Radiography is now 100 years old. Very recently, it took a new direction when it was shown at the ESRF that the coherence of a third generation facility beam makes it possible, in an easy way, to break the usual restriction to absorption and to visualise phase features. The ESRF X-ray beam has:

  • a high spatial coherence, i.e. the size and the divergence of the beam are very small,
  • a good temporal coherence after monochromatisation.

These characteristics of the beam, due to its exceptional brilliance, allow new techniques to be developed in the X-ray field:

  • phase-contrast imaging and tomography,
  • in-line holography,
  • photon correlation spectroscopy.

Some examples are given below, demonstrating these techniques used with various materials and biological samples.



Characterisation of the beam coherence by a new interferometric technique



The coming of new interferometric techniques using the high degree of coherence of the ESRF X-ray beam like speckle spectroscopy, phase-contrast imaging, and holography created a demand for a quantitative measurement of the beam coherence. In classical optics the coherence can be measured by a double-slit experiment or by means of other wavefront-splitting interferometers like Fresnel's double mirror or Fresnel's double prism. In the high energy X-ray domain, it is difficult to get a clean interference pattern with slits because these should be very thick with regard to the small absorption which requires a lot of effort to achieve a good alignment. Mirrors, on the other hand, when used with hard X-rays, are far from perfect in terms of surface roughness. Thus the mirrors themselves will influence the beam coherence. Nevertheless, an interferometer based on mirrors was built and successfully operated. The diffraction patterns, however, are fairly complicated and difficult to interpret.A very simple interferometric technique is applied to characterise the transversal beam coherence of a monochromatic beam by recording the in-line hologram of a calibrated fibre with known radius. This set-up avoids all additional optical elements but the monochromator. Reflection at a perfect crystal apparently does not deteriorate the beam coherence. The recorded hologram has a system of interference fringes of decreasing spacing with increasing distance from the fibre shadow. The smallest spacing is a function of the source size and the ratio of the source-to-fibre distance to the fibre-to-detector distance (magnification factor). By measuring the smallest fringe-spacing we can therefore calculate the source size. The source size is then related to the lateral coherence of the beam by the van Cittert-Zernicke theorem.

On the optics beamline BM5, 10 to 12 fringes were measured in the vertical and 5 to 6 fringes in the horizontal direction with a smallest spacing in the vertical direction of 1.8 µm (Figure 1). The hologram was recorded on a high resolution Kodak film, which has a spatial resolution of about l µm. The fibre-to-film distance was r1 = 1 m and the source-to-fibre distance r0 = 40 m. Thus we find a source size of about 80 µm vertically. This value is in good agreement with independently-known source characteristics for 0.5 % coupling of the electron beam.

With the same set-up the influence of optical elements on the coherence properties of the beam can be measured. To characterise the influence on the coherence of an X-ray mirror, the fringe pattern of the fibre was recorded with and without the mirror in place. The rms roughness of the mirror measured with an optical interferometer was found to be less than 0.3 nm and the slope error was less than 3 µrad. With the mirror in front of the fibre only the first 5 fringes were detected vertically. To draw preliminary conclusions from this experiment, it can be said that state-of-the-art mirrors can degrade the coherence of the beam and higher quality mirrors are needed if the high degree of coherence of the source needs to be preserved.

Photo: On the Optics beamline (BM5)




[1] A. Snigirev (a), Optics for High-Brightness Synchrotron Radiation Beamlines II, L.E. Berman, J. Arthur, Eds., Proc. SPIE 2856 (1996)

[2] I. Schelokov (b), O. Hignette (a), C. Raven (a), A. Snigirev (a), I. Snigireva (a), A. Souvorov (a), Multilayer and Grazing Incidence X-ray/EUV Optics III, R.H. Hoover, A. B. C. Walker, Eds, Proc. SPIE 2805, 282-292 (1996)

[3] A. Snigirev (a), V. Kohn, (c), C. Raven (a), I. Snigireva (a), A. Souvorov (a), to be published

(a) ESRF

(b) Institute of Microelectronics Technology RAS (Russia)

(c) Kurchatov Institute (Russia)




Phase-contrast imaging and computed tomography



As shown a year ago and as exemplified now with the better quality of the source (Figure 2), the coherence properties of the ESRF X-ray beam can easily be used to increase the contrast in imaging methods like radiography and tomography. Depending on the energy of the X-rays and the geometry of the set-up, phase-contrast imaging can be developed from a simple edge-enhancement method of absorption radiograms by recording the images at a small distance behind the objects to in-line holography of transparent samples by increasing the recording distance.

