At the end we found it:

the first beam exiting from the termination of an X-ray waveguide.

All this happened in 1994, when the ESRF was due to start operation. We got to know Stefano Lagomarsino from IESS-CNR in Rome, who had read the article
“Resonance-enhanced X-rays in thin films – A structure probe for membranes and surface-layers”
by J. Wang, M.J. Bedzyk and M. Caffrey in SCIENCE 258 (5083): 775-778 (Oct. 30, 1992).
Jin Wang had found the resonance in a layer thickness of roughly 100 nm. And as Stefano had learned from us, that we could sputter thin films in carbon of similar thickness, he posed this simple question:
Shouldn’t the resonance enhanced X-rays travel in and leave the thin film at its end, where they would be confined in the waveguiding layer and then form a submicron source for X-ray microscopy?
We could not find a plausible argument for the answer NO. Thus we immediately decided to give it a try! In order to see, whether our sputtering technique would provide suitable thin films, we (S. Di Fonzo, B.R. Müller, W. Jark from Sincrotrone and S. Lagomarsino, A. Cedola from CNR (Rome)) first tried to repeat the experiment of Wang et al. We were successful, even though we used only window glass as the substrate, we sputtered carbon and we had available only a standard laboratory X-ray source, and not an undulator beamline. Then we wrote a proposal for beamtime at the ESRF (MI50: X-ray micro-beam production by resonance effect in total reflection region) for the September 1994 deadline. We hadn’t seen any waveguided beam in the laboratory setup, but we had no doubts anymore about its existence, and Christian Riekel from the ESRF microfocus beamline at ID13 believed us.
We were finally awarded 9 shifts of beamtime at ID13 (Microfocus) starting already March 15th, 1995. This did not leave much time for the waveguide preparation and now we were in trouble: a maternity leave and the expiration of a fellowship had reduced the Sincrotrone team to the minimum of a single collaborator.
And we still had to prepare the final waveguides in our multilayer sputter facility at ELETTRA. Now we started to ask questions like:
But would the flattest and smoothest pieces of cheap float glass (i.e. window glass, category: consumables), which we could find in the laboratory, be appropriate waveguide substrates?
They needed to, as they were readily available and we did not have the time and the money for ordering specially fabricated substrates.
What does the waveguide exit need to look like?
Similarly simple answer: we just broke the glass slide after the deposition as we did not have any tools for the preparation of the “perfect” waveguide termination.

At the end we had a few samples to take to the ESRF with guiding layers between 80 and 110 nm thick. Here we (Stefano Lagomarsino, Alessia Cedola and Werner Jark) met essentially skeptical colleagues.
In the absence of a guesthouse we had to sleep in downtown Grenoble, if we wanted to. We reduced this option to a minimum, sometimes just to make breakfast, which we couldn’t have on-site at that time.
We started with assembling, disassembling and reassembling of goniometers, tilt stages etc. And we only had a pin-diode for finding the exiting beam. We choose 13 keV photon energy at ID13. Good luck! Strategy: efficient intensity feeding to waveguided modes requires minimum sample reflectance. First let us find this angular position with the pin-diode while we illuminate the end of the sample. Then let us block upstream of the waveguide exit the reflected beam and the beam passing over the sample without blocking the beam, which was expected to exit from the waveguide! This latter job was supposed to be done by a drop of silver print the closest possible to the waveguide termination. Then once the respective reflectivity minimum is found one would just move the waveguide laterally slightly from a no-drop position to the drop position. Now the diode should register only the exiting waveguided beam. This is the theory, in practice the diode insistently continues to show only noise for the first 71.5 hours of 72 scheduled hours!
It is almost 07:00 a.m. on March 18th, Saturday, i.e. the end of the beamtime assigned to us. Another sleepless night has passed. Frustration comes up: but why doesn’t the waveguide do its job, i.e. guide the wave. Where did we make a mistake? In the choice of the substrate or in the preparation of the termination? Nobody has an idea. So let’s make the last breakfast before returning home. Only Stefano does not want to surrender and does not join us. Great decision! He finally moves the right motor and there is the beam!! But what to do? No way to inform us, as we did not yet work with mobile phones and SMS. So upon our return we unexpectedly find him very excited with no sign of tiredness and frustration left. He shows us, how to find the beam again in just a few minutes. Nobody wants to leave anymore! We obviously did everything right! So one can make an X-ray waveguide almost for free.
Fortunately the next users will start only on Monday. So we can continue our experiment. Our flight tickets get changed, the hotel has rooms for another night. And from now on we always find the exiting beam within one hour after the installation of a new waveguide. Finally we can even identify the beam in pictures, which were read with a frame grabber card from a home-made X-ray camera. These pictures are unpublishable and we have to recognize that much more photon flux would have passed through a simple slit with the same aperture. But who cares, we produced an X-ray beam (1 Å wavelength) with unprecedented size of 100 nm and nobody has a slit with the same dimension! No doubt that we will do better next time! And we did!