with high coherence. An ERL is characterized by high-brilliance, high-coherence and short-duration pulses which are not available at SR rings. Although a 5 GeV ERL has the potential to provide nanometre-sized short X-ray pulses (<100 fs) with an average brilliance of 10²³ in the hard X-ray range (<40 keV), only prototypes or test machines have been constructed at the time of writing.

Higher harmonic generation (HHG) by focusing a femtosecond laser onto a gas can provide coherent EUV light, which can be downsized to a cavity-free tabletop device using a local field enhancement induced by resonant plasmons (Kim et al., 2008). Accelerators (FELs or optical klystrons in storage rings) are also used to amplify femtosecond seed pulses (λ < 10 nm) produced by harmonic generation in gas by ultrafast laser illumination (Prazeres et al., 1991). More recent studies using a mid-infrared laser demonstrated that HHG can produce intensely coherent X-rays by enabling X-ray waves to constructively interfere, stepping forward to a tabletop X-ray light source (Popmintchev et al., 2012). R&D efforts and feasibility studies for tabletop short-pulse X-ray light sources based on different approaches are in progress. One of these approaches is based on inverse Compton scattering, which can produce highly directional X-rays (gamma rays) by colliding accelerated electrons with photons (Graves et al., 2014). A tunable microwave undulator based on Thomson scattering utilizing microwaves instead of a periodic magnet array is also a promising approach to a tabletop free-electron laser (Tantawi et al., 2014). Another approach to a tabletop short-pulse X-ray source is plasma wakefield acceleration of electrons, which enables a device of a few centimetres in length to accelerate electrons to the range of a few GeV (Leemans et al., 2014). The average spectral brightness of various free-electron lasers (FELs) in shown in Fig. 1 compared with a selection of modern synchrotrons (envelopes of undulator radiation), high-harmonic sources (HHG), inverse Compton scattering (ICS) and electron-impact sources (X-ray tubes, rotating-anode and metal-jet anode microfocus X-ray sources).

Figure 1 Average spectral brightness of various free-electron lasers (FELs) compared with a selection of modern synchrotrons (envelopes of undulator radiation), high-harmonic sources (HHG), inverse Compton scattering (ICS) and electron-impact sources (X-ray tubes, rotating-anode and metal-jet anode microfocus X-ray sources). HHG and ICS plots were taken from the literature (Chapman, 2014). The indicated two-order-of-magnitude ranges show the approximate variation that can be expected among stationary-anode tubes (lower end of the range), rotating-anode tubes (middle of the range) and rotating-anode tubes with microfocusing (upper end of the range), taken from Figs. 2–9 in Kim (2009).

Note: The proofs for this chapter were kindly checked by Professor C. T. Chantler ([email protected]), to whom correspondence may be addressed.