Soon after the first phase-contrast images were recorded it became obvious that the contrast improvements resulting from the coherent illumination should also be applicable to tomography. As an example, 61 images of a 100 µm boron fibre in a 60 keV beam were recorded at the High-Energy beamline. The fibre has a tungsten core, which was etched from one side. A cross-sectional tomographic reconstruction of a fibre with and without core are shown in Figure 3. The spatial resolution is limited by the CCD camera to about 8 µm. Even with a non-adapted tomographic algorithm originally developed for absorption tomography the core is clearly visible, in both cases, with the highly absorbing tungsten core and with the hollow core. Efforts have since been made to improve the reconstruction quality by implementing an algorithm dedicated to the special nature of the phase-contrast images.




[1] A. Snigirev (a), I. Snigireva (a), V. Kohn (b), S. Kuznetsov (c), I. Schelokov (c), Rev. Sci. Instrum. , 66(12), 5486-5492 (1995)

[2] A. Snigirev (a), I. Snigireva (a), V. Kohn (b), S. Kuznetsov (c), NIM, A370, 634-640 (1996)

[3] C. Raven (a), A. Snigirev (a), I. Snigireva (a), P. Spanne (a), A. Souvorov (a), V. Kohn (b), Appl. Phys. Lett., 69(13) (1996)

(a) ESRF

(b) Institute of Microelectronics Technology RAS (Russia)

(c) Kurchatov Institute (Russia)


Phase-contrast computed microtomography of biological materials using in-line holographic techniques



It is of particular interest to apply phase contrast to organic materials, since for such samples radiation damage resulting from high absorbed doses often limits both achievable contrast and spatial resolution. One data collection strategy which can be used to perform three-dimensional imaging and where phase contrast has the potential to contribute new image information is that of computed microtomography.

A simple «in-line holography» set-up, using only a monochromator before and an area detector behind the sample, can be used to generate images exhibiting contrast due to phase shifts in the samples. With this method, abrupt changes in refractive index, occurring at interfaces in the sample that are tangential to the X-ray beam propagation direction, cause interference fringes to appear if the detector is placed 0.5-2 m behind the sample. If the X-ray energy is sufficiently high for absorption to be small or negligible, edge-enhanced projection images result. Extension of the technique to computed tomography at first sight seems to violate a fundamental requirement in that contrast is recorded only when the X-ray beam is close to parallel with the interface. The success of the imaging of the boron fibre can be explained by an analogy with absorption tomography of a sample where the interfaces between materials are modelled as thin shells. In this case the attenuation is negligible except for beam propagation directions close to parallel with the interface.

The claims for improved contrast with phase shift detection as compared to absorption imaging with the same or smaller absorbed dose in the sample are particularly interesting for medical imaging and microscopy of biopsies and other tissue specimens. An experiment imaging a segment of a formalin-fixed human coronary artery with a fatal plaque and thrombosis was performed to demonstrate the potential of phase contrast in medical and biomedical imaging of wet tissues. The sample was obtained in a collaboration with Pathologisches Institut, University of Bern (T. Schaffner and J.A. Laissue) with the purpose of investigating the feasibility of three-dimensional characterisation of the morphology of plaques in coronary artery specimens using computed microtomography techniques. A data set suitable for tomographic three-dimensional reconstruction of the artery was obtained by making 1125 X-ray projection images distributed over 180 degrees around the sample. 25 keV X-rays monochromatised with a Si(111) Bragg monochromator were used in the experiment. At this energy, which is more than twice as high as the optimum energy for absorption imaging, only the calcified plaque shows any substantial absorption contrast. Each projection image consisted of 1024 x 1024 pixels. The projection images were acquired using the FRELON fast CCD camera, developed by the ESRF Detector Group, in combination with a high spatial resolution scintillator and light optics. The resulting pixel size was 6.3 x 6.3 µm2 .The frame readout time for this camera is at present 0.2 s, allowing very rapid generation of large data sets for three-dimensional imaging.

