With the ever increasing quality of the source, i.e. smaller source size and beam divergence, the tolerances for the optical elements adapting the beam properties to the particular requirements of the experiments become more and more severe. In particular, for a certain class of experimental techniques such as topography or tomography, it is crucial to avoid any kind of wavefront perturbation that would decrease the degree of coherence of the X-ray beams. Present-day mirror surfaces degrade the coherence properties of the source [1]. Non-ideal surfaces limit the optics performances for microfocusing experiments when attempting to decrease the focal spot size to the sub-micrometre range. At the ESRF, one strategy to microfocus X-rays is to bend flat mirror substrates [2] and presently a 1 µm spot size is achieved with this technique.

It is therefore important to continuously improve the surface quality of mirrors and multilayer substrates, i.e. the microroughness on a short spatial scale (~ µm) and the slope errors and shape errors on the medium and long scale (mm to m). While the first should stay below 0.1 nm, the second and the third should not exceed 0.1 µrad and 1 nm, respectively (rms values). The best commercially available silicon mirrors presently exhibit 0.1 nm microroughness, 10 nm shape and 1 µrad slope error for 300 mm long flats, therefore a tenfold improvement in quality is needed.

We have recently succeeded in optimising our chemical-mechanical polishing process to achieve similar results. The art consists firstly of producing an excellent figure (low slope and shape errors) with a "hard" tool and then to preserve this flatness as far as possible when finishing with a "soft" tool to decrease the microroughness. The proper choice of the polishing pad material, the pressure, the rotation speed, the type and the supply rate of the slurry are all important parameters that must be optimised to converge to an acceptable result within a reasonable time. We can now polish a 100 mm diameter silicon specimen to a microroughness of about 0.06 nm (see Figure 127), a slope error around 1 µrad and a shape error around 10 nm in less than 8 hours.

The remaining challenge is to decrease the slope and shape errors by roughly one order of magnitude. To achieve this goal, ion beam figuring is presently studied by the Optics Group in collaboration with industry where a superpolished silicon substrate can be ion beam milled. A 300 mm long substrate was micromachined or, rather, nanomachined by the company Carl Zeiss (Oberkochen), according to a surface height error map determined beforehand by precise metrology. Figure 128 shows the surface height error of the substrate before (blue, standard deviation of 18 nm over 275 mm), and after ion beam machining (red and green, standard deviation of 1.2 nm). The corresponding slope errors were 0.88 and 0.11 µrad respectively, an improvement by a factor of eight. Significantly, the microroughness remained unaffected by the nanomachining process at a level of 0.08 nm rms!

The high-precision metrology used was the X-ray long trace profiler developed earlier on the ESRF's BM5 beamline [3]. Its precision limitations are depicted by the discrepancy between the red and green curves (Figure 128). Precision and accuracy are estimated to be around 0.3 nm and improvements are needed to provide an error map leading to a further quality gain by a second ion beam smoothing. In conclusion, with this remarkable improvement we have taken an important step forward towards our ambition of full coherence preservation by reflective optics, opening the way to new applications both for third and for fourth generation X-ray sources.

[1] I. Schelokov, O. Hignette, I. Snigireva, A. Snigirev, A. Souvorov, SPIE Proceedings, 2805, 282-292 (1996).
[2] ESRF Highlights 1997/1998, p 102.
[3] O. Hignette, A. Freund, E. Chinchio, SPIE Proceedings, 3152, 277-286 (1997).

O. Hignette (a), E. Chinchio (a), A. Rommeveaux (a), P. Villermet (a), A.K. Freund (a), H. Handschuh (b).

(a) ESRF
(b) Carl Zeiss Oberkochen