A tomographic reconstruction of the artery clearly demonstrated the high information content that can be obtained with the in-line holography set-up when applied to computed microtomography of tissue samples (Spanne et al., to be published). This was evident even though the reconstructed image exhibited ring artefacts caused by imperfections of the source and thermal instability of the monochromator crystal. Especially promising for further development is the fact that the artery wall is clearly delineated as a thin line in the image, demonstrating that interfaces between wet tissue components with small or minimal absorption differences can be imaged with high contrast. Inside the artery the tomograms exhibited plenty of details as well as high contrast for the edge of the plaque and the microcalcifications. Both types of contrast prove the importance of evaluating phase-contrast imaging for both biomedical and medical imaging applications.

In Figure 4 a shaded surface rendering of a part of the plaque and microcalcifications is shown.

It will be of course possible to apply phase shift generated contrast in many other imaging applications in various fields of research where the atomic composition differences between sample constituents do not allow efficient absorption contrast imaging. One very well-suited application is the study of liquid-air interfaces in unsaturated soil. The curvature of these interfaces are of fundamental importance for several problems in soil science and previously they have not been directly observed in realistic soil samples.

Angiography/tomography monochromator during installation on the Medical beamline




[1] C. Raven (a), A. Snigirev (a), I. Snigireva (a), P. Spanne (a), A. Souvorov (a), Proc. National Synchrotron Radiation Instrumentation Conf., Argonne (USA) 18-20 October 1995

[2] C. Raven (a), A. Snigirev (a), I. Snigireva (a), P. Spanne (a), A. Souvorov (a), V. Kohn (b), Appl. Phys. Lett., 69(13) (1996)

(a) ESRF

(b) Institute of Microelectronics Technology RAS (Russia)



Application of phase-contrast imaging to materials science



In phase-contrast imaging, the observation of the local variations of the optical path-length is related to Fresnel diffraction, or, in analogy with electron microscopy, to defocusing, with the peculiarity that in the X-ray case there is no "focused" position, except at the sample itself. It can also be called in-line holography. The results published definitely show that good images feature rather sharp phase variations with a resolution approaching 1 µm if the detector is a high-resolution film, or 8 µm with a CCD camera. Images can be obtained with modest exposure times of about one minute on film. They show in addition that two "regimes" can be distinguished:

1. an "edge detection" regime, where each border is imaged independently, but which does not allow the measurement of the local phase, and

2. a "holography" regime, where the image of the object is deformed, but which can give access to the local phase when combining the images recorded at different distances in an adapted algorithm.

Consequently, the following steps are, the imple-mentation of such an algorithm to reconstruct the local phase from the recorded images and the application of the "edge detection" version of this technique, combined with computer-assisted tomography as developed for the absorption case, for the three-dimensional investigation of singularities in materials (or biological objects). The present report is concerned with this second aspect, i.e. some of the uses of phase tomography in materials science, and results from a co-operation between the ESRF and several laboratories of the INSA in Lyon, France (CNDRI, CREATIS, GEMPPM).

One of the important topics in materials science is the study of the damage mechanisms. At the present time, these investigations rely on surface observations, performed on loaded samples. But the damage initiation inside materials clearly involves phenomena in a tri-axial stress-strain field, about which there exist practically no experimental data. Damage occurs either inside reinforcement material (fibres or precipitates) or at the borders of the matrix (interfaces). The damage then propagates through a percolation process, the resulting crack being created by the junction between micro-damages. A truly non-invasive, three-dimensional method with high spatial resolution is therefore strongly desirable.

The possibilities of phase-contrast imaging to investigate cracks is indicated by two examples performed at ID19, using 25 keV X-rays and a sample-to-detector distance of 82 cm, on

1. a 2 mm aluminium rod reinforced with a 140 µm SiC fibre with a 50 µm carbon centre and

2. an aluminium alloy containing silicon carbide precipitates.

Both materials were submitted to tensile stress up to the rupture. Figure 5 shows the image of the aluminium rod obtained. The silicon carbide fibre, its carbon centre and porosity in the aluminium are clearly visible, but the main contrast is associated with the three reinforcement fibre cracked zones. Figure 6 shows microtomographic sections of the aluminium alloy with silicon carbide particles obtained through the tomographic reconstruction of 600 images. They correspond to different strain levels. The shape, size and distribution of the SiC particles can be extracted from these images. No difference is noticeable between the initial state and an already drastically deformed "state 1" but the first appearance of the cracks is observable in "state 2" before the actual rupture becomes clearly visible ("state 3").

This type of experiment is under further development at the long (145 m) beamline ID19, where the angular size of the source is particularly small and consequently the transverse coherence length of the beam particularly high.




[1] P. Cloetens (a, b), R. Barret (a), J. Baruchel (a), J.P. Guigay (c), M. Schlenker (c), J. Phys. D: Appl. Phys. 29, 133-145 (1996).

[2] P. Cloetens (a, b), M. Pateyron (a, d), G. Peix (e), J.Y. Buffière (f), J. Baruchel (a), F. Peyrin (a, d) and M. Schlenker (c), to be published

(a) ESRF

(b) EMAT, Univ. of Antwerp (Belgium)

(c) Lab. Louis Néel, CNRS Grenoble (France)

(d) CREATIS, INSA Lyon (France)

(e) CNDRI, INSA Lyon (France)

(f) GEMPPM, INSA Lyon (France)

Phase-contrast tomography

Tomography of objects imaged in a wide beam X-ray interferometer in many different projections generates contrast due to phase shifts in the material rather than due to absorption contrast. The technique has the advantage that for biological objects the damage is reduced, while contrast between different tissues is much higher than applying absorption contrast. The method was originally demonstrated by Momose in Japan and further developed by the group of Bonse at Hasylab. Using the ESRF, results of extremely high quality were obtained. Figure 7 shows the experimental arrangement and Figure 8 a specific example.



F. Beckmann (a), U. Bonse (a), F. Busch (a), O. Günnewig (a), T. Biermann (a), HASYLAB Annual Report II, 691-692 (1995).

(a) Institute of Physics, Univ. of Dortmund (Germany)



Scattering with coherent X-rays

Complex relaxations in disordered systems have been studied successfully by scattering of both visible light and neutrons. Neutron-based techniques (inelastic and quasi-elastic neutron scattering, neutron spin-echo) can probe the dynamic properties of matter at high frequencies from typically equal to 1014 Hz down to about 108 Hz and achieve atomic resolution. Scattering vectors between 0.02 Å-1 and 10 Å-1 are usually accessible in these experiments. Photon Correlation Spectroscopy (PCS) on the other hand can cover the low frequency dynamics ( < 106 Hz) with visible light, but only probes the long wavelength q < 4 x 10-3 Å-1 region in materials not absorbing visible light. The availability of intense coherent hard X-ray beams from third generation synchrotron radiation sources opens up the possibility for correlation spectroscopy experiments with X-rays capable of probing the low frequency dynamics (106Hz to 10-3Hz) in a q range from 10-3 Å-1 up to several Å-1, thus providing atomic resolution. Furthermore it allows the investigation of optically opaque materials.

X-ray Photon Correlation Spectroscopy probes the dynamic properties of matter by analysing the temporal correlations among photons scattered by the material. It requires the sample to be illuminated coherently, implying the need for an intense X-ray beam with sufficient transverse (typically 10 µm) and longitudinal (> 100 Å) coherence lengths. Whereas both requirements are relatively easy to fulfil with laser light, one needs a highly collimated beam in the case of X-rays which can be produced by introducing very small (µm sized) apertures in the beam. Only high brilliance sources like the ESRF can provide sufficient flux under these experimental conditions.

The staff of beamline ID10A (TROIKA) recently succeeded in producing a 1.5 Å coherent X-ray beam of 5.109 photons/sec with a transverse coherence length of 12 µm and a longitudinal coherence length of 120 Å [1]. This was possible by replacing a crystal monochromator by a pair of crossed mirrors and relaxing the bandpath of the coherent X-ray beam to the intrinsic bandwidth (1.3 %) of an undulator harmonic. Longitudinal coherence lengths of the order of 100 Å are sufficient to reach scattering vectors up to about 5 x 10-2 Å-1 in a small-angle scattering geometry. This allows a variety of disordered materials to be studied, in particular soft-condensed matter systems like simple liquids, polymers and colloids. Figure 9 shows a static speckle pattern of an aerogel sample recorded with a 20 µm pixel direct illumination CCD camera [2]. The CCD image shows that individual speckles can clearly be resolved up to the highest q-vector (1.4 x 10-2) covered in this image. The unprecedented high coherent flux permits one to study dynamics in a wide time range (10-3 s to 1000 s), addressing important phenomena such as density fluctuations in liquids, concentration fluctations in polymer blends or the short wavelength dynamics of fluids and colloids. One example is the study of the translational diffusion of colloidal palladium. The sample was a solution of colloidal palladium in glycerol with a volume concentration of 0.3 %. The colloid was coated with a stabilising agent and the sample was a completely opaque, black suspension. Figure 10 shows a time correlation function covering eight decades in time taken at q = 1.59 x 10-3 Å-1 at a temperature of 279 K. The oscillation visible at short times (t = 2.81 µs) is quantified in the lower insert and reflects the velocity of the relativistic electron beam. The slow relaxation at t > 1 s is quantified in the upper insert. It is well described by an exponential decay and corresponds to the translational diffusion of the palladium colloid. This figure clearly demonstrates the feasibility of XPCS for measuring low frequency dynamics in the millisecond regime. The correlation function for a diffusion process is given by g(q, t) = A(q)exp(-2 t)+1, where the relaxation rate = q2D and D is the diffusion constant. The diffusion coefficient determined from the data is D = 5.1 x 10-12 cm2s at 279 K. For a translational diffusion process one expects D = kT / 6 Rh, where k is the Boltzmann constant, T the temperature, the viscosity and Rh the apparent hydrodynamic radius of the diffusing particle. The temperature dependence of the relaxation rate shows that the temperature dependence of the diffusion constant is in fact dominated by the viscosity of the glycerol solvent [3].

These results show that scattering with coherent X-rays and in particular XPCS will have an impact on many areas of statistical physics, such as dynamic critical phenomena, ordering in alloys, charge density waves, liquid crystals, low-dimensional systems, quasicrystals and amorphous systems. XPCS has started to address important phenomena in soft-condensed matter systems such as density fluctuations in liquids, concentration fluctuations in polymer blends or the short wavelength dynamics of fluids and colloids.


Photo: On the Troika beamline (ID10A)



[1] G. Grübel (a), D. Abernathy (a), Proceedings of the "10th ICFA Beam Dynamics Panel Workshop on 4th Generation Light Sources", WGS 2-101, Grenoble (1996).

[2] D. Abernathy (a), G. Grübel (a), S.G.J. Mochrie (b), A. Sandy (c), S. Brauer (c), G.B. Stephenson (d), M. Sutton (e), I. McNulty (c), to be published

[3] T. Thurn-Albrecht (f), W. Steffen (f), A. Patkowski (f), G. Meier (f), E.W. Fischer (f), G. Grübel (a), D. Abernathy (a), to be published

(a) ESRF

(b) Department of Physics, MIT, Cambridge MA (USA)

(c) APS, Argonne National Laboratory, Argonne, IL (USA)

(d) Materials Science Division, Argonne National Laboratory, Argonne, IL (USA)

(e) Department of Physics, McGill University, Montreal (Canada)

(f) MPI für Polymerforschung, Mainz (Germany)